Author: Manish Goel, FCA

  • Regression to the Mean: Francis Galton’s 1886 Discovery and Why the Long-Term Equity Investor Must Tell Skill From Statistical Gravity

    Regression to the Mean: Francis Galton’s 1886 Discovery and Why the Long-Term Equity Investor Must Tell Skill From Statistical Gravity

    AFTERNOON EDITION — Mental Models

    In the summer of 1885, Francis Galton stood before the Royal Anthropological Institute in London with a curious finding. He had measured the heights of 928 adult children and 205 of their parents, and noticed that the offspring of unusually tall parents tended to be tall — but less tall than the parents. The offspring of unusually short parents tended to be short — but less short. Each generation, in other words, drifted back toward the average of the population. He published the lecture the following year in the Institute’s Journal under a title he chose with care: “Regression towards Mediocrity in Hereditary Stature.”

    For Galton, the word “mediocrity” was statistical, not pejorative. He had stumbled on the first mathematical description of a phenomenon that now sits under nearly every empirical exercise in finance, medicine, sports, education, and public policy: when a measurement is the sum of a stable component and a noisy one, an extreme observation in the first period will be followed, on average, by a less extreme observation in the second. The drift is not punishment for excellence. It is the arithmetic of noise.

    1. The model — Galton 1886, in his own words

    The canonical citation is Francis Galton, “Regression towards Mediocrity in Hereditary Stature,” Journal of the Anthropological Institute of Great Britain and Ireland, vol. 15 (1886), pp. 246–263. Galton’s own one-sentence form, set out on page 252, is the cleanest definition we have: the deviation of the offspring from the population average is, on average, two-thirds of the corresponding deviation of the parents. He called that coefficient — the two-thirds — the “ratio of regression.” When modern statisticians later rebadged the linear technique he had invented as “regression analysis,” they preserved his accidental terminology long after the population-genetics origin had faded.

    A stricter contemporary form, due to Karl Pearson’s 1903 generalisation in Biometrika, runs as follows. If X and Y are correlated with coefficient r, and both are standardised to zero mean and unit variance, then the conditional expectation of Y given X is exactly r · X. Whenever the absolute value of r is less than one — which is to say, whenever the two variables are not perfectly correlated — the predicted Y is closer to its mean than X was to its own. The mean is the gravitational centre toward which any less-than-perfectly-correlated system is pulled.

    Galton’s discovery survives one essential restatement for the modern reader: the phenomenon does not require any causal mechanism. It is a property of measurement under uncertainty. Stephen Stigler, the historian of statistics, has written that this is the single most under-taught idea in quantitative reasoning, precisely because it produces effects that the naïve mind insistently re-narrates as cause and effect (Stigler, Statistics on the Table, Harvard University Press, 1999, chapter 9). The investor who internalises only one rule from this essay should internalise that one: when an extreme draws an explanation, the explanation may simply be the arithmetic.

    Scatter of mid-parent height vs adult-child height, with the slope-2/3 Galton regression line and a 45-degree perfect-heredity line.
    Figure 1. Galton’s 1886 finding in modern dress. The 45-degree line is what perfect heredity would look like. The actual relationship is a flatter line through the centre of mass, with slope approximately two-thirds; tall parents produce tall-but-less-tall children, short parents produce short-but-less-short children. The drift is mathematical, not biological.

    2. The mechanism — why the drift is unavoidable

    The cleanest way to see why regression must happen is to decompose any observed outcome into two parts: a persistent component, call it Skill, and a transient component, call it Luck. Suppose we observe the top decile of a population — top-decile mutual-fund managers, top-decile athletes, top-decile parental heights. By definition, these observations carry unusually favourable Luck draws on top of whatever Skill they possess. In the next period, Skill persists by assumption, but Luck — being the noisy component — is, by definition, drawn afresh from a distribution centred on zero. The expected new observation is therefore Skill plus zero, lower than Skill plus favourable Luck. The top decile must, on average, fall back toward the mean.

    The size of the drift is governed by one ratio: the share of total variance contributed by Skill. If a domain is mostly Skill — Olympic sprint times for elite athletes after years of selection — regression will be small. If it is mostly Luck — single-quarter mutual-fund returns — regression will be enormous. Michael Mauboussin’s 2012 book The Success Equation (Harvard Business Review Press) titled this the “skill-luck continuum” and made the point that the investor’s first job in nearly every empirical domain is to estimate that ratio before extrapolating a single number.

    It follows that regression is asymmetric in a useful way. Extreme outcomes regress most strongly. Performance near the mean barely regresses at all. The investor who learns to feel the pull harder at the tails — both the celebrated top and the punished bottom — is doing the work of the model. And it follows, too, that the drift is on the conditional expectation: there is no guarantee that any specific top-decile observation will fall back, only that the average of all top-decile observations will. The mistake of applying a population property to one individual is the ecological fallacy in reverse, and it is endemic in financial commentary.

    3. The empirical record — three exhibits from active equity

    The financial evidence on mean reversion is, on the whole, embarrassingly consistent. Three exhibits.

    Mark Carhart’s 1997 paper “On Persistence in Mutual Fund Performance” (Journal of Finance 52(1): 57–82) examined the entire universe of US diversified equity funds from 1962 to 1993, sorted them into deciles each year on the basis of their previous-year return, and tracked the next-year performance. After adjusting for the market, size, value, and momentum factors he had assembled, the persistence of top-decile alpha was statistically indistinguishable from zero. The bottom decile, by contrast, persisted — bad funds stayed bad, largely because of expenses and turnover. The implication is the one Galton would have predicted: the favoured deciles are dominated by noise, and noise regresses; the disfavoured deciles include a structural deadweight (fees) that does not.

    S&P Dow Jones Indices has run a quarterly persistence scorecard for two decades. In the U.S. Persistence Scorecard Year-End 2024, the share of top-quartile US large-cap funds that remained top-quartile across the next five calendar years was 0%. That is not a misprint. Reading the same scorecard for the mid-2025 update, only 29% of top-quartile large-cap funds maintained their position even over a subsequent two-year window. A coin flip would have predicted 25% over two years and roughly 0.4% over five. Active equity performance is now indistinguishable, in persistence terms, from random selection followed by regression. The mathematics permits a sharper statement: the noisier the signal you select on, the less of it survives the second draw.

    Bar chart contrasting observed top-quartile fund persistence (per the SPIVA scorecards) against the share implied by random selection.
    Figure 2. Observed persistence of top-quartile US large-cap funds (navy) against the share implied by random selection at 0.25 raised to the power N (gold). Five-year observed persistence is rounded to zero; random selection would deliver about 0.1 percent. Carhart’s conclusion, replicated each year by S&P, is that top-quartile fund performance regresses approximately as Galton would have predicted.

    The corporate analogue is no less stark. Robert Wiggins and Timothy Ruefli, in “Sustained Competitive Advantage: Temporal Dynamics and the Incidence and Persistence of Superior Economic Performance” (Organization Science 13(1), 2002: 81–105), studied the return on assets of 6,772 firms across 40 industries between 1972 and 1997. Of the firms that achieved “superior” returns in their stratum, only about 5% sustained that position for 10 years or more, and just 0.5% for 20 years. The default destination of an above-average return-on-assets number is the industry mean, and the rate of decay can be calibrated. McKinsey’s repeat of the exercise in Valuation (Koller, Goedhart and Wessels, 7th edition, 2020, chapter 8) finds the same shape: the median high-ROIC firm gives back about half its excess return within seven to ten years.

    4. Two historical episodes

    Israeli flight school, 1969. Daniel Kahneman, then a young consulting psychologist for the Israeli Air Force, was lecturing senior flight instructors on the established behavioural finding that praise produces better learning outcomes than punishment. A grizzled instructor objected. With respect, sir, he said in effect, what you are saying is for the birds: I have many times praised flight cadets for the clean execution of some aerobatic manoeuvre, and the next time they tried it they did worse; I have often screamed into a cadet’s earphone for bad execution and on the next try he improved. Kahneman writes that the moment was the most important insight of his early career. The cadets’ performance was a noisy signal around a stable mean. A spectacularly clean manoeuvre was, by definition, mostly luck on top of skill; the next attempt would regress whether the instructor praised or screamed. The instructor had been the unwitting witness to thirty years of regression to the mean, mistaking statistical gravity for causation. The episode is recounted in Kahneman, Thinking, Fast and Slow (Farrar, Straus and Giroux, 2011), chapter 17, and its formal version had appeared four decades earlier in Kahneman and Tversky, “On the Psychology of Prediction,” Psychological Review 80(4), 1973: 237–251.

    Sports Illustrated cover jinx. From 1954 onward the legend within the magazine was that any athlete or team gracing the cover would subsequently underperform. In a 2002 internal review the editors counted 913 covers; 37% had been followed by some “decline.” Statisticians who examined the data — including Schaffer (2002) and Smith and Smith (2011) — found no jinx at all, only the regression Galton had described 116 years earlier. Cover athletes were, almost by selection, drawn from the upper tail of recent performance; mean reversion guaranteed that the next month would be statistically less impressive than the month that had earned them the cover. The “jinx” was a narrative built around the arithmetic of selection.

    Both episodes carry the same warning for the investor: when an observation is selected because it is extreme, the next observation will, on average, be less extreme. Any narrative that explains the change in causal terms — the cover cursed him, the praise spoiled her, the new CEO destroyed the franchise — is a narrative that may simply be re-describing regression. The mind reaches for a story; the spreadsheet would have suggested gravity.

    5. Application to long-term equity investing

    Three operating disciplines fall directly out of Galton’s mathematics.

    Discipline one: never project the past five years of profit margins straight into the future. The single most reliable mean-reverting series in financial history is the corporate profit share of national income — what GMO’s Jeremy Grantham has called, only half-joking, the most mean-reverting series in finance. The mechanism is the one Adam Smith identified in The Wealth of Nations (1776, Book I, Chapter VII): high margins attract competition, low margins repel it. The empirical record in the United States, where post-1947 National Income and Product Accounts data permits a long view, shows after-tax corporate profit margins oscillating in a relatively narrow band around 6 to 8 percent of GDP, with each excursion to either extreme corrected within roughly a decade. A discounted-cash-flow model that capitalises peak margins as a terminal-year assumption will, in regression-to-the-mean terms, systematically overstate intrinsic value at cycle peaks and understate it at troughs. The corrective is mechanical: stress-test every long-duration model with a margin path that reverts to a sector mean within ten years, and require the investment thesis to survive that test.

    Discipline two: when selecting active managers — including selecting oneself as one’s own active manager — discount the persistence of recent outperformance to roughly nil after five years. The Carhart finding, replicated in every multi-year SPIVA scorecard, is that top-decile performance over one-, three-, and five-year windows is almost entirely a noise phenomenon, with one important exception: costs and structural disadvantages — high fees, poor execution, persistent leverage at the wrong points in the cycle — produce real persistence on the downside. The investor’s manager-selection model should therefore be asymmetric. Be sceptical of celebrated past returns; take negative persistence seriously as a structural signal rather than a temporary embarrassment.

    Discipline three: at extremes of valuation, the price-multiple itself becomes the dominant mean-reverting variable. Robert Shiller’s cyclically adjusted price-earnings ratio (CAPE), constructed in Campbell and Shiller, “Stock Prices, Earnings, and Expected Dividends,” Journal of Finance 43 (1988), has the unhappy distinction of explaining roughly 40 percent of the variance in subsequent ten-year real US equity returns since 1881. The mechanism is again Galton’s: peak multiples, like peak margins, are by definition the joint product of fundamentals and noise, and the noisy component regresses. This is not a market-timing claim — short-horizon predictability is essentially zero — but a discipline against starting positions at extreme starting multiples without a compensating margin of safety. The investor who buys at the 95th percentile of CAPE and waits ten years must expect that the median outcome will be set largely by multiple compression, not by underlying earnings growth.

    Two decay curves showing how excess return on assets above the industry mean half-lives away over 20 years for an average firm and for a moated franchise.
    Figure 3. Stylised decay of excess return on assets above the industry mean, calibrated to the orders of magnitude in Wiggins and Ruefli (2002) and McKinsey’s Valuation (2020). The average top-decile firm halves its excess in about five years; a moated franchise can stretch the half-life to roughly fifteen. Neither curve flattens at a permanent plateau; the industry mean is the gravitational floor.

    6. How the long-term equity tradition has used it

    Howard Marks, in his July 2003 memo “The Most Important Thing” and the May 2001 memo “You Can’t Predict, You Can Prepare,” made regression to the mean the engine of his cycle theory. The Oaktree pendulum, Marks wrote, swings not from euphoria to despair because anyone wills it to, but because the very behaviours that produce extreme valuations contain the seeds of their reversal: high valuations attract supply of paper and erode prospective returns until the marginal buyer rebels; low valuations starve supply and improve prospective returns until the marginal seller capitulates. In his 2011 book of the same name, The Most Important Thing: Uncommon Sense for the Thoughtful Investor (Columbia Business School Publishing, 2011), Marks devotes an entire chapter — chapter 8, “Being Attentive to Cycles” — to the proposition that the investor who fails to internalise mean reversion will be most aggressive when prospective returns are lowest and most defensive when they are highest. His operating heuristic, articulated again in the September 2014 memo “Risk Revisited,” is to scale risk-taking inversely with prevailing valuations, precisely because of the Galton mechanism.

    Jeremy Grantham at GMO has built the firm’s seven-year asset-class forecast on the same idea. In a sequence of quarterly letters from 1994 onward, and in the February 2012 letter “The Longest Quarterly Letter Ever,” Grantham observes that profit margins and price-earnings ratios are the two great mean-reverting variables in equity markets, and that GMO’s forecasts assume both will return to their long-run averages within seven years. His June 2017 piece, “This Time Seems Very, Very Different,” extended the framework with a candid admission: in the platform-monopoly era, the speed of regression in margins has slowed; he now models a fifteen-to-twenty-year half-life for profit-share reversion rather than the seven years that prevailed from 1900 to 1997. The model survives; only the time constant has changed. The GMO seven-year forecast remains the most public mean-reversion betting card in the industry, and its track record — broadly accurate on direction across decadal windows, frequently early on timing — is precisely what one would expect from a model that uses Galton’s gravity correctly but cannot pin the exact moment of return.

    Warren Buffett, characteristically, has acknowledged the same gravity while resisting its full implications for the highest-quality franchises. In his 1989 Berkshire Hathaway letter (“Mistakes of the First 25 Years”) he wrote that he had repeatedly paid too little attention to the tendency of high returns on capital to attract competition, and too much attention to apparently cheap statistical bargains where the underlying economics were quietly regressing toward unprofitability. His later doctrine — pay a fair price for a wonderful business — is in part an acknowledgement that some businesses, by virtue of structural moats, regress more slowly than the average, but not that they regress not at all. The 1999 Sun Valley speech, reprinted in Fortune on 22 November 1999 under the title “Mr. Buffett on the Stock Market,” is a sustained warning that aggregate US after-tax corporate profits have been mean-reverting against GDP for the entire post-war period and will not, contrary to the late-1990s consensus, settle permanently at a higher plateau.

    The discipline these three practitioners share is not market timing but probability-weighting. Each builds his portfolio around the prior that extreme observations — of returns, of margins, of multiples — will, on average, fall back toward a mean that one can estimate from a long enough history. The variance of each individual outcome remains large; the directionality of the conditional expectation does not.

    7. Key takeaways

    Galton’s regression is a property of measurement under uncertainty, not a causal force; the mind insists on re-narrating it as one and the investor must resist that re-narration. The strength of the pull is set by the ratio of Skill variance to Luck variance, and that ratio must be estimated domain by domain before extrapolating any extreme observation. In equity markets, the two great mean-reverting variables are profit margins and valuation multiples, and every long-duration model should be stress-tested with a path that reverts both within ten to fifteen years. Manager-selection should be asymmetric: discount celebrated outperformance toward random, but take poor performance and high costs as the persistent signals they are. In cycle terms, the Oaktree pendulum and the GMO forecast both encode the same Galton insight; the investor who internalises it tends, over decades, to take risk when others will not and to take risk off when others crowd in. The model is 140 years old; its tax on the investor who ignores it is paid afresh in every cycle.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Money Illusion: Shafir, Diamond and Tversky’s 1997 Discovery and Why the Long-Term Equity Investor Routinely Confuses Nominal Returns With Real Wealth

    Money Illusion: Shafir, Diamond and Tversky’s 1997 Discovery and Why the Long-Term Equity Investor Routinely Confuses Nominal Returns With Real Wealth

    The NorthPath Letter · Behavioural Finance · Afternoon Edition · 29 May 2026

    The bias: a 1997 paper that named what Irving Fisher had already seen

    The long-term equity investor’s relationship with money is supposed to be straightforward. A rupee, a euro, a dollar is a unit of account; capital is what it can buy. The portfolio is a claim on future real consumption, and only the real return — the return after inflation — matters for whether the investor’s family will retire comfortably, fund a child’s university place, or endow a charitable trust. None of this is controversial. It is taught in the first month of any finance course. And it is, for most investors most of the time, quietly ignored.

    The bias that explains the gap between the textbook treatment and lived practice has a name. In May 1997 the Quarterly Journal of Economics published a paper titled simply “Money Illusion” by Eldar Shafir, Peter Diamond and Amos Tversky. Through a series of carefully designed surveys — on house purchases, salary negotiations, bonus framing and contract evaluation — the authors showed that ordinary people, including those with formal training, systematically evaluate transactions in nominal terms even when they have been told, explicitly, that the relevant comparison is the real one. The bias was robust, replicable across populations, and present in domains where the welfare consequences of getting it wrong were large.

    The paper did not invent the concept. Irving Fisher’s 1928 book The Money Illusion, published in New York by the Adelphi Company, had defined the phenomenon as the “failure to perceive that the dollar, or any other unit of money, expands or shrinks in value” — a definition that has not been improved upon in the century since. What Shafir, Diamond and Tversky added was the experimental architecture: a way of demonstrating, in clean laboratory conditions, that even when both the nominal and real numbers were placed directly in front of the respondent, the nominal frame remained the dominant influence on the choice. Money illusion was not, in their account, a failure of arithmetic. It was a failure of representation — a deep feature of how the mind encodes value.

    That distinction is what makes the bias important for the long-term equity investor. The horizon over which a portfolio compounds is precisely the horizon over which inflation does its work. A 6% nominal return at 2% inflation and a 9% nominal return at 5% inflation are economically the same trade; one feels much better. A pension pot that doubles in twenty years while the cost of living triples has shrunk in real terms by a third; the statement, in nominal currency, looks like a triumph. The bias the 1997 paper described is therefore the bias most likely to be active when the investor is reviewing the very horizon on which the strategy is built.

    The mechanism: dual representation, with the nominal frame doing most of the work

    Shafir, Diamond and Tversky proposed a specific cognitive architecture. People, they argued, hold simultaneous representations of economic quantities in nominal and real terms. Asked directly, in a vacuum, which representation is “correct,” most respondents will name the real one. But choices, judgements and emotional reactions are influenced by both representations, and the nominal representation tends to dominate because it is salient, simple, immediate and shared with everyone else who uses the same currency. The real representation requires an act of mental adjustment that the brain treats as effortful, optional and easily skipped under time pressure or affective load.

    This dual-representation account has held up well under three decades of follow-on work. Replication studies — including a 2020 multi-country extension published in the Journal of Behavioral and Experimental Economics and a 2024 Brazilian replication in the same journal — recovered the core effects with consistent magnitudes. The bias is documented in wage negotiations (workers strongly prefer a 2% raise at 4% inflation to a 2% pay cut at 0% inflation, even though the real outcome is identical), in housing decisions (sellers anchor to the nominal price they paid, ignoring the price level that has moved beneath them), in bond evaluation (retail investors describe high-coupon bonds during inflationary periods as “safe income” even when the real coupon has gone negative) and, most importantly for the present subject, in equity returns (investors compare the index level today with the index level at purchase and call any positive number a gain).

    The architecture also explains why money illusion is so resistant to education. Knowing that real returns are what count does not remove the nominal frame; it adds a second frame on top of it. Under deliberation, the educated investor can override the nominal reading. Under stress, fatigue or strong emotion — exactly the conditions in which most portfolio decisions are actually taken — the override fails, and the nominal frame wins by default. The 1997 paper did not phrase it this way, but the implication is unavoidable: money illusion is not solved by knowing about money illusion. It is solved, if at all, by building procedural defences that do the real-terms translation for the investor before the decision is taken.

    The empirical record: regulators, surveys and three decades of measured loss

    Two regulator anchors from two regions establish that the bias is neither academic nor historical. In the United States, the Treasury introduced Treasury Inflation-Protected Securities (TIPS) in January 1997 — the same year the Shafir-Diamond-Tversky paper appeared in print. TIPS exist precisely because nominal Treasury securities are not a complete instrument set for an investor who wants to express a view about real, rather than nominal, future cash flows. The Federal Reserve’s Survey of Consumer Finances and the long history of TreasuryDirect retail data confirm that even after a quarter-century of availability, retail allocation to TIPS remains a single-digit percentage of household fixed-income holdings, while retail allocation to nominal Treasuries and certificates of deposit has remained an order of magnitude higher. The instrument that solves the problem exists; the bias keeps most retail investors from using it.

    In the United Kingdom, the Financial Conduct Authority’s Financial Lives 2024 survey put a sharp number on the same phenomenon for cash savings. Among UK adults holding at least £10,000 in cash but no investments, only 59% agreed with the statement that money held in cash savings decreases in value because interest rates usually do not keep pace with inflation. The remaining 41% — a population whose holdings, in aggregate, run into the hundreds of billions of pounds — either disagreed, were uncertain, or had never considered the question. The FCA’s Consumer Duty rules, in force since July 2023, and the new Consumer Composite Investments disclosure regime scheduled to replace the UK PRIIPs framework on 6 April 2026, both require firms to communicate the real-terms consequences of cash drag and inflation more clearly. The fact that such rules have to be written, and re-written, two and three decades after the academic work was complete, tells the long-term equity investor everything about how durable the bias is.

    Academic measurement has converged on the same conclusion through a different channel. The work of Eugene Fama and others on the Fisher hypothesis, the long literature on the equity risk premium measured in real terms, and the regular monitoring of household real wealth by central banks all show the same gap between the nominal accounts that investors keep and the real accounts that ultimately determine their consumption. The accounts disagree most violently during inflationary episodes, when the nominal numbers are flattering and the real numbers are punishing. They disagree most quietly during disinflations, when the nominal numbers are flat and the real numbers are slowly improving — and the investor concludes, wrongly, that nothing is happening.

    Money illusion mechanism — nominal vs real representations and where the override fails
    Figure 1. Dual-representation mechanism. The nominal frame is salient, shared and immediate; the real frame requires an adjustment step that fails under fatigue, affect or time pressure.

    Two historical episodes: the 1970s US and the 2021-2024 global inflation

    The 1970s in the United States are the canonical case. Between January 1972 and December 1981 the US Consumer Price Index rose by approximately 134%; over the same decade the S&P 500 price index rose by approximately 38%. Investors who looked at their statements saw nominal gains, modest dividends, and a portfolio that had not collapsed. Investors who priced their portfolios against the cost of the lives they were actually living saw something quite different: a real loss of roughly 40% in equity values, accompanied by a similar real loss in nominal bond portfolios. The decade was, for the long-term equity investor, one of the worst in the modern record. The nominal statements obscured it. Warren Buffett, in his May 1977 Fortune article “How Inflation Swindles the Equity Investor,” explained the mechanism: stocks, in economic substance, are perpetual instruments paying a real coupon, and inflation reduces that coupon both by raising the cost of the assets the business must replace and by raising the return required from competing securities. The piece is read today not as an investment recommendation but as the clearest contemporaneous diagnosis of money illusion in equity markets ever written.

    The 2021-2024 episode supplies the modern counterpart. US CPI inflation, having been around 1.4% in January 2021, rose to a peak of 9.1% in June 2022 — the highest reading in four decades. Eurozone harmonised inflation peaked at 10.6% in October 2022. United Kingdom CPI inflation peaked at 11.1% in October 2022. India’s headline CPI inflation, while less extreme, averaged above the Reserve Bank of India’s 4% target throughout 2022 and into 2023. During the same three-year window most global equity indices delivered nominal returns that, viewed in isolation, looked respectable. Adjusted for the local cost of living, a substantial fraction of those returns disappeared. Cash deposits, money-market funds and short-duration nominal bonds — all of which appeared on retail statements as “safe” — delivered real returns that were, for an extended interval, materially negative. The fact that the bias was active in this episode, despite forty-eight years of accumulated academic and journalistic warning since the 1977 Fortune article, is the strongest available evidence that knowledge alone does not dissolve money illusion.

    Howard Marks’s December 2022 Oaktree memo “Sea Change” is the practitioner reading of the same episode. Marks argued that the four-decade environment of disinflation and falling interest rates had ended, and that investors who had built mental models — and portfolio structures — calibrated to that environment would have to rebuild them for a regime in which positive real returns from credit instruments became possible again and the equity premium had to be re-earned. The memo’s importance for the present essay is that it framed the regime change explicitly in real terms. Most of the commentary it provoked, in the financial press and on retail platforms, did not.

    Nominal vs real returns in two inflationary episodes — 1970s US and 2021-2024 global
    Figure 2. Two episodes, the same pattern. Nominal account balances looked acceptable; real purchasing power eroded materially before the disinflation arrived.

    The counter-measure framework: three disciplines that translate before the decision is taken

    The bias is durable. The defences therefore have to be procedural rather than cognitive — built into the workflow of portfolio review, not into the goodwill of the reviewer. Three disciplines, in combination, do most of the work.

    Discipline one — keep two sets of books and report on the real one first. Every periodic portfolio review should open with the real-terms account and only then turn to the nominal-terms account. The real account expresses the portfolio’s value, contributions and withdrawals in the units of a chosen base period (any base will do; consistency matters more than the choice). The deflator is published — for India by the Ministry of Statistics’ CPI series and the RBI’s WPI series; for the eurozone by Eurostat’s HICP; for the United Kingdom by the Office for National Statistics; for the United States by the Bureau of Labor Statistics’ CPI-U. The investor builds a spreadsheet, fills it monthly, and looks at the real column first. The simple act of putting the real number above the nominal number in the page layout removes most of the bias most of the time.

    Discipline two — express every long-term target in real terms, and refuse to translate. The investor who says “I want my portfolio to fund €48,000 of annual spending in today’s purchasing power for 30 years from retirement” has stated a target the brain can verify against the real account. The investor who says “I want a portfolio worth €2 million” has stated a target the brain can hit nominally while losing real ground. The first form is harder to formulate, harder to communicate to a spouse, and harder to celebrate when reached, precisely because it is honest. The second form is easy, comforting and frequently wrong. The discipline is to write the target down in the first form, share it with the household in that form, and refuse — including to oneself — to translate it into a nominal headline. Where a nominal headline is unavoidable for administrative purposes (a will, a contribution limit, a tax declaration) the real-terms equivalent is recorded alongside it in the same document.

    Discipline three — pre-commit to inflation-linked allocations as a structural, not tactical, position. The instruments exist. US TIPS, UK index-linked gilts (in issue since 1981), French OATi and OAT€i, Italian BTP€i, German Bund€i, Brazilian NTN-B, Chilean UF-linked deposits, Indian Inflation-Indexed Bonds: each is a direct claim on real, rather than nominal, future cash flows. The discipline is to decide in advance, in writing, what fraction of the long-term portfolio will be held in inflation-linked instruments and to hold to that fraction across regimes. The fraction will be wrong much of the time — inflation-linked instruments under-perform nominal ones during disinflations and the gap can persist for years — but the structural commitment removes the bias-driven temptation to abandon real-terms protection precisely when it begins to be needed. The 1997 paper’s central finding, that nominal evaluation dominates real evaluation under cognitive load, is the reason such commitments must be pre-written rather than left to in-the-moment judgement.

    Three counter-measure disciplines for money illusion
    Figure 3. Three procedural disciplines. None of them depends on the investor remembering, in the moment, that real is what counts.

    How long-term equity practitioners have addressed it

    Benjamin Graham treated the question in the chapter on “The Investor and Inflation” in the 1973 fourth edition of The Intelligent Investor. Graham acknowledged the natural impulse to look at the dollar amount of a portfolio and to be satisfied with growth in that amount, and then insisted on a real-terms view: the investor’s purchasing power, not the dollar count, was the proper test of investment success. Graham did not use the modern psychological vocabulary; he treated the bias as an obvious error of accounting that the disciplined investor would simply refuse to make. The chapter remains, half a century later, the most concise statement of the real-terms discipline in the practitioner literature.

    Warren Buffett’s 1977 Fortune article was a more explicit attack on the same problem at the level of the security rather than the portfolio. The essay defines the “equity coupon” — the real return generated by the underlying businesses — and argues that during inflationary periods the equity coupon is squeezed from both sides: by the rising cost of the fixed assets the business must replace, and by the rising returns available on competing instruments. The piece’s specific empirical claim, that the long-run return on equity was anchored near 12% nominal regardless of inflation, has been debated since; the diagnostic frame has not. Buffett returned to the theme in shareholder letters throughout the late 1970s and again, briefly, in the inflationary episodes of 2021-2023, restating the point in the language of owner economics: only the real change in the value of the underlying businesses matters; the inflation-adjusted accounts are the only accounts that tell the truth.

    Howard Marks’s December 2022 memo “Sea Change” provides the contemporary practitioner reading. Marks argued that the long disinflation of 1982-2021 had produced a generation of investors whose mental models, portfolio structures and risk frameworks were calibrated to a falling-rate, low-inflation world. The regime change, in Marks’s account, was not a tactical opportunity but a requirement to rebuild those mental models in real-terms language. Where the memo is read carefully, the prescription is identical to Graham’s: maintain the real account, set the targets in real terms, and accept that inflation-linked allocations are a structural rather than tactical commitment. Three generations of long-term investors, each writing in the vocabulary of their own era, have produced, on the question of money illusion, the same framework.

    Key takeaways

    • The bias is one of representation, not arithmetic. Shafir, Diamond and Tversky’s 1997 paper showed that the nominal frame dominates even when the real numbers are placed in front of the respondent. Education alone does not remove it.
    • The instrument set that solves the problem exists. TIPS (US, 1997), index-linked gilts (UK, 1981), eurozone OATi/BTP€i/Bund€i, Indian IIBs and inflation-linked deposits in several emerging markets are all direct claims on real, not nominal, cash flows. Retail allocation to them remains small.
    • The 1970s and 2021-2024 are the same lesson, half a century apart. Nominal accounts looked acceptable in both episodes; real purchasing power eroded materially. The repetition is the proof that knowledge alone does not dissolve money illusion.
    • The defence is procedural. Three disciplines — keep two sets of books and report on the real one first, state long-term targets in real terms and refuse to translate, pre-commit to inflation-linked allocations as a structural position — do most of the work because they take the real-terms translation out of the moment of decision.
    • Graham, Buffett and Marks converged on the same framework. Three generations of long-term equity practitioners, writing in different decades and different vocabularies, produced the same prescription. Money illusion is the bias that long-term equity investing was, in part, invented to defeat.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The Central Limit Theorem: Laplace’s 1810 Memoir and Why the Long-Term Investor’s Friend Is Aggregation, Not Prediction

    The Central Limit Theorem: Laplace’s 1810 Memoir and Why the Long-Term Investor’s Friend Is Aggregation, Not Prediction

    AFTERNOON EDITION — MENTAL MODELS · Essay No. 03 in the Mental Models series · The NorthPath Letter · 28 May 2026 · Tallinn

    The Model — Laplace, 1810

    The Central Limit Theorem is the most consequential theorem in probability theory for the long-term investor, and almost no investor knows its exact statement. In ordinary language it says: when you add up a large number of independent random influences, none of which is overwhelmingly large compared with the others, the distribution of the sum is approximately Gaussian — bell-shaped — no matter what the individual distributions look like. The bell curve is not a fact about nature. It is a fact about aggregation.

    The first published general version appears in Pierre-Simon Laplace’s Mémoire sur les approximations des formules qui sont fonctions de très grands nombres et sur leur application aux probabilités, read to the Institut de France in April 1810 and printed in the Mémoires de l’Académie des Sciences later that year. Laplace generalised an earlier special case proved by Abraham de Moivre in The Doctrine of Chances (second edition, 1738; first stated in a 1733 supplement), in which de Moivre derived the bell-shaped approximation to the symmetric binomial. Stephen M. Stigler, in The History of Statistics: The Measurement of Uncertainty Before 1900 (Harvard, 1986, chapters 2 and 3), credits Laplace with extending the result to sums of independent variables drawn from arbitrary distributions and with embedding it in his programme of inverse probability. Lucien Le Cam’s monograph article “The Central Limit Theorem Around 1935” (Statistical Science, vol. 1, no. 1, 1986, pp. 78–96) traces the modern Lyapunov–Lindeberg–Feller rigorisation, which fixes both the conditions under which the theorem holds and, equally important, the conditions under which it fails.

    The one-sentence form for an equity practitioner is this: the average of many small, independent, finite-variance shocks looks Gaussian even when each shock is not — and that fact is the entire architecture of risk management, factor models, Sharpe ratios, and modern portfolio theory. Strip the theorem away and almost every quantitative technique on a typical asset-management desk goes with it.

    Convergence to Gaussian: distribution of sums of independent uniform variables for n=1, n=2, n=5, n=30, showing the bell curve emerging from aggregation
    Figure 1. The Central Limit Theorem in action. The distribution of a single uniform draw is flat; sum two and it is a triangle; sum five and the bell shape is visible; sum thirty and the histogram is, for practical purposes, Gaussian. The shape of the individual contributors is irrelevant; the geometry of repeated convolution is the entire story.

    The Mechanism

    Why does aggregation produce a bell curve? Intuition first, formality second. Each random influence contributes some mean and some variance to the sum. When you add a great many of them, the means stack linearly but the variances also stack linearly — so the standard deviation of the sum grows only as the square root of n. The relative dispersion shrinks. What is left, once you standardise by that shrinking dispersion, is determined not by the shape of the individual contributors but by a deeper geometric fact about the convolution of probability densities. Convolution is a smoothing operation; repeated convolution drives the result toward the unique shape that is invariant under further convolution and standardisation. That fixed point is the Gaussian.

    A more careful statement: if X₁, X₂, … are independent identically distributed random variables with finite mean μ and finite variance σ², then the standardised sum (X₁ + … + Xₙ − nμ) ⁄ (σ√n) converges in distribution to a standard normal as n grows without bound. The Lindeberg–Feller refinement weakens the identical-distribution assumption and replaces it with a condition that no single variable dominates the sum, formalised as the Lindeberg condition that the contribution of any individual term to the total variance must vanish in the limit.

    Two things are essential and the long-term investor must internalise both. First: independence. The theorem says nothing about dependent variables that share a common shock. Second: finite variance. The theorem says nothing about variables drawn from distributions whose second moment does not exist. Both qualifications are central to what follows, and both are violated routinely in the financial world.

    The Empirical Record

    Equity returns over short horizons emphatically do not follow a normal distribution. The classic stylised facts, catalogued by Rama Cont in “Empirical Properties of Asset Returns: Stylized Facts and Statistical Issues” (Quantitative Finance, vol. 1, no. 2, 2001, pp. 223–236), are: heavy tails — excess kurtosis of daily returns is routinely above ten — volatility clustering, leverage effects, and what Cont labels “aggregational Gaussianity,” the empirical observation that the distribution looks more bell-shaped at monthly and quarterly horizons than at daily horizons. The Central Limit Theorem does operate on equity returns, but slowly, and only because daily returns are neither truly independent nor drawn from a stationary distribution.

    Eugene F. Fama’s 1965 PhD work, “The Behavior of Stock Market Prices” (Journal of Business, vol. 38, no. 1, pp. 34–105), found that daily stock returns are leptokurtic and rejected the simple Gaussian model. Benoit Mandelbrot’s earlier “The Variation of Certain Speculative Prices” (Journal of Business, vol. 36, no. 4, 1963, pp. 394–419) had already proposed stable Paretian distributions with infinite variance — distributions to which the classical CLT does not apply. The empirical picture six decades on is that the Gaussian arrives at aggregation horizons measured in months and years, not days, and even then only as an approximation that breaks down in the tails.

    The Bank for International Settlements Quarterly Review of December 2019 noted that the September 2019 US repo-market spike, which a standard one-factor Gaussian model would have placed at roughly a 1-in-10⁹ probability, had in fact occurred within ten years of the previous comparable dislocation. The US Office of Financial Research’s Annual Report (2020) made the same point for equities: the 12 March 2020 single-day −9.5% S&P 500 close was, under a Gaussian volatility regime calibrated to the prior year, a roughly 10-sigma event — once in many billions of years on a Gaussian planet. We are not on a Gaussian planet at the daily frequency, but we drift toward one as the aggregation window widens.

    The European Securities and Markets Authority’s annual Trends, Risks and Vulnerabilities Report (2024 edition) reaches a complementary conclusion from the regulator’s perspective. Across the 2015–2023 period, single-day European blue-chip equity moves of greater than four standard deviations occurred roughly nine times more often than a constant-volatility Gaussian model would predict, and the excess was concentrated in clustered episodes — March 2020, September 2022, March 2023 — that violated independence within the cluster while otherwise looking benign. The supervisor’s operational conclusion is the one a thoughtful investor should already have reached: the Gaussian framework is a useful default for setting capital under normal conditions, and a dangerously misleading default for setting capital under stress conditions.

    Bar chart comparing empirical S&P 500 daily return distribution to fitted Gaussian: body fits closely, tails diverge by orders of magnitude beyond ±4σ
    Figure 2. The empirical CLT verdict on equity returns. Stylised representation of S&P 500 daily return frequencies versus the best-fit Gaussian on a log frequency scale. The body of the empirical distribution tracks the bell curve. The tails diverge by orders of magnitude. The lesson for the operator: trust the bell in the middle, distrust it at the edge.

    Two Historical Episodes

    The collapse of Long-Term Capital Management in September 1998 is the textbook study of misapplied CLT. Roger Lowenstein’s When Genius Failed: The Rise and Fall of Long-Term Capital Management (Random House, 2000) and the President’s Working Group on Financial Markets report “Hedge Funds, Leverage, and the Lessons of Long-Term Capital Management” (April 1999) document the firm’s value-at-risk machinery, which assumed that daily P&L was approximately Gaussian with variance estimated from a rolling five-year window. Convergence trades — long off-the-run US Treasuries against short on-the-run; long Italian government bonds against short German Bunds; equity-pair arbitrages — were sized so that a one-day standard deviation of book P&L was roughly forty-five million dollars on equity of four-and-three-quarter billion. The model implied that a one-billion-dollar daily loss carried a probability of approximately one in 10²⁴. The fund lost five hundred and fifty-three million dollars on a single day, 21 August 1998, and was insolvent within five weeks. The independence assumption had failed: when the Russian sovereign default touched off a global flight to liquidity, every supposedly independent trade became one and the same bet on the willingness of leveraged intermediaries to provide funding.

    The 19 October 1987 crash is the older episode. The Dow Jones Industrial Average fell 22.6% in a single trading session. Under the lognormal model that underpinned the Black–Scholes pricing of the portfolio-insurance strategies that contributed to the cascade, a one-day move of that magnitude was a roughly 20-sigma event — frequency-equivalent to once in many times the age of the universe. The Brady Commission’s Report of the Presidential Task Force on Market Mechanisms (January 1988) attributed the cascade to feedback among index futures, programme trading, and portfolio insurance — three features that violated the independence assumption simultaneously. Mark Rubinstein, one of the architects of the insurance approach, later acknowledged in his Frank J. Fabozzi Memorial Lecture (2000) that the model had treated the insurance-driven order flow as exogenous when in fact, at scale, it was the dominant endogenous shock. The operational point is that risk frameworks built on the CLT manufactured a false sense of safety in regimes where independence was the first thing to break.

    Application to Long-Term Equity Investing — Three Operating Disciplines

    The first discipline is to know whether you are in CLT territory before you trust an average. Aggregation across many independent positions, holding periods, or business cycles is the equity investor’s friend. Aggregation across positions that share a common factor is a false friend. A portfolio of fifty mid-cap equities held for twenty years across multiple credit and policy cycles has many of the independence properties the CLT requires. A portfolio of fifty European peripheral-sovereign-exposed banks held over a quarter does not, because every name in it is a single, repeated bet on one shared variable. The first practical test before relying on a portfolio-level Sharpe ratio or standard deviation is to ask: in the bad case, do these positions move together?

    The second discipline is to keep finite variance on your side. Variance is finite when the worst possible single outcome is bounded — when individual position size is capped, when leverage is bounded, when any single illiquid concentrated bet does not exceed a defined fraction of capital. Variance becomes effectively infinite the moment a single trade can wipe out the book. This is the operational meaning of Munger’s “rule of intelligent compounding”: survive each year, and the CLT will reward you over decades. It is also the meaning of Buffett’s two rules about not losing money: a single ruinous draw turns the iterated multiplication of returns from a long-run Gaussian-in-logarithms into a zero, and the theorem cannot save a series whose product has been multiplied by nought.

    The third discipline is to distrust the bell curve in the tail and trust it in the body. The mid-distribution behaviour of well-diversified equity portfolios at multi-year horizons is genuinely close to Gaussian, and Sharpe ratios, mean-variance optimisation, and standard deviation are useful descriptive tools there. In the tail — drawdowns of thirty per cent, forty per cent, fifty per cent — the Gaussian model understates frequency by orders of magnitude. The long-term investor uses the bell curve for the body of the distribution to set position sizes and to evaluate strategies, and uses non-Gaussian thinking — scenario analysis, stress testing, leverage limits, liquidity buffers, written-down pre-mortems — for the tail.

    Three-card framework: discipline 1 manage independence, discipline 2 cap variance, discipline 3 separate body from tail
    Figure 3. Three operating disciplines that translate the Central Limit Theorem into an equity operating manual. Manage independence so the n in √n is real; cap variance so the theorem’s preconditions hold; split the bell curve into a body framework and a tail framework, and refuse to use one tool for the other.

    How the Long-Term Equity Tradition Has Used It

    Warren Buffett has been disarming about distributional assumptions. In the Berkshire Hathaway 1993 chairman’s letter, discussing the use of beta and standard deviation as proxies for risk, he wrote that academic definitions of risk wander off into absurdity once they require treating market volatility as the relevant measure for a long-term owner of a business. The implicit critique is that the Gaussian framework is the right tool for some questions — portfolio-level dispersion over many independent owners and many quarters — and the wrong tool for others, notably the probability of permanent loss of capital on a single concentrated position. The 2002 letter, in the famous passage on derivatives as “financial weapons of mass destruction,” sharpens the same point: when independence breaks, the standard deviation of P&L is no longer a meaningful summary statistic, because the joint distribution has collapsed to a single shared move.

    Howard Marks’s Oaktree memo “Risk” (January 2006), later reprinted as a chapter in The Most Important Thing (Columbia University Press, 2011), takes the same view from the tail-of-the-distribution side. Marks argues that risk is the probability of an unacceptable outcome, not the dispersion of outcomes — a definition that is dual to the CLT. His later memo “Investing Without People” (June 2018) returns to the idea in the context of passive-vehicle flows: as more capital is allocated to vehicles that trade in lockstep, the independence assumption underlying any portfolio-diversification benefit shrinks, and the effective n in the √n denominator of the CLT collapses.

    Charlie Munger’s “A Lesson on Elementary, Worldly Wisdom as It Relates to Investment Management and Business,” delivered at the USC Marshall School in April 1994 and reprinted in Poor Charlie’s Almanack (Donning, 2005), makes the most economical case for why the long-term investor must understand the CLT. Munger argues that compounding — the long-term equity investor’s central mechanism — is intelligible only as the iterated multiplication of independent returns, and that the geometric structure of the result is, in logarithms, essentially Gaussian. The discipline is to keep the inputs roughly independent, keep their variance bounded, and let arithmetic do the rest.

    Nick Sleep’s Nomad Investment Partnership letters (2001 to 2014, collected and reprinted in Nomad Investment Partnership Letters to Partners, 2021) repeatedly invoke aggregation logic: fewer decisions, smaller variance per decision, more time for compounding. Sleep’s portfolio concentration is unusual, but the principle — keep the number of large independent bets finite and well-understood — is a deliberate refusal to be ambushed by the failure modes of large-n CLT thinking, which silently assumes a great many small bets are genuinely independent when in fact they are correlated through factor exposure. François Rochon’s Giverny Capital quarterly letters, archived on the firm’s website since 1998, make the same observation about diversification from the other direction: beyond roughly twenty-five well-understood positions, the marginal CLT benefit is exhausted, and additional names dilute analytical attention without reducing systematic variance.

    Key Takeaways

    The Central Limit Theorem is not a description of nature. It is a description of what happens when you average independent, finite-variance random variables. The two assumptions — independence and finite variance — are the entirety of the engineering safety question for any portfolio that treats Gaussian statistics as its working language.

    Equity returns approach Gaussian shape only at long aggregation horizons and only in the body of the distribution. The tails remain heavier than the bell curve for as long as anyone has measured them, and the documented stylised fact of aggregational Gaussianity is exactly that — an approach, not an arrival.

    The CLT explains why broad, long-duration index investing tends to look close to its model, and why concentrated, short-duration, leveraged trading tends not to. The investor who builds one operating system around the CLT’s body and a separate operating system around its tail is treating the theorem honestly. The investor who applies the body’s tools to the tail will, sooner or later, blow up.

    Two of the most studied failures of CLT-based risk management — Long-Term Capital Management in 1998 and the 1987 crash — both stemmed from violating the independence assumption, not from any defect in the theorem itself. Independence is the assumption that breaks first in a panic, because a panic is by definition the moment when one common factor swamps every supposedly idiosyncratic shock.

    The long-term equity tradition — Buffett, Munger, Marks, Sleep, Rochon — has converged on a simple operating discipline that is the CLT made human: keep position sizes bounded so variance stays finite, keep correlations honest so independence stays approximately true, and let aggregation across many years do the work the theorem promises.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The Hot-Stove Effect: Denrell and March’s 2001 Adaptive-Sampling Model and Why a Single Punishing Draw Permanently Distorts an Investor’s Estimate of an Entire Category

    The Hot-Stove Effect: Denrell and March’s 2001 Adaptive-Sampling Model and Why a Single Punishing Draw Permanently Distorts an Investor’s Estimate of an Entire Category

    The NorthPath Letter  ·  Behavioural Finance  ·  Afternoon Edition

    In 1894, in Following the Equator, Mark Twain wrote a sentence that a hundred and seven years later would become the title of an academic paper in Organization Science. “We should be careful to get out of an experience only the wisdom that is in it — and stop there; lest we be like the cat that sits down on a hot stove-lid. She will never sit down on a hot stove-lid again — and that is well; but also she will never sit down on a cold one any more.” Twain was making a literary observation. Jerker Denrell and James G. March, writing in 2001, turned it into a formal model of how an adaptive learner, sampling outcomes from a set of options whose true qualities are uncertain, can converge on a stable and systematically wrong belief about the world.

    The argument is the single most important diagnostic any long-term equity investor can carry through a career, because it explains, more cleanly than any other behavioural construct, why the categories that an investor has been most recently burned by are precisely the categories the investor is most likely to misprice for the next decade. It is not a story about emotion. It is a story about information. And once the mechanism is understood, the discipline of counter-measure becomes a deliberate operational protocol rather than a feat of psychological self-overcoming.

    The bias, defined

    The hot-stove effect, as formalised by Denrell and March in “Adaptation as Information Restriction: The Hot Stove Effect” (Organization Science 12(5), 2001, pp. 523-538), is the proposition that adaptive learning — the simple rule of returning to options that have rewarded you and avoiding options that have punished you — produces, in the presence of any noise in outcomes, a systematic and permanent under-estimate of the quality of risky and novel alternatives. The mathematics is elementary. The implication is severe. An option’s estimated value, in the mind of an adaptive learner, is the running average of the outcomes the learner has sampled from it. The learner’s sampling rule is to continue drawing from options whose running average is high and to stop drawing from options whose running average is low. The crucial observation is that the running average for any high-variance option will, sooner or later, dip below the threshold — not because the true mean of the option is low, but because the realised draws from a noisy distribution will occasionally include a string of bad outcomes. Once the running average dips below the threshold, the learner stops sampling. The estimate is then frozen at the punishing value forever. No further data arrives to correct it.

    Denrell extended the construction in “Adaptive Learning and Risk Taking” (Psychological Review 114(1), 2007, pp. 177-187), showing that the same mechanism produces an apparent and durable preference for low-variance options over high-variance options of equal true mean — without invoking any utility-curvature explanation. The agent in the model is not risk-averse in the Bernoulli sense. The agent is simply a sampler whose sampling has been truncated by its own past actions. The output looks like risk aversion. The underlying cause is information restriction.

    The mechanism

    It is worth pausing on the asymmetry, because the asymmetry is the entire phenomenon. Consider an option whose true expected return is positive but whose realised returns include a left tail of painful losses. An investor allocates a starter position. The first draw is one of the left-tail outcomes. The investor’s estimate of the option’s quality is now anchored on a deeply negative number. The investor’s decision rule — one that no professional would find unreasonable, because it is the rule by which prudent capital is supposed to be allocated — is to reduce or eliminate exposure to options that have disappointed. The investor disengages. Crucially, the investor’s disengagement is not a temporary suspension. It is the silent termination of all future data collection on that option. Every subsequent outcome that the option would have produced — including the right-tail outcomes that would have corrected the estimate — never enters the investor’s record at all. The estimate is locked at the punishing draw forever, and the investor will defend the lock with the entirely sincere argument that the data, such as it is, supports the decision.

    Compare this with the corresponding case of an option whose first draw happens to be a right-tail outcome. The investor adds to the position, samples more data, and the running average regresses to the true mean. The bias is not symmetric, because the sampling rule is not symmetric. Continued engagement provides more information; disengagement provides none. Across the full universe of options the investor considers, those that delivered a punishing first draw become permanently and systematically under-estimated. Those that delivered a rewarding first draw converge to their true means. The investor’s portfolio — and the investor’s expressed preferences for what to consider in future — is the equilibrium of this asymmetric learning. It is a portfolio shaped not by the world as it is, but by the world as the investor stopped looking at it.

    The further implication, drawn out by Denrell across both papers, is that the bias is heavier where outcome variance is heavier. Low-variance options — bonds, money-market instruments, defensive consumer staples held over very long periods — produce running averages close to their true means after very few draws, and the investor’s estimate is reasonably well-calibrated. High-variance options — cyclical equity, deep value, emerging markets, smaller-capitalisation securities, recovering sectors after a credit event — produce running averages whose first few draws can be far from the true mean, and these are precisely the categories most likely to be discarded after a single bad season and never re-sampled. The hot-stove effect is, in this sense, a structural reason why an entire class of investors will under-own precisely the categories that academic finance suggests offer the deepest long-horizon premia. It is not a story about courage. It is a story about who stops counting and when.

    Two-panel mechanism diagram: rewarded first draw vs. punished first draw.
    Figure 1. The asymmetric-sampling mechanism behind the hot-stove effect. Continued engagement after a rewarded draw lets the estimate regress to the true mean; disengagement after a punished draw freezes the estimate at the punishing realisation forever.

    The empirical record

    The cleanest large-sample evidence comes from regulators who have, in the years since the global financial crisis, methodically measured how retail households allocate when they have the resources to invest but the experience to remember being burned. The United Kingdom’s Financial Conduct Authority Financial Lives 2024 Survey, conducted between February and June 2024 with 17,950 respondents, documents that sixty-one per cent of adults with more than ten thousand pounds in investible assets hold at least three-quarters of those assets in cash rather than in any form of invested instrument. The same survey records that twenty-four per cent of all United Kingdom adults — some 13.1 million people — had low financial resilience as of May 2024. The FCA reports the headline in the language of cash hoarding and product hesitation. The hot-stove reading is sharper: a cohort of households, having lived through the 2008 financial crisis, the 2011 European sovereign episode, the 2020 pandemic drawdown and the 2022 gilt convulsion, has effectively withdrawn from the equity sampling exercise altogether. Their estimate of the long-horizon premium is locked at whatever the worst of those episodes told them. The locked estimate has not been corrected by the 2009-2021 equity recovery for the simple reason that they were not in the market to receive it.

    The complementary anchor is the United States Federal Reserve’s Survey of Consumer Finances, whose 1989-2022 triennial waves provide the longest comparable cross-section of household balance sheets in the world. The 2022 wave, published in October 2023, records that the equity participation rate among households headed by adults under thirty-five recovered only gradually after the 2008 crisis: in 2010 the participation rate for that cohort stood near a multi-decade low, and although the 2022 wave shows a meaningful rebound — with twenty-two per cent of Millennial and older-Gen-Z households reporting direct stock ownership and a further nine per cent holding pooled funds — this recovery took fourteen years, during which the broad United States equity market roughly quadrupled on a total-return basis. A generation of households, hot-stoved by the 2008 draw, were absent for most of the right-tail draws that followed. The cost of that absence, measured in lifetime wealth, is the most expensive consequence of a single bias in the long history of household finance.

    Three empirical anchors: FCA Financial Lives 2024 (61% UK cash hoard), SCF 1989-2022 (14-year recovery), Morningstar Mind the Gap (1-2% annual gap).
    Figure 2. The hot-stove effect at population scale, as recorded by three independent regulator and industry data series.

    Two historical episodes

    Japan, 1989 to 2020. The Nikkei 225 peaked at 38,915.87 on the final trading day of December 1989. Over the next two decades and a half, the index drifted, crashed, briefly recovered, crashed again and stayed below the 1989 high until 22 February 2024 — a period of more than thirty-four years. The behavioural consequence for Japanese households was the closest pure-form expression of the hot-stove effect ever recorded. Bank of Japan Flow of Funds data show that household financial assets shifted decisively into cash and cash-equivalents from 1990 onward: by 2010 roughly fifty-four per cent of household financial assets were held as cash and bank deposits, against approximately ten per cent in equity and investment trusts combined, a ratio that did not materially shift until the long-running Suga-Kishida policy push to encourage equity participation through the NISA tax-advantaged account framework. The cat had sat on the stove in 1989, and three decades of nominal cold draws were not enough to coax it back. Japanese households were not irrational. They were sampling the data the world had given them, and the data, restricted by the sampling rule itself, never produced a draw large enough to overwhelm 1989. The lesson is not that Japanese households should have ignored the 1989 outcome. It is that, once a single outcome had been treated as an estimate of the equilibrium mean, no information short of total re-sampling could dislodge it.

    The United States retail investor, 2000 to 2013. The dot-com correction that began in March 2000 and concluded around October 2002 erased roughly seventy-eight per cent of the Nasdaq Composite’s value at its trough. The 2008 financial crisis, beginning the long retracement in October 2007 and concluding in March 2009, took the S&P 500 down by approximately fifty-seven per cent peak-to-trough on a total-return basis. Investment Company Institute data on long-term mutual-fund flows show that net flows into United States equity mutual funds were essentially zero or negative every year from 2008 through 2012, even as the index recovered all of its losses by 2013. Retail investors who lived through the 2000 and 2008 draws had hot-stoved out of equity. Households who took the 2009 right-tail draw — those who held positions through the trough — saw the running average corrected. Households who disengaged in 2008 took no further draws and saw nothing corrected. The Morningstar “Mind the Gap” series, which has measured the investor-return gap each year since 2005, has consistently found that retail-investor cash-flow-weighted returns lag fund time-weighted returns by between one and two per cent per annum, with the largest gap concentrated in the most-volatile fund categories. The gap is not a story about manager selection. It is the hot-stove effect, measured in basis points.

    The counter-measure framework

    The first task in counter-acting a bias whose mechanism is asymmetric sampling is to construct a sampling rule that does not allow disengagement to be the default response to a punishing draw. Three operational disciplines, in combination, accomplish this without requiring the investor to abandon the broader prudential framework of position sizing and risk control.

    The pre-committed re-sampling protocol. Before a position is taken, the investor specifies, in writing, the conditions under which the position will be re-sampled following a loss. The conditions are stated in terms of objective triggers — a defined calendar interval, a defined drawdown level, a defined change in the fundamental thesis — rather than in terms of post-loss conviction, which the hot-stove mechanism reliably corrupts. The point of the pre-commitment is to ensure that the second draw is taken at all. The size of the re-sample need not match the size of the original position; in many cases a re-sample at twenty to thirty per cent of the original allocation will produce sufficient new information to update the estimate without exposing the investor to a position that may, on the second look, still be unattractive. The discipline is not “buy more after a loss.” The discipline is “look again after a loss,” with size and timing pre-specified.

    The decoupling of the security-specific lesson from the category lesson. A punishing draw on a single security carries two distinct pieces of information: information about the security and information about the category to which the security belongs. The hot-stove mechanism conflates the two, because the investor’s sampling rule is applied at the category level — “I have been burned by emerging markets,” “I have been burned by Indian small-caps,” “I have been burned by biotech” — rather than at the security level. The discipline is to write down, before the position is taken, what the position is being held to learn about. If the thesis is security-specific, the loss carries no category information at all and must not be allowed to terminate sampling at the category level. If the thesis is category-level, the loss carries security-specific noise that must not be allowed to dominate the category estimate. The bookkeeping is unglamorous but it is the only available defence against the asymmetric translation of one draw into a permanent verdict on an entire universe.

    The shadow-position register. For every category from which the investor has disengaged, the investor maintains a shadow allocation — a notional position, marked to market, that records what the disengaged category has subsequently done. The shadow allocation is not a trade; it is a sampling instrument. Its purpose is to ensure that the data the disengaged category produces is collected even when the actual portfolio has stopped collecting it. At a quarterly or annual review, the shadow register is examined against the live portfolio, and any category whose shadow has materially outperformed the realised allocation is flagged for re-sampling review under the protocol described above. The shadow register is the only mechanism by which an investor can reliably know that the cat has been sitting on a cold stove for some time, and that the discomfort of returning to it is no longer informational but reflexive.

    Three counter-measure disciplines: pre-committed re-sampling protocol, decouple security from category, shadow-position register.
    Figure 3. The three counter-measure disciplines that together prevent the investor’s sampling rule from terminating after a punishing draw.

    How long-term-equity practitioners addressed the bias

    Sir John Templeton built one of the longest sustained outperformance records in twentieth-century equity investing by operationalising, decades before it had an academic name, a refusal to allow a punishing draw to terminate his sampling of a category. His most celebrated single decision — the purchase of small parcels of one hundred and four United States equities trading below one dollar per share in 1939, financed with borrowed money on the day Germany invaded Poland — was a deliberate re-sampling of a category that the broader investor base had hot-stoved out of after the 1929-1932 collapse and the long depression that followed. Equally instructive is Templeton’s 1962 entry into Japanese equity, at a time when most Western capital had concluded that Japan was a country whose post-war recovery had stalled. Templeton wrote, in correspondence later collected by his successor at Templeton Funds, that the discipline was not to seek out hated assets for their hatedness; it was to ensure that no asset class was excluded from the consideration set by virtue of having most recently disappointed. The point of the discipline of “buying at the point of maximum pessimism,” in Templeton’s words, was not contrarianism. It was the construction of a sampling rule that did not permit disengagement to be permanent.

    Howard Marks, in three decades of Oaktree Capital Management memos to clients, has returned repeatedly to the asymmetry that Denrell and March formalised. The 2008 memo “The Limits to Negativism,” written in October of that year as distressed credit prices made forced-selling histories at multiples of recovery value, opens with the observation that the same intelligence that allows an investor to identify legitimate reasons for caution can, when extended without limit, prevent the investor from ever re-engaging with categories that have priced in those reasons. The 2012 memo “What’s Behind the Downturn?” reiterates the point. The 2020 memo “Latest Thinking,” written in the middle of the COVID-19 drawdown, makes it explicit: the question facing the investor in a punished category is not “is this asset dangerous?” but “is the price compensating me for the danger?” The discipline at Oaktree, as Marks has described it across two books and forty years of memo correspondence, is to require that any decision to remain disengaged from a category be re-justified on every cycle on the basis of current valuation, not on the basis of the most recent disappointment. The shadow-position register described above is, in effect, the Oaktree practice generalised to a single investor’s portfolio.

    Key Takeaways

    • The hot-stove effect is not a story about emotion or courage; it is a story about asymmetric sampling. Continued engagement with a category produces more information, and disengagement produces none. Estimates of categories from which the investor has disengaged after a punishing draw are systematically and permanently biased downward.
    • The bias is heaviest where outcome variance is highest, which means the categories most likely to be misjudged are precisely the cyclical, recovering, and high-tail-premium categories that academic finance suggests offer the deepest long-horizon returns.
    • Regulator data — the FCA Financial Lives 2024 Survey, the Federal Reserve Survey of Consumer Finances 1989-2022 series, and the Investment Company Institute fund-flow series — all document the same phenomenon at population scale: cohorts hot-stoved by a single drawdown remain underweight equity for a decade or more, missing the bulk of the subsequent recovery.
    • The counter-measure is procedural, not psychological. Three disciplines — the pre-committed re-sampling protocol, the decoupling of security from category, and the shadow-position register — together ensure that the sampling rule does not allow disengagement to be the default response to a punishing draw.
    • The defining practitioners of long-horizon equity — Sir John Templeton through six decades of cross-cycle category re-entry, and Howard Marks through forty years of memo correspondence on the asymmetric cost of sustained negativism — operationalised these disciplines decades before the academic literature gave the underlying mechanism its name. The architecture is available. The only requirement is the deliberate construction of a sampling rule that the investor’s own punishments are not permitted to terminate.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Graham’s Seven Tests for the Defensive Investor: The 1973 Quantitative Screen That Still Filters the Global Equity Universe

    Graham’s Seven Tests for the Defensive Investor: The 1973 Quantitative Screen That Still Filters the Global Equity Universe

    The NorthPath Letter · Value Investing · Morning Edition · 28 May 2026

    In the 1973 fourth edition of The Intelligent Investor, Benjamin Graham did something most authors of investment classics never do. He wrote a chapter that any reader could mechanically apply on a Saturday afternoon, at any kitchen table, in any country, against any equity market, without consulting a broker, an analyst, or a chart. He called it “Stock Selection for the Defensive Investor” (Chapter 14). It contained seven numerical tests — size, balance-sheet strength, earnings stability, dividend record, earnings growth, price-to-earnings ratio, and price-to-book ratio. A stock that survived all seven was, in Graham’s words, a candidate for a portfolio that the investor could hold “without losing sleep and without paying close attention.”

    The seven tests are not a stock-picking system in the modern sense. They are a filter on the population of listed equities, designed to leave behind only those for which the probability of a permanent loss of capital over a multi-year holding period is structurally low. The screen does not promise outperformance. It promises that you have begun the process of analysis from a universe in which most of the worst outcomes have already been excluded. That is a strikingly modest claim. It is also, in our experience after twenty-two years of practising the discipline, the only claim about a stock screen that does not eventually break.

    1. The principle

    Graham defined the “defensive investor” as someone “whose chief emphasis will be on the avoidance of serious mistakes or losses; second goal, on freedom from effort, annoyance, and the need for making frequent decisions.” The seven tests were the operational expression of that definition. Each test sets a numerical minimum (or maximum) that a candidate stock must clear. The thresholds, in Graham’s 1973 wording, were:

    Test 1 — Adequate size. Annual sales of at least USD 100 million (or, for a public utility, total assets of at least USD 50 million). Graham noted explicitly that this threshold was conservative even in 1973 and that “the result is to exclude small companies which may have unusual vicissitudes.” In 2026 currency, USD 100 million of 1973 sales is approximately USD 700 million to USD 1 billion of inflation-adjusted sales. We will return to this translation problem.

    Test 2 — Sufficiently strong financial condition. Two sub-tests, both of which must clear: current ratio of at least 2.0, and long-term debt no greater than the net current asset value (current assets minus all liabilities). Together these put a hard floor under the company’s ability to survive a working-capital squeeze or a refinancing crisis without resort to equity dilution.

    Test 3 — Earnings stability. Some earnings for the common stock in each of the past ten years. Note the precise wording: not earnings growth, not consistent returns on capital, simply positive reported earnings every year for a decade. The criterion excludes cyclicals at the peak of their cycle and start-ups that have not yet earned across a full business cycle.

    Test 4 — Dividend record. Uninterrupted payments for at least the past twenty years. Graham’s deepest test, and the one that, in our work today, narrows the qualifying universe more aggressively than any other. A twenty-year unbroken dividend record is corroborating evidence that reported earnings have, over two decades, been backed by free cash flow large enough to support a payout through wars, recessions, and changes of management.

    Test 5 — Earnings growth. A minimum increase of at least one-third in per-share earnings over the past ten years, using three-year averages at the beginning and end. The three-year averaging is the part most modern screens drop, and it is the part that does the real protective work: it removes the influence of cyclical peaks at either endpoint.

    Test 6 — Moderate price-to-earnings ratio. Current price not more than fifteen times average earnings of the past three years.

    Test 7 — Moderate price-to-asset ratio. Current price not more than one and one-half times the most recently reported book value. Graham added a softening clause: a lower P/E may justify a higher P/B, and a lower P/B may justify a higher P/E, provided the product of the two multiples does not exceed 22.5. That product test is the origin of what later authors christened “the Graham Number.”

    Seven tests. Each test is a number. Each number is verifiable from the audited annual report. The screen requires no access to management, no proprietary database, no view on macroeconomic conditions, and no judgement about industry attractiveness. Graham’s deliberate intention was to construct a filter that a non-expert could apply, that an expert could not improve upon by adding intuition, and that would, in his words, “not leave the investor much exposed to the dangers of the market.”

    2. The mechanism — why seven independent filters work

    The intellectual content of the seven tests lies not in any single criterion but in their conjunction. Each criterion in isolation is unremarkable. A P/E below fifteen, a P/B below one and a half, a current ratio of two, a dividend record — in any modern stock screener these are standard fields. What makes the Graham screen different is that it requires all seven to clear simultaneously, and the criteria are constructed so that they fail on different kinds of company.

    Consider the failure modes. A young company with no operating history fails tests 3 (ten-year earnings) and 4 (twenty-year dividend). A cyclical at the top of its cycle fails test 5 (three-year-averaged growth) and probably test 7 (P/B will be elevated). A leveraged roll-up fails test 2 (long-term debt versus net current assets). A declining business with melting earnings fails test 5 (growth measured between three-year averages). A glamour stock at the top of a bull market fails tests 6 and 7 (P/E and P/B). A capital-light business that has spent the last decade returning cash via buy-backs rather than dividends fails test 4. A small-cap with limited analyst coverage fails test 1.

    The seven tests are therefore not redundant. They cover seven distinct sources of permanent capital impairment. The probability that a single stock passes all seven by accident is much lower than the product of the seven individual pass probabilities, because the conditions that produce high scores on one test (cheap multiples, for instance) tend to correlate negatively with the conditions that produce high scores on another (long dividend records, strong balance sheets). The conjunction is, statistically, a powerful filter.

    This is why subset versions of the screen — “Graham’s low-P/E screen,” “the Graham Number screen,” “deep-value screens” that drop tests 3 through 5 — tend to perform worse than the full seven over multi-decade windows. Removing any one of the seven dilutes the filter’s ability to exclude a specific failure mode.

    Figure 1. How each successive Graham test reduces the qualifying universe. Each test addresses a distinct source of permanent capital impairment; the conjunction is the filter.
    Figure 1. How each successive Graham test reduces the qualifying universe. Each test addresses a distinct source of permanent capital impairment; the conjunction is the filter.

    3. The empirical record

    The most thorough academic test of Graham’s seven criteria remains Henry Oppenheimer’s “A Test of Ben Graham’s Stock Selection Criteria,” published in the Financial Analysts Journal, September–October 1984. Oppenheimer applied subsets of Graham’s criteria to all stocks traded on the New York and American Stock Exchanges between 1974 and 1981 and computed the realised total return of equally weighted portfolios formed from each subset. The full seven-criteria portfolio produced an arithmetic mean return of approximately 26 percent per annum over the eight-year period, against approximately 14 percent for the equally weighted CRSP universe. The premium was not driven by a single year; the seven-criteria portfolio outperformed the broad market in seven of the eight years.

    Subsequent academic work has corroborated the result while disentangling the source of the excess return. Cliff Asness, Andrea Frazzini, and Lasse Pedersen of AQR Capital Management, writing in their 2015 paper “Quality Minus Junk,” reconstruct a measurable “quality” factor from criteria closely related to Graham’s tests 2 through 5 — profitability, growth, safety, and payout — and find that this quality factor has produced positive returns across twenty-four developed-market equity indices since the early 1980s. The Asness factor is not identical to Graham’s screen, but the overlap is substantial. The economic content of the seven tests has not decayed.

    Jason Zweig’s 2003 commentary in the revised edition of The Intelligent Investor ran the seven-criteria filter against the US equity universe of the early 2000s and produced a sobering finding: the criteria, applied strictly, narrowed the universe to fewer than fifty stocks at most points in the 1995–2002 period. That is, in our view, a feature, not a defect. The screen is supposed to produce a small qualifying universe. A screen that produces five hundred candidates is no longer a filter; it is the market.

    It is worth being precise about what the empirical record claims. The seven criteria do not promise that any individual stock passing them will outperform. They claim that, applied as a portfolio formation rule across many years and many regions, the resulting basket has historically produced returns above the market with downside volatility no higher than the market. That is the only claim a defensive screen is meant to make. It has, on the evidence, made it consistently.

    4. Two historical episodes

    The seven tests are designed to be applied in any market, in any year, but their character is most clearly visible in environments where the qualifying universe either expands or contracts dramatically.

    The United States, 1974. The bear market of 1973–74 cut the S&P 500 by roughly 46 percent peak-to-trough and was the deepest equity drawdown since the 1930s. Graham’s fourth-edition revision was published in the middle of that bear market. He wrote in the new chapter that “the recent market decline has produced, for the first time in many years, a substantial number of stocks that meet our criteria.” Oppenheimer’s 1984 study, which began its sample period in 1974, captured exactly this moment. A defensive investor who applied the seven tests in 1974 and held the resulting portfolio for the next decade compounded at roughly twice the rate of the market while taking less drawdown risk. The screen was not predicting the 1975 recovery. It was simply identifying the population of stocks for which the probability of survival was high and the price paid was modest. Survival and modest price did the work.

    Japan, 1990–2003. After the Nikkei 225 peaked at 38,915 in December 1989 and proceeded to fall to under 8,000 by 2003, the Japanese equity universe became, for over a decade, the global home of the Graham defensive screen. The combination of a long bear market, a culture of strong balance sheets and high cash balances, twenty-year-plus dividend histories at large industrial groups, and price-to-book ratios systematically below 1.0 across hundreds of companies produced more seven-criteria qualifiers than any market in financial history. The screen was widely written about in international value newsletters during these years (Marc Faber’s Gloom, Boom & Doom Report, Tweedy Browne’s investor letters, Fidelity’s overseas analyst notes), and the period offered an unusually clean test of the discipline: an investor who applied the seven tests to Japanese equities in 2003 and reinvested dividends would have substantially out-compounded the Topix index over the subsequent fifteen years. The discipline travelled. It did not need American conditions to work.

    The two episodes are useful precisely because they look so different at the surface. The 1974 case is a normal cyclical bear market in the world’s deepest capital market. The Japanese case is a sui generis fifteen-year deflationary unwinding. The screen produced qualifying baskets in both, and the qualifying baskets produced acceptable long-term returns in both. A discipline that works only in one market structure is brittle. The seven tests, on this evidence, are not brittle.

    Figure 2. The qualifying universe expands when markets fall and contracts when they rise. A small universe is a feature of the screen, not a defect.
    Figure 2. The qualifying universe expands when markets fall and contracts when they rise. A small universe is a feature of the screen, not a defect.

    5. The application framework — three practitioner disciplines

    First, apply the criteria additively, not selectively. The seven tests are a logical conjunction. They are not a buffet from which the investor may pick the four that look most attractive in the current regime. We see this error frequently: investors retain tests 6 and 7 (cheap multiples) because the result feels concrete, drop tests 3 and 4 (earnings stability, dividend record) because they feel old-fashioned, and end up holding cyclicals near the top of their cycles. The screen’s protective character disappears entirely when the conjunction is broken. The discipline is binary: either all seven, or none of them as Graham intended.

    Second, translate the formula to the holding period. Test 5 measures ten-year earnings growth using three-year averages at the beginning and end. That is a thirteen-year measurement window. The screen was not designed for monthly turnover. An investor who applies the seven tests with the intention of trading the resulting basket over twelve months is applying a screen calibrated for a different problem. Practitioners we respect — Walter Schloss, Tweedy Browne, the partners at Marathon Asset Management — have historically held the median position for three to seven years. The screen has a natural holding period embedded in its arithmetic, and an investor whose horizon is shorter than that horizon should expect the screen to underperform for them.

    Third, adjust the size threshold to today’s units, not to today’s stories. The USD 100 million sales threshold of 1973 is roughly USD 700 million to USD 1 billion in inflation-adjusted 2026 terms. It is not a small-cap threshold. Graham himself wrote that the size criterion was the “weakest” of the seven and could be relaxed safely by an investor capable of doing additional balance-sheet work on smaller companies. Tests 2 through 5, by contrast, must not be relaxed. If a candidate company has only seven years of positive earnings rather than ten, the candidate does not pass test 3, and the response is not to lower the threshold — it is to wait three years, or to look at a different company. The numerical thresholds are the discipline; relaxing them is the equivalent of moving a goalpost.

    6. How practitioners actually applied it

    The seven tests have been used by serious long-term equity investors for more than five decades. Two case studies illustrate how the discipline survives in real portfolios.

    Walter Schloss (Walter J. Schloss Associates, 1955–2002). Schloss was a junior analyst at the Graham–Newman Corporation in the late 1940s and went on to run an investment partnership for forty-seven years applying a Graham-style screen. His audited returns, published in Buffett’s 1984 Columbia Business School address “The Superinvestors of Graham-and-Doddsville,” show 15.7 percent annual gross returns over the 1955–1983 period against 11.2 percent for the S&P 500. Schloss did not apply all seven tests mechanically; he placed primary weight on tests 2 (financial strength), 3 (earnings stability), and 7 (price-to-book), with the price-to-book test acting as the binding constraint. He held an unusually large number of positions — often a hundred or more — precisely because his criteria produced a steady stream of qualifying candidates and he refused to concentrate. Schloss ’s case is instructive because it shows that the seven tests can be customised to a practitioner’s temperament without losing their protective character, provided the conjunction of strength, stability, and price remains intact.

    Tweedy, Browne Partners (founded 1920, value-investing house since 1959). Tweedy Browne’s 1992 (updated 2009) white paper “What Has Worked in Investing” is the most thorough academic-style retrospective of Graham-style criteria from a practising firm. The paper compiles forty-four separate empirical studies of Graham-derived screens applied across the United States, the United Kingdom, Continental Europe, and Japan from the 1930s through the 2000s. The conclusion is, in our reading, the single most useful sentence written about the seven tests: “Investments characterised by some combination of low price relative to current asset value, low price-to-earnings ratios, low price-to-book ratios, significant insider buying, and consistent dividend records have produced returns substantially above the broad equity market over long measurement periods in every developed market studied.” The firm itself has applied a near-Graham screen to its Global Value Fund since 1993 with audited returns broadly in line with global equity benchmarks while taking materially lower drawdown.

    The two firms’ experience converges on the same conclusion. The seven tests are robust to customisation, robust to market structure, and robust to changes in the underlying economy. They are not robust to dilution — that is, to the practice of dropping inconvenient tests — and they are not robust to a holding period shorter than the measurement window they were designed for.

    Figure 3. A single-metric P/E screen and Graham's full seven-test conjunction. A wide qualifying universe is a sign that the discipline has been relaxed, not that the market is cheap.
    Figure 3. A single-metric P/E screen and Graham’s full seven-test conjunction. A wide qualifying universe is a sign that the discipline has been relaxed, not that the market is cheap.

    7. Key takeaways

    The defensive investor’s seven tests are a single discipline, not seven separate ones. The conjunction is the filter; dropping any test breaks it.

    The screen was designed in 1973 for a holding period of five to ten years. Applying it on a shorter horizon transfers the screen’s benefit to someone else.

    The qualifying universe will be small — tens of stocks in a normal market, hundreds in a bear market, near-zero at the top of a bull market. A wide qualifying universe is not a feature; it is a signal that the discipline is being relaxed.

    The thresholds (current ratio of two, twenty-year dividend record, ten-year earnings, P/E of fifteen, P/B of one and a half) are not arbitrary. Each test addresses a specific failure mode. Adjustments to the size threshold for inflation are defensible. Adjustments to the others are not.

    The seven tests are implementable on free regulatory data — SEC EDGAR in the United States, Companies House and the FCA register in the United Kingdom, EDINET in Japan, the MCA filings and audited annual reports in India, the official gazettes in Continental Europe. The screen does not require expensive databases. It requires patience, arithmetic, and the willingness to do nothing for long periods.

    The premise behind all seven tests is unchanged from 1949: in any market, in any decade, the investor’s first task is to assemble a population of candidates from which the worst outcomes have already been excluded. Graham’s discipline does that. After fifty-three years, the screen has earned the right to remain the starting point.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Reading the BRSR: India’s ESG Disclosure Framework, What’s Useful, What’s Noise

    Reading the BRSR: India’s ESG Disclosure Framework, What’s Useful, What’s Noise

    If you open the FY2024-25 annual report of any large Indian listed company and scroll past the auditor’s report, the Board’s report, the Form AOC-1 we discussed in the last letter, the corporate-governance report and the management discussion, you will eventually arrive at a long, separately paginated document with a navy or green cover called the Business Responsibility and Sustainability Report, or BRSR. It will run between thirty and a hundred pages. It will be organised under nine numbered principles. It will contain a great many tables of percentages, ratios and “Yes / No” cells. And — like Form AOC-1 a year earlier and like CARO 2020 a year before that — it will be a piece of disclosure machinery that an outside analyst with twenty minutes of training can extract real signal from, and that the median Indian retail investor will never open.

    This letter is about how to read it.

    The argument has three parts. First, what the BRSR actually is — the statutory basis, the structure, what it replaced, and the assurance overlay that has been bolted on since 2023. Second, what is useful in it for an outside analyst — the seven or eight items that genuinely carry information you cannot get elsewhere. Third, what is noise — the items that look quantitative but are not, the items where the disclosure is structurally untrustworthy, and the items whose only function is to satisfy an unreviewed checklist. The framework is denser than IFRS S1/S2, denser than the EU’s CSRD-aligned ESRS for non-listed comparables, and denser than the SEC’s now-stayed climate disclosure rule. Whether that density translates into useful signal is the question.

    The statutory basis, in one paragraph

    The Business Responsibility and Sustainability Report sits in Regulation 34(2)(f) of the SEBI (Listing Obligations and Disclosure Requirements) Regulations, 2015. SEBI’s master circular dated 10 May 2021, supplemented by the operational circular of 11 November 2022 and the BRSR Core circular of 12 July 2023, prescribes the format. The reporting obligation applies to the top 1,000 listed entities by market capitalisation, on a mandatory basis, from financial year 2022-23 onwards. Below that threshold, BRSR is voluntary. The report is required to be included as a separate section of the annual report and tagged in XBRL on the BSE / NSE filing portals. The BRSR replaced the older Business Responsibility Report (BRR), introduced in 2012, which had been mandatory for the top 500 listed entities and which was, frankly, a checklist exercise. The 2021 BRSR is materially more demanding, and the 2023 BRSR Core layer is more demanding still.

    What the document is organised into

    The BRSR is divided into three sections. Section A is general disclosures — the legal name, listing details, employee headcount split by permanent / non-permanent and male / female, locations of operations, products as a percentage of turnover, subsidiary and associate counts, and CSR applicability. It is largely descriptive and largely lifted from filings the analyst already has. Section B is management-and-process disclosures — for each of the nine NGRBC principles, whether the entity has a policy, whether it is board-approved, whether it is publicly available, the web-link, the grievance redressal mechanism, and a self-rated maturity assessment. Section B is, almost in its entirety, structurally weak — more on that below. Section C is principle-wise performance disclosures — the actual quantitative content, organised under the nine principles of the National Guidelines on Responsible Business Conduct, with a mix of essential indicators (mandatory) and leadership indicators (voluntary).

    The nine NGRBC principles, in compressed form: P1 ethics, transparency and accountability; P2 sustainable and safe goods and services; P3 employee well-being; P4 stakeholder responsiveness; P5 human rights; P6 environment; P7 responsible public-policy advocacy; P8 inclusive growth and equitable development; P9 customer value. This taxonomy is older than the BRSR itself — it dates from 2011 — and it is best treated as a piece of architectural scaffolding rather than as a substantively meaningful split.

    What BRSR Core layered on top

    The 2023 circular introduced two important changes. The first was the carve-out of a subset of the BRSR’s Section C — labelled BRSR Core — covering nine attributes and roughly forty key performance indicators (KPIs) that are, in SEBI’s judgment, the items most worth third-party assurance. The nine attributes are: greenhouse gas footprint, water footprint, energy footprint, embracing circularity (waste), enabling gender diversity in business, enabling inclusive development, fairness in engaging with customers and suppliers, openness of business, and gross wages paid to women.

    The nine BRSR Core attributes
    Figure 1. The nine BRSR Core attributes — the subset of Section C carved out by SEBI’s 12 July 2023 circular for mandatory reasonable assurance.

    The second change was a phased reasonable-assurance mandate over BRSR Core. The phasing is: top 150 listed entities by market capitalisation from FY 2023-24; top 250 from FY 2024-25; top 500 from FY 2025-26; top 1,000 from FY 2026-27. The 2024 industry standards from the joint Industry Standards Forum (ASSOCHAM, CII, FICCI) standardised the assurance evidence base. The choice of word matters. “Reasonable” assurance is the higher of the two assurance grades — a positive opinion based on sufficient appropriate evidence — and is the same standard the statutory auditor opines under for the financial statements themselves. The lower grade, “limited assurance”, produces only a negative-form conclusion (“nothing has come to our attention”) and is the standard under which most global sustainability reports are issued. The Indian framework is the more demanding standard.

    BRSR Core assurance phasing 2023-2027
    Figure 2. The four-year roll-out of reasonable assurance over BRSR Core — from the top 150 listed entities in FY 2023-24 to the top 1,000 in FY 2026-27, c. 85% of NSE/BSE market cap.

    The third change, also from the 2023 circular and refined through 2024, was a value-chain disclosure obligation: the top 250 listed entities were required to report Section C Core KPIs for their value chain (suppliers and customers individually contributing two per cent or more of upstream / downstream purchases / sales) from FY 2024-25, on a comply-or-explain basis. As of the most recent circulars, the value-chain reporting layer continues to be deferred year over year in practice — SEBI has consistently extended timelines under industry pressure — but the structural intent stands.

    How the BRSR compares to other jurisdictions

    The cleanest comparison is to IFRS S1 and S2, the International Sustainability Standards Board’s general-requirements and climate standards, mandatory in jurisdictions that have endorsed them — the UK from 2025, Singapore, Hong Kong, Australia from 2025 — and against which the EU’s ESRS, the SEC’s climate rule (stayed in 2024), and the Indian BRSR can all be benchmarked. IFRS S2 prescribes Scope 1, 2 and material Scope 3 emissions with a comply-or-explain timeline; IFRS S1 prescribes governance, strategy, risk-management and metrics-and-targets disclosures across sustainability matters. The Indian BRSR Core covers more ground than IFRS S2 alone (it has nine attributes, not one) but less than the full ESRS suite. The Indian framework is principle-wise rather than topic-wise, which makes it more cumbersome to navigate but more comprehensive on social factors than IFRS S1/S2.

    The EU’s CSRD, mandatory from FY 2024 reporting (large companies in 2025), is the broadest framework — twelve ESRS standards covering twelve topical areas, double-materiality assessment, value-chain disclosure built in — but applies only to EU-domiciled entities and large non-EU groups with EU operations. The American SEC climate rule, finalised March 2024 and stayed by the Fifth Circuit in April 2024, would have required Scope 1, 2 and certain Scope 3 disclosures for large filers but is in litigation limbo. Against this landscape, India’s BRSR is one of the few mandatory, audited, multi-year frameworks now actually in force at scale.

    What is useful — the seven things to actually look at

    What follows is the seven items in a BRSR that, in my experience reading them across roughly forty large Indian listed entities, carry genuine analytical information. Treat the rest as scaffolding.

    Useful item one: Scope 1, 2 and 3 greenhouse-gas emissions year over year

    Section C Principle 6 question 7 reports total Scope 1 and Scope 2 emissions in tonnes of CO2-equivalent, with intensity ratios per rupee of turnover and per unit of physical production where available. From FY 2023-24 onwards, BRSR Core requires that these numbers carry reasonable-assurance certification for the top 150 entities, with the threshold expanding annually. A multi-year run of audited Scope 1 + 2 emissions, intensity-normalised, is the single most analytically useful piece of an Indian BRSR — it lets the analyst track decarbonisation pace against company-stated transition targets and against peer benchmarks.

    Scope 3 disclosure is required only where material and is, in practice, where most large Indian groups still under-report. Where Scope 3 is provided, treat the absolute number with caution and the year-over-year delta with more confidence — methodology drift is large but consistency within a single reporter is usually better.

    Useful item two: Water withdrawal by source and water-intensity ratio

    Principle 6 question 3 reports water withdrawal in kilolitres by source (surface, ground, third-party, sea-water, others) with a water-intensity ratio per rupee of turnover. For water-stressed sectors — textiles, paper, chemicals, beverages, semiconductors, thermal power — this is genuinely useful. The water-source mix tells you something about a plant’s exposure to regulatory action (ground-water depletion notices are now common in Tamil Nadu, Maharashtra and Gujarat) and the multi-year intensity trend tells you whether the operator is improving on a per-rupee basis or merely growing into more water consumption.

    Useful item three: Waste generation and intensity, by type

    Principle 6 question 9 reports waste generation in metric tonnes split by plastic, e-waste, bio-medical, construction-and-demolition, battery, radioactive and other hazardous categories, with the proportion recycled / re-used / safely disposed. For sectors with extended-producer-responsibility exposure under the 2022 plastic-waste-management and 2022 battery-waste-management rules, the disclosure has compliance bite. The metric to watch is the recycled / re-used percentage trend, not the absolute generation number, which scales with volume.

    Useful item four: Female participation rates in workforce, by management level

    Principle 5 question 1 reports the gender split of permanent employees, permanent workers and key management personnel, alongside the gender split of new hires. Indian listed entities have a chronic under-representation problem at senior levels — the median large-cap reports female participation of seven to eleven per cent at board level and substantially lower at executive committee level. The number itself is interesting; the year-on-year trend is more interesting; the dispersion between hiring-pool female percentage and KMP female percentage tells you whether the company is making promotion-pipeline progress or hiring at the bottom and losing at the middle.

    Useful item five: Workplace safety — lost-time injury frequency rate and fatalities

    Principle 3 question 11 reports the lost-time injury frequency rate (LTIFR) per million person-hours worked for employees and for workers, the total recordable work-related injuries, and — critically — the number of work-related fatalities for both employees and workers, separately for permanent and contract / outsourced. For industrials, mining, construction, oil-and-gas and chemicals, this is a non-financial KPI that maps directly to operational discipline. The contract-worker fatality count is the line that gives most signal — it is the variable that companies with weaker contractor-management systems most often disclose poorly on.

    Useful item six: Ratio of remuneration — CEO and median worker, plus board-to-median

    Principle 5 question 3 reports the ratio of the remuneration of the median employee to the remuneration of the Chief Executive Officer, of the Chief Financial Officer, of the Whole-Time Directors and of the Chairperson, and the percentage change year-on-year. This is the Indian equivalent of the SEC’s Item 402(u) pay-ratio disclosure introduced in 2017 under Dodd-Frank Section 953(b), and it is more granular than the SEC version because it disaggregates the executive denominator across multiple roles. For a country with the income-distribution profile India has, the ratio is a useful piece of context — though the analyst should resist the temptation to draw cross-company conclusions from absolute pay-ratios without adjusting for the underlying workforce composition (a company with a large contract workforce can post a misleadingly low ratio).

    Useful item seven: Open complaints — consumer, employee, supplier — and human-rights complaints

    Principle 9 question 3 reports the number of consumer complaints received, the number pending at year-end and the breakdown by category (data privacy, advertising, cyber-security, restrictive trade, unfair trade, other). Principle 5 questions 5 and 6 report sexual-harassment complaints filed under the Sexual Harassment of Women at Workplace (Prevention, Prohibition and Redressal) Act, 2013 (POSH Act), complaints disposed, and complaints pending beyond ninety days. Principle 3 reports grievance redressal mechanisms for employees, workers and contract workers. These four complaint tallies — consumer, POSH, employee grievance, supplier grievance — are the items most worth tracking year-on-year, because the absolute count is less meaningful than the ratio of pending-to-received and the multi-year disposition trend.

    The seventh item — and the analyst’s reward for getting through to Principle 7 — is the public-policy advocacy disclosure, which lists the trade and industry chambers the entity is affiliated to (CII, FICCI, ASSOCHAM, NASSCOM, ICC and so on) along with whether the affiliation is paid and whether the entity has engaged in policy advocacy. This is one of the few disclosures globally that requires a company to admit its lobbying footprint. It is rarely read.

    What is noise

    What follows is the items in a BRSR that I find structurally weak — items where the disclosure looks quantitative but is essentially performative, items where the data definition is so loose as to allow material discretion, and items where the self-rated nature of the disclosure makes year-over-year and cross-entity comparison meaningless.

    Section B in its entirety is mostly noise. The nine-principle policy-coverage table — “Whether the entity’s policy / policies cover each principle of NGRBC and its core elements” — is a Yes / No grid. Almost every large entity ticks Yes for almost every cell. The web-link column is supposed to provide a public URL to the underlying policy; in practice many of these URLs go to corporate-intranet pages or to PDFs of one or two pages that say very little. The self-rated maturity assessment of policies, where present, is similarly self-served.

    The “training on principles of NGRBC” table (Principle 1, question 6 in most years) reports the percentage of board members, KMP, employees and workers trained on the nine principles. Almost every large entity reports figures in the 90-to-100 per cent band. The metric does not survive even a cursory comparison test — the definition of “trained” is loose enough to capture an annual five-minute online-module click-through.

    The “intentional and unintentional spills” table (Principle 6, question 12) reports spills of materials by type. Outside of a small number of oil-and-gas and chemicals reporters, the disclosure is almost universally “Nil”, which on a country-wide basis is structurally implausible.

    The “value-chain emissions” tables as currently reported are noise for most Indian groups, because the methodology is non-standard and the supplier-survey coverage is, in practice, far below the population the entity does business with. Treat the number as directional at best.

    The “research and development investment as a percentage of turnover” line (Principle 2) is structurally arbitrary because the line that distinguishes capex from R&D in Indian accounts varies materially by sector and by reporter.

    The Section C “leadership indicators” — the voluntary indicators sitting under each principle — are a different kind of noise: most large entities skip them, and the entities that do report them are the same entities that already over-disclose elsewhere, so the data set is selection-biased.

    A practitioner’s seven-pass routine for reading the BRSR

    The same seven-pass discipline that worked for Form AOC-1 in the last letter works here. The compression is approximately ten minutes for a first BRSR read, twenty for a thorough one. The passes:

    Seven-pass practitioner playbook
    Figure 3. The seven-pass practitioner playbook for reading a BRSR in twenty minutes — the cells where the signal lives.

    Pass one. Open the BRSR appendix in the annual report. Note the page count. A long BRSR is not necessarily a better BRSR — the longest reports I have read are also among the most boilerplate-heavy. The signal is denser in the BRSRs of operationally heavy industrials and looser in the BRSRs of asset-light services firms.

    Pass two. Jump to Section C Principle 6 question 7 — Scope 1, Scope 2, and if disclosed Scope 3 emissions. Pull the three years of numbers (current year, comparative year, and the figure from last year’s BRSR as a cross-check). Compute the intensity per rupee of turnover and the year-on-year change. Note whether the disclosure carries the reasonable-assurance certificate (BRSR Core overlay).

    Pass three. Stay on Principle 6 — questions 3 and 9. Pull water withdrawal by source and waste generation by type, with the recycled / re-used percentages. For water-stressed and EPR-exposed sectors, these are the second-most-important reads.

    Pass four. Move to Principle 3 question 11. Pull the LTIFR for employees and workers, the recordable injuries, and the fatalities for both permanent and contract. Sit on the contract-worker fatality line for a second.

    Pass five. Move to Principle 5 questions 1 and 3. Pull the gender split at KMP level and the median-employee-to-CEO pay ratio. Compare the gender split at KMP level to the gender split of new hires; the dispersion is the signal.

    Pass six. Move to Principle 9 question 3 and Principle 5 questions 5 and 6. Pull the consumer-complaints disposition, the POSH-complaints disposition, and any pending-beyond-ninety-days counts.

    Pass seven. Move to Principle 7. Read the trade-association affiliation list and the policy-advocacy disclosure. This is the disclosure most likely to surprise you in the next two minutes of any BRSR you ever read.

    Total time, with practice: under twenty minutes per group. The output is a one-page summary on a single BRSR that contains more analytically actionable detail than the equivalent twenty minutes spent on the same group’s standalone P&L.

    The compounding utility

    The reason this matters is that the BRSR is in its third or fourth year of mandatory disclosure depending on market-cap tier, which means the data set is finally long enough to compute year-on-year trends without methodology drift drowning the signal. The assurance overlay — top 150 in FY 2023-24, expanding to top 1,000 by FY 2026-27 — means that within two reporting cycles, the entire mandatory-BRSR universe will carry reasonable-assurance certification on the nine BRSR Core attributes. That is a richer audited ESG data set than any other emerging market produces. It is comparable to, and in some respects more granular than, the audited sustainability disclosure now required in the UK, Singapore and Hong Kong under their respective ISSB-aligned regimes. India is, on this dimension, ahead of the US.

    Whether the audited data set translates into better-priced equity capital for Indian firms with stronger ESG profiles is a separate question — the empirical evidence on ESG-premia in Indian equities, as elsewhere, is mixed and contested. But the data exists; that is the relevant point for the analyst.

    Where the limits sit

    Three structural weaknesses of the framework are worth flagging, because they affect how the analyst should use the data.

    First, the population is still narrow. The BRSR applies to the top 1,000 listed entities — roughly 25 per cent of the listed universe by count and approximately 85 per cent by market capitalisation. Below that tier, equivalent disclosure is voluntary and rarely produced. The framework also does not apply to the unlisted layer of Indian corporate India, which dominates total economic activity. For comparative work across the entire Indian economy, the BRSR is a sample, not a census.

    Second, the assurance scope is narrower than the disclosure scope. The reasonable-assurance overlay applies only to BRSR Core’s nine attributes, not to the full Section C. The bulk of the social and governance disclosure remains self-reported and unverified. That changes how to read each individual data point — Scope 1 and 2 emissions in an FY 2024-25 BRSR from a top-150 reporter is an audited number; LTIFR in the same document is not.

    Third, the value-chain layer remains aspirational. The 2023 circular’s intent was to push BRSR Core reporting upstream and downstream into the supplier and customer base, but the implementation has been deferred year over year. For 2026, treat any value-chain numbers in a BRSR as indicative only.

    Why the global reader should read the BRSR

    If you analyse global businesses with India exposure — and the universe of such businesses is, by 2026, no longer small — the BRSR is the most efficient source of standardised ESG data on the Indian leg of those exposures. A European chemicals group whose Indian subsidiary is in the top 1,000 listed universe (if separately listed) or whose joint venture partner is, will produce a Schedule of disclosures on that Indian operation that materially exceeds what the parent’s CSRD report can capture. For a US investor running a fundamental long book in Indian large-cap equities, the BRSR provides a multi-year audited Scope 1 / Scope 2 series that simply does not exist for most equivalent emerging-market listings.

    The wider point — and it is the same point I have been making across this Indian Market Context series for the last fortnight — is that the architecture of Indian disclosure is denser than the country’s reputation. The auditor’s report under CARO 2020 is denser than the equivalent US auditor’s report. The subsidiary disclosure under Form AOC-1 is denser than the equivalent SEC Item 21 or UK Schedule 4. The segmental disclosure under Ind AS 108 is, in some respects, more demanding than the IFRS 8 equivalent. And the BRSR — with its principle-wise structure, its reasonable-assurance overlay, and its now-multi-year time series — is one of the more substantive mandatory ESG frameworks in force at scale anywhere in the world.

    The question is not whether the disclosure exists. It does. The question is whether the analyst opens it.

    One-line takeaway: The BRSR is most useful when read selectively — the seven items above are where the signal lives; the rest is scaffolding — and the value of the framework grows with each additional year of audited data that accumulates in the public file.

  • Form AOC-1: The One-Page X-Ray of an Indian Group

    Form AOC-1: The One-Page X-Ray of an Indian Group

    If you spend long enough reading annual reports across jurisdictions, you begin to grade them not by what they print but by what the underlying law forces them to print. On that test, Indian listed parents carry a small piece of disclosure machinery that is genuinely rare in the global filing landscape — a single sheet, prescribed line by line by the central government, that itemises the financial summary of every subsidiary, every associate and every joint venture the parent owns. It is called Form AOC-1, it is attached to the Board’s Report, and at large groups it can run to dozens of pages of rows, each row a different entity inside the holding structure. Read carelessly, it is a wallpaper of names and numbers. Read carefully, it is the closest thing an outside investor has to an x-ray of the group.

    The form is not new. It has existed in essentially its present shape since the Companies (Accounts) Rules came into force on 1 April 2014, replacing the older “Section 212 statement” that limped along under the Companies Act 1956. What is new — and what most non-Indian readers have not absorbed — is that the post-2013 regime, taken together with Section 129(3) of the Companies Act 2013 and Indian Accounting Standard 110, produces a disclosure stack denser than what the equivalent SEC filing or Companies House filing requires. The 10-K has Item 21, a flat list of subsidiary names and jurisdictions with no financials at all. The UK Companies Act requires disclosure of subsidiaries in a note to accounts, again typically without per-entity financials beyond an aggregate. IFRS 12 demands summarised financial information for material non-wholly-owned subsidiaries and associates but allows considerable discretion in what counts as “material”. Form AOC-1 demands a fixed schema: thirteen financial line items per subsidiary, every subsidiary, irrespective of materiality.

    That last point is what makes the form interesting. The Indian rule is mechanical. Materiality is not a defence. If the entity is a subsidiary as defined in Section 2(87) of the Companies Act — which captures both equity-control subsidiaries and “more than one-half of total voting power” arrangements — it appears, and its financial summary appears with it. The same is true for associates and joint ventures, captured under Section 2(6) and Section 2(76)(v) respectively, which Part B of the form addresses. For a conglomerate parent with a hundred-plus subsidiaries — and several listed Indian conglomerates carry well over that — the result is a long, dense, mechanically standardised table that, to a sophisticated reader, contains more analytical signal than the consolidated financials themselves.

    This letter walks through what Form AOC-1 contains, how to read it as a practitioner, what it gives you that other jurisdictions do not, and where its limits lie.

    The statutory basis, in one paragraph

    Section 129(3) of the Companies Act 2013 requires every company with one or more subsidiaries — including associates and joint ventures — to prepare consolidated financial statements alongside its standalone accounts. The first proviso to that sub-section adds a separate obligation: the parent shall also attach to its standalone financial statement “a separate statement containing the salient features of the financial statement of its subsidiary or subsidiaries in such form as may be prescribed.” The prescription comes from Rule 5 of the Companies (Accounts) Rules 2014, which names Form AOC-1 as the statutory format and divides it into Part A (subsidiaries) and Part B (associates and joint ventures). The form is signed by directors and certified by the auditors, and any consolidated financial statement filed without it is procedurally deficient. The Ministry of Corporate Affairs has updated the form’s content twice — once via the Companies (Accounts) Amendment Rules 2016 and again in a minor 2018 clarification — but its core schema has held.

    Part A — the thirteen financial fields per subsidiary

    The columns in Part A are these, in order: the administrative columns — serial number; name of the subsidiary; the date since when the subsidiary was acquired (added in 2016 — useful for spotting recent acquisitions); the reporting period of the subsidiary, if different from the parent’s; the reporting currency and exchange rate as on the last day of the financial year (for foreign subsidiaries). Then the financial columns — share capital; reserves and surplus; total assets; total liabilities; investments; turnover; profit before taxation; provision for taxation; profit after taxation; proposed dividend; percentage of shareholding held by the parent.

    That is the entire schema. Thirteen financial line items if you count the percentage shareholding as a financial datum, eleven if you treat it as administrative. The columns are short on context — there is no breakdown of revenue by segment, no detail of what assets comprise, no working-capital schedule — but the columns are present for every subsidiary, every year, and they tie directly to amounts that appear in the consolidated financial statements after intercompany elimination. The arithmetic does not always reconcile to the consolidated numbers because of those eliminations and because the AOC-1 reflects each subsidiary’s own audited accounts before consolidation adjustments. That difference, properly understood, is part of what the form gives you.

    The AOC-1 schema
    Figure 1. The AOC-1 schema — every field Indian law forces a parent to print for every subsidiary, associate and joint venture, grouped by what each one reveals.

    Part B — associates and joint ventures, where consolidation choice gets exposed

    Part B addresses entities the parent does not control outright but in which it holds significant influence — typically, an equity interest of 20% or more, though the test under Ind AS 28 is one of substance rather than threshold. The columns are different: name; latest audited balance sheet date; date associated or acquired; number of shares held; amount of investment; extent of holding percentage; description of how significant influence is established; reason why the associate or joint venture is not consolidated, if applicable; net worth attributable to shareholding as per the latest audited balance sheet; and the profit or loss for the year split into the portion considered in consolidation and the portion not considered in consolidation.

    That last split is what most rewards close reading. An associate’s earnings flow into the parent’s consolidated profit and loss account via the equity method only if the parent has chosen to consolidate that associate. Indian law allows two exceptions to mandatory equity-method consolidation: where the investment is held for sale, and where the associate operates under “severe long-term restrictions” that significantly impair its ability to transfer funds to the investor. Both exceptions are narrow. When a parent uses them, Part B forces it to name the entity, state the reason, and quantify the un-equityed share of profit. That number — the “profit not considered in consolidation” — is, in practice, one of the most useful red flags an outside analyst can extract from any Indian filing.

    What the form gives the analyst that nothing else does

    Set Form AOC-1 next to Item 21 of a US 10-K and the contrast is immediate. Item 21 is governed by Item 601(b)(21) of Regulation S-K: it requires a list of significant subsidiaries with the name and jurisdiction of organisation. That is the entire requirement. Most large US filers print a five-page list of names and US states. There are no financials. A Berkshire Hathaway 10-K does not tell you what GEICO’s individual turnover was; you would extract that, if at all, from segmental notes and from GEICO’s own state-insurance filings. The disclosure obligation simply does not extend to per-subsidiary financial statements.

    The UK Companies Act 2006, read with Schedule 4 of the Large and Medium-sized Companies Regulations 2008, requires more — a note to the accounts must disclose the name, principal place of business and proportion of nominal value of shares held for every subsidiary undertaking — but, again, no per-entity financial summary. A FTSE-100 parent typically presents a multi-page “list of significant subsidiaries” in a note, sometimes with country-by-country tax disclosures appended, but the AOC-1’s column-by-column financial summary has no UK equivalent. The closest match in IFRS-land is IFRS 12, which mandates “summarised financial information” for each subsidiary that has material non-controlling interests and for each material associate and joint venture. The qualifier “material” is doing heavy work in that sentence. In practice, IFRS 12 disclosures tend to surface for a handful of named entities, not for the hundred-plus that an Indian group might list.

    The result is that Form AOC-1 is, for any global investor analysing an Indian listed parent, the densest single source of structural information available in the public file. If you want to understand where the parent’s risk actually sits — what it owns, in what currency, with what profitability, in which jurisdiction — the answer is in this attachment, not in the consolidated income statement.

    Cross-jurisdiction comparison
    Figure 2. Subsidiary disclosure across four regimes — what each one mandates the listed parent to print for each subsidiary it owns.

    A seven-pass practitioner playbook

    A useful AOC-1 reading routine, after roughly two decades of doing this for Indian groups, looks like the following. None of these passes requires more than a spreadsheet and twenty minutes per group; together they will tell you more about the structural shape of an Indian conglomerate than any management presentation.

    The first pass is the count. How many entities are listed in Part A and Part B together? The number is a coarse measure of structural complexity. A clean operating company will have five to fifteen subsidiaries. A holding-company structure routinely shows fifty to two hundred. Anything over two hundred — and several large Indian groups carry this — should produce a follow-up question about why the structure has been allowed to bloom to that size. The answer is sometimes legitimate (legacy demergers, jurisdiction-specific operating requirements, regulated entities that must sit in their own vehicles). It is sometimes less so.

    The second pass is geography. Tag every entity by the jurisdiction implied in its name and its reporting currency. The form does not require a country column, but the reporting currency field is a strong proxy: USD-reporting subsidiaries are typically incorporated in the US or, more often, in a USD-denominated offshore jurisdiction. The list of jurisdictions that recur in the Indian-group context is short and worth knowing — Mauritius, Singapore, the United Arab Emirates, Cyprus, the Netherlands, the British Virgin Islands, and the Cayman Islands. Concentration of subsidiaries in any one of these jurisdictions is not by itself a red flag; many legitimate group structures route through Mauritius for treaty reasons, or through Singapore for regional headquarter purposes. But a concentration that does not match the parent’s operating footprint should produce a question.

    The third pass is currency. A parent reporting in Indian rupees that nevertheless derives material turnover from USD-reporting and EUR-reporting subsidiaries is carrying translation risk that the standalone income statement obscures and the consolidated income statement only partly reveals. The AOC-1 lets you sum subsidiary turnover by currency and ask the corresponding question about hedging policy.

    Seven-pass practitioner playbook
    Figure 3. The seven passes a practitioner makes through Form AOC-1, from coarse structural counts to per-entity impairment risk.

    The fourth pass is the reporting-period column. Indian listed parents are on an April-to-March financial year. Subsidiaries on calendar years, or on local-statutory year-ends inherited from acquisition, will show a different reporting period. Where the period differs by more than six months — which Ind AS 110 generally prohibits but which still appears in legacy structures — the subsidiary’s contribution to the consolidated accounts will be either a fitted partial period or a roll-forward, and the analyst should treat the comparable-period numbers with care.

    The fifth pass is negative net worth. Sort the entire subsidiary list by reserves and surplus. Any subsidiary where reserves and surplus is materially negative — implying accumulated losses larger than equity contributions — is a candidate zombie. Indian parents routinely keep loss-making subsidiaries operating long past the point at which they would have been wound up in a more disciplined jurisdiction, partly because Indian insolvency law is procedurally heavy and partly because the parent prefers to absorb the losses through equity infusions rather than recognise a goodwill impairment. The number to watch is whether the parent has subscribed to additional equity in the current year — a sign that the loss-makers are still drawing capital.

    The sixth pass is turnover concentration. Sort the subsidiary list by turnover. The top three to five subsidiaries usually account for 60% to 90% of the consolidated turnover. Identifying them tells you what the group actually is, in operating terms, regardless of how the parent’s segmental disclosures slice the same revenue. For a holding-company structure, this pass often reveals that the consolidated revenue is in substance the revenue of one or two large subsidiaries, with the remaining entities contributing very little.

    The seventh pass is the investment-to-net-worth comparison. For each subsidiary, take the parent’s investment shown in the parent’s standalone balance sheet and compare it to the subsidiary’s own (share capital + reserves and surplus) on the AOC-1. Where the investment is materially larger than the underlying net worth, the difference is in substance goodwill that the parent has paid above book on acquisition — and which, in consolidated accounts, is either capitalised as goodwill (subject to annual impairment testing under Ind AS 36) or written off through reserves. The standalone investment-versus-AOC-1-equity gap is a quick way to see where impairment risk sits, before opening the goodwill note in the consolidated financials.

    The disclosure quality of an Indian group is not measured by whether Form AOC-1 is present — it always is — but by whether the analyst opens it.

    Limits — what the form does not give you

    A clear-eyed read of Form AOC-1 also requires acknowledging what it does not contain. It does not segment a subsidiary’s revenue by business line or geography; it gives only the aggregate turnover figure. It does not disclose related-party transactions between subsidiaries; for that, the analyst must read the Notes to Accounts and the auditor’s report on standalone-level related-party disclosures. It does not give cash flow, working capital, or debt-equity composition; only the totals. And critically, it does not disclose subsidiary names that the parent has been able to argue are immaterial under any specific carve-out, although the rules do not actually permit such carve-outs for legal subsidiaries — every legal subsidiary, as defined in Section 2(87), must appear, regardless of size.

    The form also does not name promoter-affiliated entities that fall outside the legal subsidiary definition. A company in which the parent holds 18% — below the associate threshold under most readings of significant influence — does not appear in Part B. For that universe, the analyst must triangulate through the related-party transactions note, the shareholding-pattern filings of the affiliated entities themselves, and the SEBI Substantial Acquisition disclosures. Form AOC-1 is comprehensive within the legal-control universe; it is silent outside it.

    Two analytical case-study sketches

    Consider, in purely analytical terms and without any view on the merits of the equity, what Form AOC-1 reveals about Reliance Industries Limited’s consolidated structure: well in excess of two hundred legal subsidiaries across multiple operating verticals — petrochemicals, refining, telecommunications, retail, digital services — incorporated across jurisdictions including India, the US, the UK, the Netherlands, the UAE and Singapore. The subsidiary-by-subsidiary turnover view from AOC-1, summed by vertical, is in some years a more reliable starting point for understanding the group’s revenue mix than the segmental disclosure in the consolidated income statement, because Ind AS 108 segmental reporting reflects management’s internal reporting view, while AOC-1 reflects legal-entity reality. The two views frequently disagree at the margin.

    Consider also, in equally analytical terms, the case of a conglomerate where Part B carried a string of associates incorporated in offshore jurisdictions with reporting currencies in USD and “significant influence” descriptions that did not align with the equity holdings disclosed. That kind of pattern is exactly what the short-selling report on the Adani Group, published in January 2023, drew attention to — and the data the report relied upon to make its initial structural claims was, in large part, exactly the data Indian regulation requires to be printed annually in Part B of Form AOC-1. The structural disclosure was already in the public file. The disclosure was simply not being read.

    Why this matters for the global reader

    The wider point is that India is often characterised, by people who have not done the reading, as a jurisdiction with weak disclosure. The reality is more nuanced. Indian disclosure law in some areas — segmental reporting under Ind AS 108, related-party transactions under SEBI’s LODR Regulations, and the CARO 2020 auditor’s report on top of them — is genuinely demanding. Form AOC-1 belongs in that group. It is more granular than what the US, the UK or general IFRS regimes require, and it sits in the public filing pack of every Indian listed parent.

    What India does have, in places, is a delivery gap between the legal disclosure requirement and the practical readability of the resulting document. Form AOC-1 is sometimes printed in eight-point type, sometimes split across orientations, sometimes presented in scanned image form rather than searchable PDF, and almost always positioned where a casual reader will not encounter it. That is a presentation problem, not a regulatory one. For an outside investor who is prepared to put twenty minutes per group into the seven-pass routine described above, the form yields more structural information than almost any other single page in the file.

    One-line takeaway: Form AOC-1 is the densest source of subsidiary-level financial detail in any major listed market’s standard disclosure pack; the disclosure quality of an Indian group is not measured by whether the form is present — it always is — but by whether the analyst opens it.

  • Probability Weighting and the Fourfold Pattern: Tversky and Kahneman’s 1992 Cumulative Prospect Theory and Why the Long-Term Investor Routinely Misprices Both Tails of the Distribution

    Probability Weighting and the Fourfold Pattern: Tversky and Kahneman’s 1992 Cumulative Prospect Theory and Why the Long-Term Investor Routinely Misprices Both Tails of the Distribution

    Behavioural Finance · Afternoon Edition · 27 May 2026

    In December 1992, in a special issue of the Journal of Risk and Uncertainty, Amos Tversky and Daniel Kahneman published a thirty-page paper titled “Advances in Prospect Theory: Cumulative Representation of Uncertainty.” It revised the 1979 paper that had launched prospect theory and replaced the original probability-weighting machinery with a rank-dependent one that worked for arbitrary outcomes. Buried inside the algebra was an empirical claim with consequences for every long-term investor. Tversky and Kahneman called it the fourfold pattern. Risk attitudes, they showed, are not a property of a person. They are a property of the cell in which the person is standing. Move from a low-probability gain to a high-probability gain — or from gains to losses — and the same person flips from risk-seeking to risk-averse, or back again, without noticing. The mechanism behind the flip is a non-linear function called the probability weighting curve, which over-weights small probabilities and under-weights large ones. The investor who does not see the function distorting his judgement will mis-price both tails of the distribution he is trying to invest in.

    The bias and its canonical citation

    Prospect theory began life as a 1979 Econometrica paper that argued people evaluate gambles relative to a reference point, are loss-averse around that reference point, and treat probabilities not as objective likelihoods but as weights that are systematically distorted. The 1992 paper sharpened the third claim. Cumulative prospect theory introduced a rank-dependent weighting function that can be calibrated from experimental data. Across thousands of choices, Tversky and Kahneman estimated a median weighting function with two properties. First, w(p) is greater than p for small probabilities — a one-percent chance is treated as if it were closer to seven percent. Second, w(p) is less than p for moderate-to-large probabilities — a ninety-nine-percent chance is treated as if it were closer to ninety-four. The function is concave near zero, convex near one, and shaped like an inverted S. The four corners of the resulting decision space — small probability gain, small probability loss, large probability gain, large probability loss — produce four different risk attitudes. Risk-seeking for small-probability gains. Risk-averse for small-probability losses. Risk-averse for large-probability gains. Risk-seeking for large-probability losses. That is the fourfold pattern.

    The inverse-S probability weighting curve, calibrated from Tversky and Kahneman 1992
    Figure 1. The inverse-S probability weighting curve, Prelec single-parameter form with α ≈ 0.65, calibrated from Tversky and Kahneman (1992).

    The mechanism behind the curve

    The cognitive architecture that produces the weighting function has been studied for thirty years. Three forces combine. The first is diminishing sensitivity: the psychological difference between zero and one percent feels larger than the difference between forty and forty-one percent, which feels larger in turn than the difference between ninety-nine and one hundred percent. The endpoints of the probability scale carry a disproportionate weight. The second is category boundary effects: probabilities near zero are mentally categorised as “possible” rather than “impossible,” and the jump from impossible to possible feels qualitatively larger than the slope of the line would suggest. The same is true at the top of the scale, where the jump from “very likely” to “certain” feels qualitatively larger than the underlying gap. Kahneman labelled this last force the certainty effect in the 1979 paper. The third force is limited mental resolution: the mind handles “small,” “medium,” and “large” well, but it does not handle fine gradations of small. A one-in-ten-thousand event and a one-in-a-million event collapse into the same emotional category, which is why a single vivid news story can persuade a person that an event with a six-decimal-place probability is now imminent.

    The combination produces the inverted-S shape. The weighting function is steepest near the endpoints and flattest in the middle. In the middle of the curve — roughly the band from twenty to seventy percent — objective and subjective probability are tolerably close. Outside that band, distortion sets in. And because long-term equity investing routinely asks investors to assess events that live in the tails — the chance a small biotech reaches commercialisation, the chance a turnaround actually turns, the chance a leveraged firm survives a recession, the chance a once-in-a-decade dislocation arrives this decade — the investor spends most of his time in the part of the curve where the function is most distorted.

    The curve also explains a second puzzle that classical finance struggles with. Why are people willing to pay both for a lottery ticket and for an insurance premium against the same expected value? The expected-utility framework requires that a person be either risk-seeking or risk-averse, not both at once. The fourfold pattern dissolves the puzzle. The same individual, looking at a small-probability gain, is risk-seeking and buys the ticket; looking at a small-probability loss, is risk-averse and buys the policy. The lottery counter and the insurance counter live in the same shopping mall because the curve has the same shape on both sides of the reference point. For the equity investor the practical implication is that any thesis written entirely in the language of upside — “the probability the technology works,” “the probability the addressable market is as forecast” — will be distorted upward, and any thesis written entirely in the language of downside — “the probability the regulator forces a recall,” “the probability the litigation produces an adverse ruling” — will be distorted in the opposite direction. The cure is to write the same thesis from both ends, and to compare the two probability statements line by line.

    The empirical record

    Three families of evidence have accumulated. The first is laboratory. Camerer’s 1995 Handbook of Experimental Economics chapter reviewed several hundred replications of the original Tversky and Kahneman experiments. The fourfold pattern is one of the most robust findings in the entire experimental literature; the median weighting function recovered from a fresh sample of laboratory subjects in any decade since 1992 sits within a narrow envelope of the original curve. The Prelec single-parameter form, with α near 0.65, fits both Tversky and Kahneman’s median data and most subsequent re-estimations.

    The second family is field. Alok Kumar’s 2009 Journal of Finance paper “Who Gambles in the Stock Market?” tracked 70,000 retail brokerage accounts and showed that holders of low-priced, high-volatility, positively-skewed stocks — the precise profile predicted by the probability weighting function as attractive to lottery-loving investors — under-performed matched peers by two to three percentage points a year. Barberis and Huang, in a 2008 American Economic Review paper titled “Stocks as Lotteries,” built an equilibrium asset pricing model around the same prediction and showed that lottery-like stocks are systematically over-priced relative to a CAPM benchmark, with a measurable negative excess return. The probability weighting function is now a standard cross-sectional pricing factor in academic finance.

    Three empirical anchors for the probability weighting bias from regulators and academic literature
    Figure 2. Three empirical anchors for the probability weighting bias, drawn from regulator data and academic finance.

    The third family is regulatory. The European Securities and Markets Authority published a product intervention package in March 2018 that banned the marketing, distribution and sale of binary options to retail investors and imposed leverage and margin-close-out limits on contracts for difference. The supporting national competent authority analyses, summarised in the published rationale, showed that between seventy-four and eighty-nine percent of retail accounts trading these products lost money, with average losses per client running from sixteen-hundred euros to twenty-nine-thousand euros. ESMA’s public language was careful and clinical: the products had “structural expected negative return,” the documented retail outcomes were inconsistent with rational pricing, and the “disparity between expected return and risk of loss” warranted permanent intervention. The Financial Conduct Authority in the United Kingdom reached an adjacent conclusion in 2020. In policy statement PS20/10, the FCA prohibited from January 2021 the sale of cryptoasset derivatives and exchange-traded notes to retail clients, citing “extreme volatility,” absence of a reliable valuation basis, and the conclusion that retail consumers could not reliably assess the value and risks of the products. Two regulators, two jurisdictions, the same empirical signature: retail demand was concentrated in instruments whose return profile is structurally attractive to a probability-weighted decision-maker and structurally unattractive to an expected-value one.

    Two historical episodes

    The first episode is the 2020-2021 special-purpose acquisition company cycle in the United States. More than six hundred SPACs were listed in 2020 and 2021, raising roughly one hundred sixty billion dollars. The pitch — a blank-cheque vehicle that would, with some probability, merge with a high-growth private business at a favourable valuation — was a paradigmatic small-probability, large-payoff proposition. The fourfold pattern’s top-left quadrant predicts exactly that kind of demand: retail investors should bid for the SPAC warrants and post-merger shares at prices that over-weight the probability of a transformative deal. The data, once enough mergers had closed to permit measurement, vindicated the prediction. A 2023 review in the Yale Journal on Regulation reported that the 2021 cohort of de-SPAC mergers had lost an average of sixty-seven percent of their value from the de-SPAC price, the 2022 cohort fifty-nine percent, and the combined post-merger excess return relative to the Nasdaq was negative forty-four percent. The cycle was the most expensive demonstration in modern markets of the difference between objective probability and decision weight on low-frequency, high-skew outcomes.

    The second episode is older and structurally similar. Between the late 1990s and 2000, retail investors in the United States poured capital into low-priced internet stocks with negative or non-existent earnings and option-like return profiles. Many of these were micro-cap names with valuations supported entirely by extrapolated traffic metrics. Kumar’s subsequent academic work treated this cohort as a natural experiment in lottery-stock demand, and showed that the cross-section of underperformance among individual investors was concentrated almost entirely in this segment. The 1998-2000 episode and the 2020-2021 episode differ in vintage and technology, but the cognitive signature is the same. Both cycles concentrated retail capital in the precise corner of the outcome space where the probability weighting function is most distorted, and in both cycles the realised long-run return was deeply negative.

    A third regulatory data point is worth recording because it covers a different region and a different instrument. The French Autorité des Marchés Financiers published in 2014 a study of 14,799 active retail accounts trading forex and binary options on French-licensed platforms between 2009 and 2012. The headline finding was that eighty-nine percent of these accounts lost money over the four-year window, with cumulative net losses of approximately one hundred seventy-five million euros against gross winnings of thirteen million for the top decile. The AMF’s subsequent advertising restrictions, followed by ESMA’s 2018 pan-EU intervention, were the response. Three jurisdictions, two continents counting the United States Commodity Futures Trading Commission’s parallel binary-options enforcement actions, and the same empirical curve. The bias is not a feature of any particular market or generation; it is a property of the species.

    The counter-measure framework: three concrete disciplines

    The fourfold pattern does not yield to wishful thinking. Tversky himself observed that decades of laboratory exposure to the bias did not eliminate it in his own subjects, including those who had taught the original papers. What it yields to is procedural pre-commitment — rules that operate before the distorting machinery has a chance to engage. Three disciplines have an evidence base.

    Discipline one: decompose every thesis into base rate times payoff. Before sizing any position whose investment case turns on a low-probability, high-magnitude outcome, the investor writes down two numbers. The first is the probability the thesis implicitly assigns to the favourable scenario. The second is the closest available base-rate from the academic or industry literature. The gap between the two is the investor’s personal probability weighting distortion. If the gap is wide and the investor cannot defend it in writing with case-specific evidence, the position has not yet passed the first gate. This is the discipline behind Howard Marks’s “second-level thinking” and behind Annie Duke’s “thinking in bets” framework, both of which trace, in their methodological roots, to the cumulative prospect theory literature.

    Three procedural disciplines that counter the inverse-S weighting function
    Figure 3. Three procedural disciplines that re-anchor the investor to the base rate of the distribution.

    Discipline two: cap the tail-bet line item. Even when a low-probability, high-payoff thesis is well-defended, the appropriate position size is small. The asymmetry of the weighting function means that the investor will, on average, over-pay for such a position; capping it bounds the cost of being wrong. A defensible ceiling for tail-bet positions in aggregate — the sum of all positions whose investment case depends on a low-probability event — is in the two-to-three percent range of the portfolio. The cap is not a forecast of how often the thesis will pay; it is a recognition that the human brain pricing the bet will be working with a distorted weighting function, and that prudent sizing must include a haircut for that distortion.

    Discipline three: pre-write the kill criteria. The bottom-right cell of the fourfold pattern — risk-seeking under a high-probability of loss — is the one that converts a manageable mistake into a permanent loss. Once a position has moved significantly against the thesis, the same weighting function that overpaid for it on entry now generates risk-seeking behaviour: the investor doubles down, holds longer, refuses to crystallise the loss. The only known antidote is to write down, ex ante, the quantified evidence that would falsify the thesis — the milestone missed, the metric breached, the audit qualification issued, the legal claim filed — and to act mechanically when that evidence arrives. Investors who pre-commit in writing exit losing theses three to six months earlier than investors who do not, with a measurable saving of compounded capital.

    How long-term-equity practitioners addressed it

    Two practitioners are useful anchors. The first is Howard Marks, co-founder of Oaktree Capital. Marks has written for thirty years about asymmetric risk-reward, and his memos return repeatedly to a single discipline: ask what can go wrong before asking what can go right. Marks’s “I-know school” versus “I-don’t-know school” framing, articulated in the 2004 memo of the same title, is operationally a probability weighting counter-measure. The “I-know” investor sees a low-probability favourable scenario, weights it as if it were medium-probability, and pays the corresponding price. The “I-don’t-know” investor treats the same scenario as low-probability because the base rate says so, and pays a price that admits the wide error band. Oaktree’s long history of distressed-debt outcomes is, at the cognitive level, a long history of refusing to pay a probability-weighted price for a tail outcome.

    The second is Warren Buffett, whose insurance operations at Berkshire Hathaway constitute the most studied counter-measure in modern long-term equity investing. Berkshire’s reinsurance business, run for thirty-five years by Ajit Jain, prices catastrophe risk — large losses that occur with low probability — and is profitable on a cumulative basis precisely because it refuses to write business at prices that imply over-weighted small probabilities. Buffett’s annual letters describe the discipline plainly: the underwriter must price what the long-run loss distribution will deliver, not what the present demand curve says the market will pay. Berkshire has walked away from significant blocks of premium when, in Buffett’s phrase, “the price was too low for the risk.” The same logic governs the operating businesses. Berkshire avoids capital-intensive growth bets whose narrative case rests on a low-probability winner-takes-all outcome and concentrates instead on businesses with durable economics and a defensible base rate of survival. The portfolio is built quadrant by quadrant of the fourfold pattern, not against it.

    Key takeaways

    The fourfold pattern is a feature of the human cognitive system, not a flaw of the individual investor. Probability weighting is non-linear in everyone studied. Decades of teaching the bias do not eliminate it. The only working defence is procedural.

    The function distorts both tails of the distribution. Lottery-stock demand and over-priced insurance both follow from the same curve. An investor who has corrected for the lottery side without correcting for the certainty-effect side has done half the work.

    Regulators have already documented the bias at scale. ESMA’s 2018 binary-options and CFD intervention and the FCA’s 2020 crypto-derivative ban are, methodologically, large-sample empirical studies of probability weighting in retail markets. Their findings should be read as evidence about the strength of the bias, not as quirks of a few products.

    The discipline is base rate first, payoff second, sizing third, kill criteria fourth. All four are written down ex ante. The investor who skips any of the four is leaving the weighting function unsupervised.

    Long-term equity outcomes are dominated by the avoidance of the bottom-right quadrant. The risk-seeking response to a high-probability of loss — doubling down, refusing to sell — is what converts a recoverable mistake into a permanent capital impairment. The pre-written kill criterion is the single most underused tool in the long-term investor’s kit.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Spin-Off Investing: Joel Greenblatt’s 1997 Framework and Three Decades of Academic Evidence on Corporate Separations

    Spin-Off Investing: Joel Greenblatt’s 1997 Framework and Three Decades of Academic Evidence on Corporate Separations

    MORNING EDITION — VALUE INVESTING

    In April 1997, a forty-year-old hedge-fund manager named Joel Greenblatt published You Can Be a Stock Market Genius. Eight chapters in, he arrived at what he called “the single best area to look for opportunities” — corporate spin‑offs, the legal separation of a subsidiary from its parent, distributed pro‑rata to the parent’s existing shareholders. He told the reader, with the directness of someone reporting field data rather than offering opinion, that the average US spin‑off between 1963 and 1988 had outperformed the broad market by roughly ten percentage points per year over its first three years of independent trading.

    The claim sounds either too neat or too narrow to take seriously. Too neat because the equity market is the most heavily intermediated capital market on earth, and a ten‑percentage‑point edge that persists for three decades after publication ought to have been arbitraged away. Too narrow because the global pipeline of separations is small relative to the listed‑equity universe, and even a specialist would build a portfolio at the rate of a handful of names a year. Both reactions miss the point. The spin‑off premium is not an anomaly defying market efficiency; it is a predictable consequence of how institutional capital is mandated to behave. And while the flow of separations is small, it is steady, geographically diverse, and concentrated at the moments — conglomerate break‑ups, forced divestitures, post‑merger rationalisation — when the underlying businesses are most likely to be mispriced.

    This morning’s letter is about that framework: where it came from, why the mechanism works, what three decades of academic replication has confirmed, and how a long‑term practitioner ought to translate it into process today.

    1. The Principle and Its Primary Source

    The framework was set out in Chapter 3 of Greenblatt’s 1997 book, “Chips off the Old Stock”. A spin‑off, in legal form, is a distribution by a parent company of the shares of a wholly‑owned subsidiary to the parent’s shareholders on a pro‑rata basis. The parent receives no cash; the shareholders receive a new security whose value reflects the standalone economics of the spun‑off business. In the United States, the transaction is generally structured to qualify as tax‑free to both the parent and the shareholders under Section 355 of the Internal Revenue Code. Comparable structures exist under United Kingdom demerger relief, the German Umwandlungsgesetz, and — adapted with friction — under Section 2(19AA) of the Indian Income Tax Act.

    Greenblatt’s argument, stripped to its load‑bearing claims, was this. First, that the new spun‑off security is, at the moment of distribution, in the wrong hands. The original parent’s shareholder base was assembled to own the parent — for index reasons, for sector reasons, for size reasons, sometimes simply because the parent had been in the portfolio for a long time. The spun‑off subsidiary is, by definition, a different business in a different sector at a different size. A meaningful fraction of the inheriting shareholders are structurally compelled to sell within the first weeks: index funds whose mandate excludes the new security, large‑cap funds that cannot hold the smaller spin, sector funds whose mandate excludes the new industry, retail holders who view the unfamiliar new ticker as a clean‑up trade. Second, that this forced selling is largely price‑insensitive — the selling is driven by mandate, not by valuation, and it lands on a security with a tiny circle of natural buyers, no Wall Street research coverage for at least a quarter, no earnings history as a standalone entity, and a shareholder register that is itself in flux. Third, that the underlying business is, on average, structurally healthier after separation: a sub‑scale subsidiary that had been starved of capital under the conglomerate often emerges with sharper management focus, cleaner economics, equity compensation tied directly to its own performance, and the freedom to make capital‑allocation decisions on its own terms.

    Greenblatt’s recommended process was almost embarrassingly simple. Read the SEC Form 10 — the registration statement filed by the spun‑off entity ahead of its first day of independent trading — in full. Look for three signals: insider ownership and incentives that align management with the standalone business, a balance sheet whose debt load is appropriate to the new standalone cash flows rather than weighted with the parent’s legacy, and a business that, on its own, the patient investor would want to own at the price implied by the post‑distribution market value. Wait through the initial forced‑selling window — typically four to eight weeks — and act when the technical pressure has cleared. The discipline trades activity for asymmetric exposure: nothing is bought until the structural mispricing is visible in the tape.

    2. The Mechanism — Why the Premium Exists at All

    The premium does not exist because investors are foolish. It exists because the rules under which the world’s largest pools of equity capital are managed produce a predictable, recurring window in which a subset of listed securities is mispriced for reasons unrelated to the underlying business. Three structural features explain the mechanism, and all three are still in force three decades after Greenblatt described them.

    The first is the architecture of indexed and benchmark‑hugging capital. Globally, more than half of equity assets under management are now either passively indexed or run against a benchmark with low tracking‑error tolerance, per Morningstar industry data through 2024. When a parent in the S&P 500 or MSCI World Large Cap spins off a subsidiary that lands, on its first day, in the small‑cap index, the funds tracking the larger index are mandated to sell — not by judgement, but by rule. Active funds benchmarked to the large‑cap index face the same arithmetic. The selling lands on a security whose natural buyers — small‑cap dedicated funds — are themselves not yet positioned to take it, because most have not yet been notified by their benchmark provider that the new ticker is in their universe.

    The second is the absence of sell‑side coverage at the moment of listing. Until separation, the standalone subsidiary was a segment in the parent’s annual report. It had no analyst model of its own, no consensus earnings estimate, no broker target price. Analyst coverage usually takes one to two quarters to assemble. Until it does, the spun‑off security trades on Form‑10 filings and management presentations — both available, both requiring work to read. The bias of most short‑term trading capital is to wait for coverage to begin; the result is a window in which the price reflects the marginal seller far more than the marginal informed buyer.

    The third — and the one Greenblatt emphasised most — is the change in management incentives at the moment of independence. Inside a conglomerate, the manager of a sub‑scale division is paid against the parent’s consolidated metrics. Capital is rationed by the corporate office. After separation, the same manager runs a public company. Equity compensation is tied to the standalone share price. Capital can be redeployed without permission from a corporate parent. The same business, with the same assets and the same people, can produce materially different financial results in its first three years of independence — not because it has changed, but because the constraints under which it was being run have changed.

    Three structural forces — forced selling, coverage gap, incentive reset
    Figure 1. The three structural forces — forced selling, sell-side coverage gap, and management incentive reset — combine in the weeks following separation to produce a predictable mispricing window.

    3. The Empirical Record — Three Decades of Academic Replication

    The interesting thing about Greenblatt’s framework is not that he made the claim. The interesting thing is that the academic literature has, with considerable consistency, replicated it across three decades, three continents, and several methodologies. The exact magnitudes differ, the windows differ, and the statistical significance varies by sub‑sample; but the direction is the same.

    The first major academic study was Cusatis, Miles and Woolridge, published in the Journal of Financial Economics in 1993 under the title “Restructuring through Spinoffs: The Stock Market Evidence”. Examining a sample of 161 spin‑offs from US parents between 1965 and 1988, the authors documented a buy‑and‑hold abnormal return of roughly twenty‑five per cent over the three years following the distribution, with the bulk of the excess concentrated in the second and third years rather than the first. Desai and Jain, writing in the Journal of Financial Economics in 1999, extended the sample to 144 spin‑offs through 1995 and identified an important sub‑sample: “focus‑increasing” spin‑offs — those in which the spun‑off business operated in a different two‑digit SIC code from the parent — earned abnormal returns of roughly thirty‑three per cent over the three‑year window, while “non‑focus‑increasing” spin‑offs delivered no statistically significant excess return.

    McConnell and Ovtchinnikov, writing in the Journal of Investment Management in 2004, addressed the most pointed critique of the earlier literature: that the excess returns were a small‑sample artefact driven by a handful of outliers. They tested 311 spin‑offs from 1965 to 2000 and showed that the result survived the removal of the largest outliers; the median spin‑off, not just the mean, outperformed the market. Veld and Veld‑Merkoulova published a meta‑analysis in the International Review of Financial Analysis in 2009, pooling 26 studies across the US, the UK, and continental Europe. They concluded that the average announcement‑date return to the parent of 3.0 per cent, and the post‑separation three‑year abnormal return to the spin of roughly 9.6 per cent annualised, were robust to geography and to time period.

    Two important caveats appear in the more recent literature. The first, documented by Boreiko and Murgia in the European Financial Management Journal in 2016 for European spin‑offs between 1989 and 2013, is that the magnitude of the spin‑off premium has compressed in recent decades. The 1965‑1988 sample on which Greenblatt’s original number rests included a great many cases in which a conglomerate that traded at a meaningful discount to its sum‑of‑the‑parts handed shareholders a corrected valuation simply by separating. As conglomerate discounts narrowed in the late 1990s and as activist investors began pre‑empting many of the most obvious value‑unlocking transactions, the premium available to the patient generalist after a spin‑off compressed. The second caveat, documented by Chemmanur and Yan in the Journal of Financial Economics in 2004, is that the premium concentrates in spin‑offs that are focus‑increasing — separations that meaningfully change the strategic geography of both the parent and the child. Spin‑offs that are essentially financial engineering, in which the same management team and the same capital‑allocation philosophy continue to run both halves, have not, on average, delivered excess returns.

    The honest reading of the literature is therefore this. The spin‑off premium is real; it is global; it is persistent over a thirty‑year window. But it has compressed in magnitude, it concentrates in focus‑increasing transactions, and it requires that the practitioner read the Form 10, separate the genuinely independent spins from the cosmetic ones, and pay attention to insider incentives and balance‑sheet structure. The single number — “ten per cent per year for three years” — was always a textbook average; the practitioner’s number is whatever the practitioner’s discipline can earn within that distribution.

    Three-year abnormal returns to spun-off entities by academic study
    Figure 2. Reported three-year abnormal returns to the spun-off entity by published academic study. The premium has compressed in recent decades but the direction has been replicated across the US, the UK, and continental Europe.

    4. Two Historical Episodes

    Two episodes, separated by a quarter‑century and an ocean, illustrate the framework in action. Neither is offered as a recommendation; both are case‑studies in why the mechanism works and where it can fail.

    The first is the Marriott separation of October 1993. Marriott Corporation split itself into Host Marriott Corporation, which retained the hotel real estate and the historical long‑dated debt, and Marriott International, which retained the asset‑light hotel management and franchising business. The transaction was controversial when announced — Host Marriott carried roughly $2.9 billion of the parent’s senior debt while owning the cyclical real estate; bondholders sued; institutional shareholders who had bought a hospitality compounder found themselves holding a leveraged real‑estate vehicle they had not asked for. In the months following distribution, Host Marriott traded at a substantial discount to its underlying property value, while Marriott International traded close to a fee‑based comparable multiple. Over the subsequent five years, as the US hotel cycle recovered and the debt schedule worked through, the combined value of the two securities materially exceeded that of the pre‑split parent. The episode became the case‑study by which a generation of US analysts learned to read Form 10 carefully and to separate noisy first‑months trading from underlying business value.

    The second is the Mercedes‑Benz Group separation from Daimler Truck Holding AG in December 2021. After more than two decades during which Daimler AG had operated as a single listed entity combining the Mercedes‑Benz passenger car franchise with the Daimler Truck business, the European board concluded that the strategic geographies of the two businesses had diverged sufficiently — battery‑electric passenger vehicles on one side, heavy commercial vehicle electrification and hydrogen on the other — to warrant separation. Daimler Truck Holding AG listed on the Frankfurt Stock Exchange on 10 December 2021; the parent renamed itself Mercedes‑Benz Group AG on 1 February 2022. The forced‑selling pattern was textbook: a number of European large‑cap mandates could not hold Daimler Truck; the new entity began trading without the analyst coverage that had attached to Daimler AG for decades; insider equity grants at Daimler Truck were re‑based to the standalone share price. In the first six weeks of independent trading, the two securities together traded at a discount to the implied pre‑separation value; by mid‑2022, as European institutional capital re‑allocated and as analyst coverage of both standalone entities matured, the discount had narrowed. The episode reads as a European replication of the mechanism Cusatis, Miles and Woolridge had documented in the US three decades earlier.

    5. The Application Framework — Three Process Disciplines

    What follows is not motivational. It is process. The framework is operationally simple and analytically demanding; the simplicity is exactly what makes it survive the cycle.

    Discipline one: build a written watch‑list, refreshed quarterly, of announced and pending spin‑offs in the markets in which the investor is willing to commit capital. The pipeline is finite — globally, between 50 and 100 spin‑offs of meaningful size complete each year, with the largest concentrations in the US and Europe. Anchor the watch‑list to the announcement of the corporate separation, not to the distribution date; the announcement is publicly knowable, often six to twelve months before the actual distribution, and provides the analyst with the runway to read the Form 10 (or its UK / European equivalent) once it is filed. The watch‑list is not a buy list. It is a research queue.

    Discipline two: read the Form 10 in full, and write a one‑page memo to oneself before any decision. The memo answers three questions: whether the standalone business is one the long‑term investor would want to own at any price; whether the balance sheet handed to the spin is appropriate to its standalone cash flows or saddled with disproportionate parent legacy debt; and whether the equity compensation of the standalone management team is aligned with the standalone business’s long‑term performance. If any of the three questions returns a negative answer, the spin is not actionable for the patient investor regardless of how attractive the first‑day price action looks. The memo is the discipline that separates a framework from a tip.

    Discipline three: wait through the technical‑selling window before sizing a position, and size it within a portfolio context. The forced‑selling pattern typically clears within four to eight weeks of distribution. First‑week price action reflects index rebalancing rather than considered valuation; positions established into that window are frequently marked down before the technical pressure clears. The patient investor sizes the position only after the structural sellers have completed their work, and sizes it as a fractional holding within the broader portfolio rather than as a concentrated bet. Spin‑off investing rewards the practitioner who treats each separation as one observation in a portfolio of observations, not as a thesis to be defended.

    Four-step practitioner process for the patient generalist
    Figure 3. The four-step process — watch-list, Form 10, wait, size — that converts the structural opportunity into a disciplined portfolio allocation.

    6. How Long‑Term Practitioners Have Applied It

    Greenblatt himself remains the most documented practitioner of the framework. The Gotham Funds, which he co‑founded with Robert Goldstein, ran an explicit spin‑off strategy through the 1990s and into the 2000s; the firm’s reported audited returns from inception of the original Gotham Capital partnership in 1985 through its 1995 closure averaged approximately 50 per cent gross annualised, a figure that Greenblatt has discussed publicly and is documented in his subsequent writing including The Little Book That Beats the Market (2005). The strategy was concentrated, opportunistic, and explicitly tied to corporate events including spin‑offs, recapitalisations, and rights offerings. Greenblatt has been consistent in both the books and the interviews he has given over thirty years that the spin‑off premium is structural, not stylistic, and that the discipline of reading the Form 10 is what converts the structural opportunity into realised return.

    A second well‑documented practitioner is Murray Stahl of Horizon Kinetics, whose firm has managed long‑duration equity capital since 1994. In the firm’s quarterly commentaries across the 2010s, Stahl developed the thesis that the academic spin‑off premium concentrates in what he termed “owner‑manager spin‑offs” — separations in which the new standalone entity emerges with strongly aligned insider ownership — and effectively disappears outside that sub‑sample. His framework refines Greenblatt’s by asking the practitioner to focus the research budget on spins in which the people running the new business have a meaningful personal stake in its long‑term success, rather than on the full distribution of separations, most of which are organisational rearrangements with limited insider re‑alignment.

    The author of this letter holds no position in any spin‑off or post‑spin entity discussed above.

    7. Key Takeaways

    First, the spin‑off premium is real and globally persistent over a thirty‑year academic record. It is not an anomaly defying market efficiency; it is a structural consequence of how indexed and benchmark‑hugging capital is mandated to behave in the weeks following a corporate separation.

    Second, the magnitude of the premium has compressed since Greenblatt’s 1997 publication, and it now concentrates in focus‑increasing separations in which the parent and the child operate in materially different strategic geographies. Cosmetic spin‑offs and financial‑engineering separations have not, on average, delivered excess returns.

    Third, the framework is operationally simple: build a watch‑list, read the Form 10, write a memo, wait through the forced‑selling window. The simplicity is exactly what makes it executable by a generalist long‑term investor without specialised infrastructure.

    Fourth, the framework rewards discipline over activity. The pipeline is finite, the actionable subset is smaller still, and the patient investor will commit capital to a small number of separations a year. Treated as one process among several in a long‑term equity portfolio, it is one of the few corners of the market in which structural mispricing is documented, replicated, and still available to the practitioner willing to read the filings.

    Fifth, and perhaps most importantly, what Greenblatt’s 1997 chapter actually teaches is a habit of mind that extends well beyond spin‑offs. The habit is to ask, of any apparent mispricing in listed equity, whether there is a structural reason why a particular pool of capital is constrained from acting on what would otherwise be a profitable trade. Wherever mandate, benchmark, or coverage produces forced behaviour at known dates, the patient investor who is not so constrained holds an asymmetric option to act. Spin‑offs are one such corner. There are others. The framework belongs to anyone willing to think about the equity market as a system of constraints as much as a system of prices.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The Law of Large Numbers: Bernoulli’s 1713 Golden Theorem and the Long-Term Equity Investor

    The Law of Large Numbers: Bernoulli’s 1713 Golden Theorem and the Long-Term Equity Investor

    Afternoon Edition · Mental Models · Essay No. 10 · 26 May 2026 · Tallinn

    1. The model

    In Ars Conjectandi, published posthumously in Basel in 1713 from a manuscript Jakob Bernoulli had worked on for at least twenty years, the Fourth Part contains what its author called the Theorema Aureum—the Golden Theorem—and what every modern textbook now calls the (weak) law of large numbers. Bernoulli proved that the relative frequency of a binary outcome, observed across an increasing number of independent trials, converges in probability to the true underlying probability of that outcome. In the language he himself used: if a bag contains a fixed but unknown proportion of white and black pebbles, then the more pebbles you draw with replacement, the closer the observed white-fraction will come to the true white-fraction, with arbitrarily high probability, as the number of draws becomes large.

    The result was the first formal demonstration that observed frequencies tell you something reliable about the world that produced them. It was so important to Bernoulli that he refused to publish the rest of the work without it; his nephew Niklaus eventually edited the manuscript for the 1713 edition, eight years after Jakob’s death. Modern probability theory distinguishes a weak form (Khintchine, 1929—convergence in probability) and a strong form (Borel, 1909; Kolmogorov, 1930—almost-sure convergence). Both say the same thing in plain English: the sample mean of independent and identically distributed random variables with a finite expectation tends to that expectation as the sample size grows. The historian of statistics Stephen Stigler, in The History of Statistics (Harvard University Press, 1986, chapter 2), treats Bernoulli’s 1713 theorem as the foundation stone of the entire frequentist edifice.

    For the long-term equity investor, the one-sentence form is this: the longer you sample a process, the closer your observed average will come to the process’s true expected value—but only if the process you are sampling is stationary, the draws are independent, and the expected value exists. All three of those conditions matter for what an investor can and cannot conclude from a track record. The model is more often misused than used. The investing literature is full of casual appeals to “the long run” that smuggle in unproven assumptions about stationarity. A central purpose of this essay is to separate the law itself from those misuses, and to show how Bernoulli’s theorem, applied with care, becomes one of the strongest pillars of a long-term equity discipline.

    2. The mechanism

    Why does the law work, and what makes it brittle? Consider a portfolio of n independent and identically distributed positions whose individual return X has finite expectation μ and finite variance σ2. The arithmetic mean of the n positions is X̄n = (X1 + X2 + … + Xn) / n. Two elementary facts about the distribution of X̄n drive everything that follows. The expectation E[X̄n] equals μ itself, irrespective of n: the expected value of the average is the true mean. The variance Var(X̄n) equals σ2 / n: the standard deviation of the average shrinks as 1/√n.

    The shrinkage in 1/√n is the mathematical engine. Doubling the sample size reduces the standard error by a factor of √2, not 2. Quadrupling it halves the standard error. This is why convergence is real but slow: getting from a standard error of ten percent down to one percent requires a hundred-fold increase in sample size, not a ten-fold one. It is also why the most common quantitative claim in investing—”this strategy has worked over a five-year backtest”—is, in many cases, a single noisy observation rather than statistical evidence.

    Standard error of the sample mean shrinks as one over the square root of n. Table shows that going from ten observations to one thousand observations multiplies precision by ten, not one hundred.
    Figure 1. Convergence is real but slow: the standard error of the sample mean shrinks as 1/√n. To halve the noise the investor must quadruple the sample, not double it. Source: author’s calculation from elementary sampling theory.

    The mechanism rests on four assumptions, each of which is fragile in real markets. The first is the independence of draws: cross-correlated positions—every Indian small-cap, every European bank stock, every US high-multiple software name—do not provide n independent observations; they provide some smaller effective sample size neff. The second is identical distribution: if the underlying process changes over the sampling window—a regulatory regime change, a structural shift in interest rates, the entry of a new disruptive technology—what looks like one long sample is actually two short samples glued together. The third is finite expectation: for a few important investment-relevant distributions, notably power-law-tailed return distributions in the spirit of Mandelbrot (1963) and Taleb (2020), the theoretical mean exists but the sample mean converges very slowly; for distributions without finite variance, the central limit theorem fails altogether. The fourth is a long enough horizon: convergence is asymptotic, and at any finite n the sample average remains a random variable around the true mean.

    Amos Tversky and Daniel Kahneman, in their 1971 paper “Belief in the Law of Small Numbers” (Psychological Bulletin, vol. 76, no. 2, pp. 105–110), documented that even trained statisticians systematically overestimate the reliability of small samples—they treat n = 20 as if it were the asymptotic case. The behavioural literature has replicated this finding many times since. For the investor, the takeaway is that the law of large numbers is silent at the sample sizes most investors care about; only the law of small numbers is operative.

    3. The empirical record

    The most striking empirical record of the law of large numbers in equity markets concerns the wide dispersion of individual stock returns and the consequently large sample sizes required before broad-market averages stabilise. Hendrik Bessembinder, in “Do Stocks Outperform Treasury Bills?” (Journal of Financial Economics, vol. 129, 2018, pp. 440–457), computed lifetime returns for the universe of CRSP US common stocks from 1926. Of roughly 26,000 individual stocks, just over half (51.6 per cent) delivered lifetime returns below those of one-month Treasury bills. The aggregate equity premium over Treasury bills since 1926 was driven by a small minority: the top 4 per cent of stocks accounted for the entirety of the net dollar wealth creation; the median stock destroyed wealth relative to T-bills. In the global update (Bessembinder, Chen, Choi & Wei, “Long-Term Shareholder Returns: Evidence from 64,000 Global Stocks,” SSRN working paper, 2023), 60.9 per cent of 64,000 international stocks underperformed cash over their lives.

    Bessembinder’s numbers are the law of large numbers in operation. The market-cap-weighted aggregate is well-behaved because n is enormous—tens of thousands of names over a century, with very high effective sample size. The index-level mean reliably reflects the equity-premium expectation. But the same arithmetic implies that a portfolio of ten names is not a meaningful sample of the equity-return distribution. The standard error of the average return of ten randomly-drawn stocks is roughly thirty per cent of the true σ; for one hundred names it is roughly ten per cent. This is the source of Meir Statman’s “diversification ratio” empirical result (“How Many Stocks Make a Diversified Portfolio?”, Journal of Financial and Quantitative Analysis, vol. 22, no. 3, 1987, pp. 353–363): something on the order of thirty stocks captures most diversifiable variance, but residual idiosyncratic risk remains material.

    SPIVA bar chart: share of US large-cap active mutual funds that underperformed the S&P 500 over one, five and fifteen year horizons. Underperformance rises from sixty per cent at one year to nearly ninety per cent at fifteen years.
    Figure 2. As n grows, apparent skill compresses toward zero net of costs. US large-cap active mutual fund underperformance vs. the S&P 500, by horizon, mid-year 2024. Source: S&P Dow Jones Indices, SPIVA U.S. Mid-Year 2024 Scorecard.

    The other empirical anchor is the S&P Indices Versus Active Funds (SPIVA) scorecard, published semi-annually since 2002. The SPIVA U.S. Mid-Year 2024 Scorecard reports that, over the fifteen-year window through June 2024, 89.9 per cent of large-cap actively-managed equity mutual funds underperformed the S&P 500. The single-year figure for the twelve months ended mid-2024 was around 57 per cent. The five-year figure was 77 per cent. As the horizon lengthens—as n grows—the dispersion in apparent skill compresses dramatically, and the share of funds that look skilful approaches the share that one would expect from pure noise net of the cost drag. Both data sources point to the same operating fact for the long-term equity investor: at short horizons, almost anything can happen in the sample mean; at long horizons, structural truths assert themselves. The discipline is not to confuse the two regimes.

    4. Two historical episodes

    4.1 The Nifty Fifty, 1968–1974

    Through the late 1960s and into 1972, a roughly forty-stock set of US growth franchises—Polaroid, Eastman Kodak, Xerox, IBM, Avon, Coca-Cola, Johnson & Johnson, Procter & Gamble, McDonald’s, Disney—traded at price-earnings multiples between fifty and ninety, on the proposition that their durable growth justified essentially any starting multiple. The empirical evidence then cited was their immediate post-war record: roughly two decades of high and apparently stable earnings growth. The argument was framed as a long-run truth.

    It was a short-run sample. The sample period chosen (1949–1969) was a unique structural episode: a US export franchise into a war-flattened world, the bedding-in of the post-war consumer economy, and a long disinflation. When the 1973–74 bear market began and the underlying stagflation revealed itself, the Nifty Fifty stocks fell forty to eighty per cent from peak; several—Polaroid, Avon, Eastman Kodak—never recovered their 1972 highs in real terms. Jeremy Siegel’s two retrospectives (“Valuing Growth Stocks: Revisiting the Nifty Fifty,” AAII Journal, October 1998, and “The Nifty Fifty Revisited,” Journal of Portfolio Management, vol. 21, 1995) showed that the basket as a whole did, eventually, justify its 1972 multiples over thirty years—but only as an aggregate, with extreme dispersion within the basket and decades of underwater holding for many individual names. The episode is the canonical example of treating a small, regime-specific sample as if it were the asymptotic case.

    4.2 Long-Term Capital Management, 1994–1998

    LTCM’s swap-spread and convergence trades were sized using volatility estimates from a roughly five-year sample of post-Maastricht European data, in which sovereign spreads had been gently grinding tighter. The bet was that the empirical volatility of that period was representative of the underlying process. Roger Lowenstein’s When Genius Failed (Random House, 2000) and Donald MacKenzie’s reconstruction in An Engine, Not a Camera (MIT Press, 2006, chapter 8) both document that LTCM’s leverage was calibrated to volatility numbers from a benign regime that excluded both the 1987 crash and the 1998 emerging-market crises that followed. When Russia defaulted on its rouble-denominated debt in August 1998, the realised volatility was an order of magnitude above the modelled volatility; the convergence trades widened rather than converged; and the fund—with capital of $4.7 billion at peak and notional positions over $1.25 trillion—required a $3.6 billion Fed-coordinated bailout to wind down without forcing a systemic event.

    LTCM is not a story about the law of large numbers failing. It is a story about the assumption of stationarity failing. The sample size was, mathematically, adequate for narrow inference; what was inadequate was the assumption that the next draw came from the same distribution as the prior draws. Both episodes—the Nifty Fifty and LTCM—teach the same operating lesson: it is not n that matters, it is whether the n draws come from a distribution that resembles the distribution that will generate the next draw.

    5. Application to long-term equity investing

    Three operating disciplines follow directly from the law of large numbers for any investor with a multi-decade horizon.

    Discipline 1: Concentrate, but ensure enough independent bets to let convergence work. A one-stock portfolio has, by construction, an effective sample size of one. The standard error of its annual return is the standard error of a single name—for individual stocks, that has historically been roughly thirty to fifty per cent per year (Bessembinder, 2018). A thirty-stock portfolio of well-diversified independent exposures has an effective n closer to thirty, and a sample-mean standard error roughly five to six times smaller. The trade-off between conviction (concentrate) and convergence (diversify) is genuine, but the relevant variable is effective n, not nominal n. Forty correlated bank stocks are still one bet. The right test for any new candidate is whether its primary economic exposure is materially different from the exposures already in the book.

    Discipline 2: Demand long horizons before judging skill. The SPIVA data implies that even five-year returns provide weak evidence of skill, because the noise dominates the signal. The relevant unit of sample in investment skill is not the trade or the quarter but the cycle. Michael Mauboussin’s The Success Equation (Harvard Business Review Press, 2012) shows that for activities where luck plays a substantial role, the required sample size to detect a one-percentage-point edge with reasonable confidence is in the dozens of cycles, not the dozens of months. The honest implication is that an investor must judge their own process more by the discipline of the inputs (research depth, position sizing, behavioural restraint) than by the trailing returns of the outputs over any short window.

    Three operating disciplines drawn from the Law of Large Numbers: independent bets, long horizons before judging skill, and a regime-change check before extrapolating.
    Figure 3. Three operating disciplines for the long-term equity investor that follow from Bernoulli’s theorem. The first manages the n; the second manages the time; the third manages the assumption that the process has not changed underneath the data.

    Discipline 3: Distinguish stationary from non-stationary processes before extrapolating. Most investment “rules”—sector beta, factor premia, sovereign spread relationships, currency mean-reversion—are stationarity assumptions wearing the costume of statistical inference. The questions to ask, before applying any historical relationship to capital, are: what regime produced this sample?, what would change the regime?, and would I notice the regime change in time? If the answers are unclear, the sample is short, and the prudent posture is humility about the inference. Warren Buffett’s 1996 owner’s manual injunction—that Berkshire avoids situations where it must “be precise about a number that we don’t really understand”—is, at its root, a statement about non-stationarity: when the data-generating process can shift in ways we cannot anticipate, no amount of historical data delivers asymptotic comfort.

    These three disciplines do not produce a strategy. They produce a posture: the long-term equity investor is one who accepts that her edge is statistical, that statistical edges only manifest over many independent observations, and that the cost of forgetting this is the destruction of the very compounding she was trying to harvest.

    6. How the long-term equity tradition has used it

    Warren Buffett has invoked the law explicitly, if informally, throughout the Berkshire Hathaway chairman’s letters. In the 1991 chairman’s letter (Berkshire Hathaway Inc., 1991 Annual Report, dated 28 February 1992), Buffett described the insurance underwriting franchise as one whose results would, “with a long-enough horizon and a wide-enough underwriting book, revert to the underlying actuarial truth.” The thought is repeated, in different forms, in the 1996 owner’s manual and again in the 2014 letter marking Berkshire’s first fifty years: investment skill manifests across a sample of decades, not a sample of months. Berkshire’s own structure—permanent capital, no redemption pressure, a willingness to hold concentrated positions for thirty years—is engineered to let the law operate without interruption. The insurance float strategy in particular is a literal application of Bernoulli’s theorem: across a sufficiently large book of independent risks, the underwriting result converges to the underlying actuarial expectation, and the float earns a return in between.

    Howard Marks has built much of his published thinking around the same statistical core. The Oaktree memo “Risk” (January 2006) frames investment risk as the distribution of possible outcomes around an expected value, and warns explicitly against treating a small realised sample as evidence about the distribution. In “How the Game Should Be Played” (Oaktree Capital, August 2017) and again in Mastering the Market Cycle (Houghton Mifflin Harcourt, 2018, chapter 1), Marks returns to the same point: a single year, a single trade, a single cycle is one draw from a wide distribution; the investor’s job is to think probabilistically about all the draws that could have happened, not just the one that did. Bernoulli’s theorem is the formal expression of why this discipline matters: the next draw is information, but it is not the truth.

    Charles Ellis, in “The Loser’s Game” (Financial Analysts Journal, July–August 1975, pp. 19–26), made the same argument earlier and in stronger form. Ellis’s central observation was that the proliferation of professional investors and the falling cost of information had moved equity markets from a winner’s game (where skill systematically rewards itself in the short run) to a loser’s game (where the dominant variable is the cost of mistakes). The implication, framed in our terms: in a loser’s game, the long-run statistical result is determined by who can afford to wait for n to become large enough for the mean to assert itself, net of fees and frictions, and who has the temperament to resist acting on small-n signals. The rise of indexed and long-only patient capital in the four decades since is, in a sense, the institutional embodiment of Ellis’s reading of Bernoulli. The intellectual chain from Bernoulli to Ellis to Buffett to Marks is direct. It is not a chain about specific stock picks; it is a chain about what kind of evidence about investment skill it is rational to demand, and on what time scale.

    7. Key takeaways

    The law of large numbers is the formal justification for taking the long view, but it is silent on whether any particular sample is large enough. The honest investor decomposes “long-run” claims into a precise n and a precise assumption about stationarity. Standard error shrinks as 1/√n, not 1/n: doubling a sample halves the standard error by a factor of about 1.41, not 2, and most published track records are at sample sizes where most of the variation is still noise. Independent observations are the input—correlated positions are not; effective n in a portfolio is almost always materially below nominal n, and the first test of any new position is its marginal contribution to true independence. The Nifty Fifty and Long-Term Capital Management are the same mistake in different clothing: both treated a short, regime-specific sample as a description of the underlying process. The long-term equity tradition, from Bernoulli through Ellis to Buffett and Marks, has never been about predicting the next outcome; it has been about earning the right to wait for the law to operate.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Extrapolative Expectations: Why the Future the Investor Forecasts Looks Like the Past the Investor Just Survived

    Extrapolative Expectations: Why the Future the Investor Forecasts Looks Like the Past the Investor Just Survived

    NorthPath Advisory  ·  The NorthPath Letter  ·  Behavioural Finance  ·  Afternoon Edition

    In March 2000, with the NASDAQ Composite at 5,048, the American Association of Individual Investors’ weekly survey recorded a six-month bullish reading of 75 percent — a record at the time. In March 2009, with the same index at 1,269 and having lost three-quarters of its value over the preceding nine years, the same survey recorded a bearish reading of 70 percent. The first reading came at a market top; the second at a market bottom. Neither was wrong about the recent past. Both were precisely wrong about the future. This is not a coincidence. It is the most reliably documented finding in the empirical literature on investor expectations, and it has a name: extrapolative expectations.

    The argument of this essay is that long-term equity investors who do not understand how their own brain forecasts returns will, with reliable regularity, contribute capital at the worst times and withdraw it at the worst times. The mechanism is older than markets and shared across every domain in which humans must predict an uncertain future from a finite past. The cost, when applied to a multi-decade investment horizon, is measurable, persistent, and large. And the defences — three of them, none of them new — are available to anyone willing to write down what they expect before the market tells them what to expect.

    1. The bias: what extrapolative expectations are, and where the idea came from

    An expectation is extrapolative when the forecast of the next period is formed by projecting the trend of the past few periods forward. The phrase predates behavioural finance by a century; Alfred Marshall used it in the 1890s to describe how merchants estimated demand. What is specific to the modern behavioural finance literature is the empirical claim that investor expectations of stock-market returns are extrapolative in a strong sense — that they rise after a market has risen and fall after a market has fallen — and that this pattern is sharp enough to dominate the more deliberate, model-based forecasts that any disciplined valuation framework would generate.

    The canonical citation is Robin Greenwood and Andrei Shleifer, “Expectations of Returns and Expected Returns,” Review of Financial Studies, vol. 27, no. 3 (March 2014), pp. 714–746. Greenwood and Shleifer assembled six independent survey datasets covering the period 1963 to 2011: the Gallup investor survey, the AAII weekly sentiment poll, the Graham–Harvey CFO survey, Investor Intelligence, Robert Shiller’s individual-investor survey, and the University of Michigan consumer survey. The six measures were positively correlated with one another and with the level of the stock market, and were strongly positively correlated with returns over the trailing twelve months. They were, however, negatively correlated with the model-based expected returns implied by every standard valuation framework — dividend yield, earnings yield, the cyclically adjusted price–earnings ratio. In plain language: when objective measures said future returns would be low, investors expected them to be high; when objective measures said future returns would be high, investors expected them to be low.

    The result is not a curiosity of one survey. It survives across countries, across investor types, and across cycle. A related literature finds that the same pattern holds for chief financial officers in the Graham–Harvey survey (Itzhak Ben-David, John Graham, Campbell Harvey, “Managerial Miscalibration,” Quarterly Journal of Economics, 2013) and for institutional fund managers (Nicholas Barberis, Robin Greenwood, Lawrence Jin, Andrei Shleifer, “Extrapolation and Bubbles,” Journal of Financial Economics, 2018). The earlier intellectual ancestor is Werner De Bondt’s 1993 paper “Betting on Trends,” International Journal of Forecasting, which demonstrated that even financial professionals predicted that past trends would continue.

    Six investor-expectation surveys, 1963-2011 — correlation signs
    Figure 1. Greenwood & Shleifer (2014): across six independent investor-expectation surveys, optimism is positively correlated with the past 12 months of returns and negatively correlated with model-based expected returns.

    2. The mechanism: why the brain extrapolates

    The cognitive machinery underneath extrapolative expectations is the same machinery Daniel Kahneman and Amos Tversky identified in 1972 under the label of the representativeness heuristic. When asked to forecast an uncertain quantity, the mind does not solve the prediction problem from first principles. It searches recent memory for a similar pattern and treats the most available match as representative of the underlying distribution. Recent returns are the most available pattern; they are also, by definition, the easiest to retrieve. The output of this process is a forecast that looks like the recent past with the trend extended.

    Three properties of this process matter for investing. First, the heuristic is unconscious and fast. It runs before any deliberate valuation work begins, and it sets the frame within which the deliberate work then takes place. A valuation that disagrees with the extrapolated forecast is registered, at the level of the brain that decides what feels right, as the work of a contrarian — and contrarian conclusions require additional energy to hold. Second, the heuristic ignores base rates. The historical fact that high price–earnings ratios have, on a forward ten-year basis, been associated with lower subsequent returns is a base-rate fact, and base rates lose to vivid recent narratives in almost every laboratory study ever conducted. Third, the heuristic generates over-confidence in the forecast it produces. Subjects in calibration experiments report higher subjective confidence in extrapolative forecasts than in the corresponding base-rate forecasts, even when the latter are more accurate.

    The 2018 Barberis–Greenwood–Jin–Shleifer model formalises this. They build a two-agent market with fundamentalists who price on cash flows and extrapolators who price on momentum. The extrapolators’ demand is not perverse; it is rational given their belief that recent trends will persist. The model reproduces the empirical regularities of bubbles — large run-ups, sharp peaks, slow declines — without assuming any agent is irrational in the strong sense. The bias is in the belief-formation process, not in the trading rule applied to those beliefs. This is the precise diagnosis a long-term investor needs: the failure mode is upstream of the spreadsheet.

    3. The empirical record: what the bias costs

    The most widely cited measurement of the cost of extrapolative expectations is the gap between published fund returns and the dollar-weighted returns actually earned by investors in those funds. Morningstar publishes this gap annually under the title “Mind the Gap.” The 2025 edition, covering the ten years ended December 2024, found that the average dollar invested in US mutual funds and exchange-traded funds earned 1.1 percentage points less per year than the funds themselves earned — equivalent to approximately 15 percent of total return surrendered to timing alone. The gap was wider for sector equity funds (150 basis points) than for allocation funds (10 basis points), which is consistent with the bias: the more volatile the category and the more vivid the recent trend, the larger the extrapolation error.

    The complementary measurement is DALBAR’s Quantitative Analysis of Investor Behavior. The 2025 QAIB report found that in 2024 the average equity-fund investor earned 16.54 percent against the S&P 500 total return of 25.05 percent — a gap of 848 basis points, the second-widest in a decade. Withdrawals from equity funds occurred in every quarter of 2024, with the largest outflows preceding the strongest rebounds. DALBAR’s “Guess Right Ratio,” the proportion of months in which an investor’s net flow direction matched the market’s subsequent direction, fell to 25 percent — statistically indistinguishable from a coin toss biased against the investor.

    The two regulator anchors needed for any responsible treatment of this material illustrate how different jurisdictions have intervened in different ways. In India, the Securities and Exchange Board of India requires every mutual fund advertisement that quotes performance to display, in the principal body of the communication, a warning that past performance is no guarantee of future results, and prohibits the use of performance ranking unless accompanied by methodology and the universe from which the ranking was drawn. The relevant master circular consolidates the requirements first issued in 1996 and tightened in 2018 and again in 2024. In the European Union, the Packaged Retail and Insurance-based Investment Products regulation (Regulation 1286/2014) required the introduction of the Key Information Document, and the European Securities and Markets Authority’s 2021 reform of the PRIIPs Regulatory Technical Standards (Commission Delegated Regulation 2021/2268, in force from 1 January 2023) moved the performance scenarios away from a purely past-data extrapolation toward a model-based set of stress scenarios precisely because the previous regime had been shown to systematically over-promise future returns in periods following a market rally. Both regimes are, at root, regulator responses to extrapolative expectations.

    Time-weighted fund return vs dollar-weighted investor return
    Figure 2. The investor-return gap. Left: Morningstar Mind the Gap 2025 (10 years ended Dec 2024). Right: DALBAR QAIB 2025 (calendar 2024).

    4. Two historical episodes

    The first episode is the United States in 1998 to 2002. The NASDAQ Composite rose from 1,419 at the end of 1997 to 5,048 in March 2000 — a 256 percent gain over twenty-seven months — and then fell to 1,114 by October 2002. The fund-flow data, compiled by the Investment Company Institute, shows that net new cash to US equity mutual funds in calendar year 1999 was USD 187 billion, the largest in history to that point; in the first quarter of 2000 the run-rate accelerated. Net new cash in 2001 turned negative for the first time since 1988. The aggregate dollar that entered the market did so almost exactly at the index peak and departed near the trough. The retail investor who held the average diversified equity fund for the full period earned, on a dollar-weighted basis, an annualised return materially below the time-weighted return reported by the funds — a gap calculated at the time by Bogle Financial Markets Research Center as approximately 7 percentage points per year over the 1996–2002 window.

    The second episode is India in 2017 to 2019. The S&P BSE SmallCap Index rose from 11,460 at the end of 2016 to 20,183 at the close of January 2018 — a 76 percent gain in thirteen months — driven in significant part by retail systematic investment plans into small- and mid-cap mutual funds. Net monthly mutual fund inflows from individual investors in India crossed INR 200 billion in January 2018 for the first time. The index then fell to 11,896 by February 2019, a 41 percent peak-to-trough decline, and to 9,123 by March 2020. Net monthly flows from individuals into equity schemes turned negative in mid-2019 and remained so for most of the next four quarters. The pattern is identical to the 1999–2002 American experience in every respect that matters: the peak of conviction coincided with the peak of price, the trough of conviction coincided with the trough of price, and the dollar-weighted return of the average participant fell short of the time-weighted return of the underlying funds by a margin that was, on a five-year measurement window, several hundred basis points per year.

    5. The counter-measure framework: three concrete disciplines

    The corrective disciplines all share a single design property: they force a forecast to be committed to before the market provides the answer, so that the extrapolative impulse can be observed and discounted rather than acted upon.

    Discipline one — the pre-commitment forecast. Before any purchase, write down a one-page forecast covering five elements: the expected revenue growth rate for the next five years; the expected operating margin at the end of that period; the expected return on incremental invested capital; the multiple at which the business is expected to trade at the end of the holding period; and the implied annualised return at the current price. The discipline is to date the document, file it, and review it on a fixed annual schedule against the realised path. Holdings that have benefited from the realised path are not to be celebrated; the question is whether the original forecast assumptions still apply or whether the price has merely extrapolated the recent trend. This discipline is associated with the formal investment-process literature; the cleanest exposition is in Howard Marks’s The Most Important Thing (Columbia Business School Publishing, 2011), chapter 12.

    Discipline two — the base-rate file. For every category of forecast — revenue growth above twenty percent, return on capital above thirty percent, gross margin expansion across a five-year window — maintain a written record of the empirical base rate from comparable historical situations. Michael Mauboussin has documented these base rates extensively in his Credit Suisse Global Financial Strategies reports, in particular “The Base Rate Book” (September 2016). The discipline is to compare any internal forecast against the documented base rate before submitting it. A forecast more than one standard deviation above the base rate is not necessarily wrong, but the burden of evidence required to justify it must be commensurate with the gap. Extrapolative expectations are precisely the engine that produces forecasts well above base rates; the file is the friction that slows the production.

    Discipline three — the asymmetric review. Schedule the longer and more rigorous of the two annual portfolio reviews for the period after a market or sector decline, not after a rally. This inverts the natural human tendency to scrutinise positions when they are causing pain and to leave them alone when they are providing pleasure. The mechanism it defeats is the asymmetry of attention that extrapolation creates: a position that has compounded at twenty-five percent for three years feels like a thesis confirmed, when it is at least as likely to be a thesis priced. The review need not be punitive; it need only be honest about whether the present price has begun to imply, by way of multiple expansion, a forward path that exceeds the base rate. The asymmetric review is the formalisation of John C. Bogle’s lifelong observation, repeated in Common Sense on Mutual Funds (Wiley, 1999), that “reversion to the mean is the iron rule of the financial markets.”

    Three disciplines that limit extrapolation
    Figure 3. Three disciplines that limit extrapolation: pre-commitment forecast, base-rate file, asymmetric review.

    6. How two long-term-equity practitioners addressed it

    John C. Bogle, the founder of the Vanguard Group and the author of Common Sense on Mutual Funds, treated extrapolative expectations as the principal antagonist of the retail investor and built an institution around the assumption that the antagonist could not be defeated by exhortation. His chosen instrument was the index fund — an investment vehicle that mechanically rebalances away from the most-extrapolated assets and toward the least-extrapolated assets, without requiring the investor to override any cognitive process. Bogle’s own writing was unambiguous about the diagnosis. In the 1999 edition of Common Sense on Mutual Funds, chapter four, he documents that the dollar-weighted return of the average US equity fund investor between 1984 and 1998 had been 7 percentage points per year below the time-weighted return of the funds themselves, and he attributes the gap directly to the chasing of recent performance. The remedy was structural: take the choice of when to act out of the investor’s hands.

    Howard Marks, the founder of Oaktree Capital Management, addressed the same problem from the opposite direction, through the medium of the memo. Beginning in 1990 and continuing for thirty-five years, Marks’s quarterly memos to Oaktree’s clients have systematically described where, in the cycle of investor mood, the market sits at the time of writing. The memos do not contain forecasts of price levels or returns. They describe what Marks calls the “pendulum of investor sentiment” — the alternation between greed and fear, between extrapolation upward and extrapolation downward — and locate the present moment on that pendulum. Marks’s 2018 book Mastering the Market Cycle (Houghton Mifflin Harcourt) collects the analytical framework that the memos rehearse. The discipline is not to predict the turn but to know where in the cycle one is standing, so that the natural extrapolative impulse can be inspected before it is acted on. Both practitioners arrived at the same conclusion: extrapolative expectations cannot be reasoned away, but they can be structurally limited or behaviourally observed.

    7. Key Takeaways

    • Investor expectations of future stock-market returns are systematically extrapolative. Six independent survey datasets covering 1963–2011 show that expectations rise after the market rises and fall after the market falls, and are negatively correlated with model-based expected returns (Greenwood & Shleifer, Review of Financial Studies, 2014).
    • The measured cost of this bias is approximately 1 percentage point per year, or 15 percent of total return over a decade, per Morningstar’s “Mind the Gap” 2025 study. DALBAR’s 2025 QAIB report found an 848-basis-point gap in the single year 2024 alone.
    • The bias is generated upstream of any spreadsheet. It is the output of the representativeness heuristic (Kahneman & Tversky, 1972) operating on recent returns, which means it cannot be corrected by improving valuation work alone.
    • Three disciplines limit its operation: the dated pre-commitment forecast, the written base-rate file, and the asymmetric review that scrutinises winners more than losers.
    • Two practitioner traditions illustrate the structural and the behavioural responses respectively: Bogle’s index-fund architecture removes the discretion that the bias would otherwise capture; Marks’s memos formalise the inspection of one’s own sentiment before any large allocation decision.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Reading Indian Segmental Disclosures: Ind AS 108, the Management Approach, and the Conglomerate X-Ray

    Reading Indian Segmental Disclosures: Ind AS 108, the Management Approach, and the Conglomerate X-Ray

    Indian Market Context  ·  26 May 2026  ·  Issue 8

    The single consolidated income statement of an Indian group tells you almost nothing about its economics. The segmental note tells you almost everything.

    Open the annual report of a large Indian listed company — Larsen & Toubro, Reliance Industries, Mahindra & Mahindra, ITC, Aditya Birla Capital, Grasim, Tata Chemicals, Bharti Airtel — and the consolidated profit and loss statement will show you one number for revenue, one number for cost of materials, one number for employee benefits, one number for finance costs, one number for profit before tax. From that aggregate you can compute a group-level margin and a group-level return on equity, and from those you can construct a plausible-sounding sentence about the business. The sentence will almost always be wrong.

    It will be wrong because almost none of those companies is a single business. Larsen & Toubro is a contracting business, a defence-electronics business, a hi-tech-manufacturing business, an IT-services business (LTIMindtree), a financial-services business, a developmental-projects business (concessions, transmission), and a hydrocarbon-services business — and each of those segments earns a different return on capital, deserves a different multiple, and has a different cyclical pattern. Reliance Industries earns money from oil-to-chemicals, from telecom (Jio), from organised retail, from digital media, from oil-and-gas exploration, and from a financial-services arm being spun off — and each line has a different competitive shape. ITC earns most of its EBIT from cigarettes and a small slice from hotels, paperboards, agribusiness, and FMCG-others — and the FMCG-others segment that the market values at one multiple is being subsidised by the cigarettes segment that the market values at a much lower multiple. To talk about “the ROCE of ITC” without knowing the segmental split is to talk about nothing.

    The instrument that lets the reader pull apart the consolidated number into its economic constituents is Ind AS 108 — Operating Segments, notified by the Ministry of Corporate Affairs in 2015 under Rule 3 of the Companies (Indian Accounting Standards) Rules, and effective for accounting periods beginning on or after 1 April 2016. Ind AS 108 is the Indian convergence of IFRS 8, the international standard the IASB issued in 2006 in convergence with the FASB’s SFAS 131 (now codified at ASC 280) in the United States. The three standards are now substantively identical: the segmental disclosure block in an Indian annual report, a Korean K-IFRS annual report, a UK plc annual report, and a US 10-K is being prepared under the same conceptual framework. This is a fact worth carrying with you, because it means the moment you learn to read one country’s segmental note, you have learned to read every country’s segmental note.

    This letter is a working primer on how to read those disclosures: what the standard actually requires, where the management gets discretion, what the most useful columns are, what the most common red flags look like, and how to translate the segmental table into the four or five economic statements that should be in the head of anyone analysing an Indian conglomerate.

    The architecture: the management approach

    The single most important word in Ind AS 108 is management. The standard requires that the segments a company reports externally be the same segments it uses internally to allocate resources and to assess performance. Paragraph 5 of Ind AS 108 defines an operating segment as a component of an entity that (a) engages in business activities from which it may earn revenues and incur expenses, (b) whose operating results are regularly reviewed by the entity’s chief operating decision maker (CODM) to make decisions about resources to be allocated to the segment and to assess its performance, and (c) for which discrete financial information is available.

    The CODM is a function, not a title. It may be the Managing Director, an executive committee, the Board, or a single Executive Vice Chairman; the standard cares only about who actually receives the internal performance pack and decides where capital and people are pushed. Whoever that is, the external segments must mirror the internal segments. The standard calls this the management approach — and it is a deliberate move away from the older IAS 14 regime, which required entities to report along industry-and-geography axes whether or not management thought about the business that way.

    The management approach has one large virtue and one large risk. The virtue is that the external reader sees the business as the operator sees it: real revenue lines, real cost allocations, real capital deployment. When Mahindra & Mahindra reports Automotive, Farm Equipment, IT Services (Tech Mahindra), Financial Services (Mahindra Finance), and Hospitality as separate segments, that is the structure of M&M’s internal P&L review, and the analyst is reading the same numbers the Group CFO is reading.

    The risk is that management has discretion over how it draws those internal lines. Two managers running structurally identical businesses can produce different segment notes, and a manager who wants to obscure a weak business can reorganise the internal reporting to bury it. The standard contains aggregation rules and threshold rules designed to limit that discretion, but those rules are themselves judgement-laden. Almost every interesting reading of a segmental note is a reading of where the discretion was exercised.

    The five required disclosure lines per segment

    For each reportable segment, the standard requires the entity to disclose, at minimum, the following items if they are included in the measure of segment profit or loss reviewed by the CODM, or if they are otherwise regularly provided to the CODM (Paragraph 23):

    a) revenues from external customers;
    b) revenues from transactions with other operating segments of the same entity (inter-segment revenue);
    c) interest revenue;
    d) interest expense;
    e) depreciation and amortisation;
    f) material items of income and expense disclosed in accordance with paragraph 97 of Ind AS 1;
    g) the entity’s interest in the profit or loss of associates and joint ventures accounted for by the equity method;
    h) income tax expense or income;
    i) material non-cash items other than depreciation and amortisation.

    It then requires disclosure of the amount of segment assets and segment liabilities for each reportable segment if these are regularly provided to the CODM (Paragraph 23 and 24), together with the amount of investments in associates and joint ventures and the amount of additions to non-current assets (essentially segmental capex).

    That is a long list. The line that almost every reader underweights is inter-segment revenue. In a true conglomerate — Reliance selling its own petrochemicals into its own retail business, Mahindra Finance lending to Mahindra dealers, an Aditya Birla cement plant buying captive power from an Aditya Birla power plant — the inter-segment line is the size of the internal economy of the group. If it is large, the consolidated revenue line is materially smaller than the sum of the segmental revenues, and the implicit transfer-pricing decisions inside that elimination are doing real work in shaping which segment looks more or less profitable than another. A reader who looks only at the external-revenue column is reading the group as if those internal linkages did not exist.

    The 10% rule and the 75% rule

    A segment is reportable if it meets one of three quantitative thresholds (Paragraph 13). It must be separately disclosed if:

    a) its reported revenue, including external sales and inter-segment sales, is 10% or more of the combined revenue of all operating segments; OR
    b) the absolute amount of its reported profit or loss is 10% or more of the greater of (i) the combined reported profit of all operating segments not reporting a loss, and (ii) the combined reported loss of all operating segments reporting a loss; OR
    c) its assets are 10% or more of the combined assets of all operating segments.

    The entity may aggregate two or more segments that individually fall below the 10% threshold if they share similar economic characteristics and meet most of the qualitative aggregation criteria in Paragraph 12 — broadly, similar nature of products/services, similar production processes, similar customer types, similar distribution methods, and similar regulatory environments.

    There is then a backstop. Paragraph 15 requires that if the total external revenue reported by the operating segments is less than 75% of the entity’s external revenue, additional operating segments must be identified as reportable until the 75% threshold is met. The intent is to make sure that the reportable-segment disclosures cover the bulk of the business, even where the 10% test produces a fragmented picture.

    The discretion in those rules clusters at two points. The first is the aggregation criterion: the phrase “similar economic characteristics” is doing a great deal of work, and an aggressive interpretation lets a manager hide a low-margin product line inside a high-margin segment. The second is the classification of a segment as a “single reportable segment”: many Indian small-caps and mid-caps decline to give a meaningful segmental disclosure on the ground that the business is “primarily one segment”, and they do so even when the customer base, product mix, or geography would justify a split. Both manoeuvres are visible to the careful reader and both are worth flagging.

    Where the discretion hides: the Unallocated line

    The single most consequential line in an Indian segmental disclosure is usually labelled Unallocated or Corporate / Others. It collects everything that management has not chosen to assign to a segment — typically corporate-office costs, treasury income, finance charges on group-level borrowing, group-level tax, and the assets that fund all of these (cash investments, head-office property, group-level intangibles).

    The size of the Unallocated line is the size of the gap between segmental reality and consolidated reality. A small Unallocated line, sitting at well under five per cent of total revenue and assets, is a sign that the segmental disclosure is doing most of the work and the reader can rely on it. A large Unallocated line — fifteen, twenty, twenty-five per cent of group assets — is a warning that a material slice of the balance sheet has been deliberately removed from segmental view, and that the segmental ROCE numbers you might compute will be flattered (because the assets are smaller) and the segmental margins will be too high (because the costs that produced those assets are sitting in Unallocated rather than against the revenue they generated).

    The careful reader’s first move on opening the segmental note is therefore to read the Unallocated row before reading any segment row. Three questions: What is in it? How big is it relative to the rest of the table? Has it grown faster than the segments? If the answer to the third question is yes, the segmental note has been quietly losing information content over time, and the reader is being asked to take more of the group on management’s word.

    The measurement question

    Ind AS 108 does not prescribe a single measurement basis for segment profit or segment assets. Paragraph 25 requires only that the disclosed amounts be measured on the basis used internally by the CODM, even where that basis differs from the IFRS-conformant numbers in the consolidated accounts. A company that uses an “adjusted EBITDA” internally is permitted (and required) to report that adjusted EBITDA as the segment performance measure externally, with a reconciliation back to the consolidated profit before tax (Paragraph 28).

    That reconciliation column matters. It is the only place in the document where the reader can see, line by line, how internal-management numbers bridge to audited statutory numbers. The bridge contains some combination of: depreciation differences (where the internal accounts use a different useful-life convention), corporate cost allocations not pushed into segments, finance costs and finance income held centrally, exceptional items management has chosen to exclude, and inter-segment eliminations.

    The single most informative consolidated-segmental analytical move is to reverse-engineer the reconciliation. Take consolidated PBT. Subtract the total segment-result column. The difference is the sum of all the items management has chosen not to put into the segments. Divide it into its components. If the corporate-cost allocation is large and growing faster than revenue, the segments are being over-flattered. If the finance-cost line is large and held centrally rather than pushed to segments, the segmental return-on-capital numbers will look better than the consolidated ROCE — and the truer picture is somewhere between the two.

    Entity-wide disclosures: geography and the 10% customer

    The segmental note is followed in every Ind AS 108 disclosure by two entity-wide information blocks that apply regardless of how segments were drawn (Paragraphs 31–34).

    The first is the revenue by geography split — revenue attributed to the entity’s country of domicile and revenue attributed to all foreign countries from which the entity derives revenue. The standard requires only “country of domicile” and “all foreign countries” by default, but most large Indian listed companies break the foreign component into regions (Americas, Europe, Middle East and Africa, Asia-Pacific). For an Indian generic pharma company, an Indian IT services company, or an Indian auto-components exporter, this row is the most important single row in the annual report: it tells you the share of revenue that is rupee-denominated and the share that is dollar-denominated, and therefore the operational currency exposure of the business.

    The second is the major customer disclosure. If revenues from a single external customer amount to 10% or more of an entity’s revenues, the entity must disclose that fact, the total amount of revenue from that customer, and the segment in which that revenue is reported (Paragraph 34). The standard does not require the customer to be named, and almost no Indian filer names them voluntarily. But the fact that a 10%-customer exists is itself a structural fact about the business: a contract manufacturer with one customer at thirty per cent of revenue is a different business from a contract manufacturer with twenty customers each at five per cent. A reader who skips this disclosure is missing the largest concentration risk in the entire annual report.

    A five-step practitioner workflow

    Five passes through the segmental note
    Figure 1. The five passes a long-term equity reader should make through any Indian segmental disclosure note.

    The five passes a long-term equity reader should make through any Indian segmental disclosure:

    Pass one — count the segments and weigh the Unallocated line. Count how many reportable segments there are. Compute Unallocated revenue as a share of total revenue, and Unallocated assets as a share of total assets. If Unallocated assets exceed fifteen per cent of total assets, treat the segmental disclosure as a partial picture rather than a complete one.

    Pass two — read inter-segment revenue. Compute inter-segment revenue as a share of total segmental revenue. If it exceeds ten per cent, the internal economy of the group is doing meaningful work, and the transfer-pricing decisions inside it are material to which segment looks profitable. Read the accounting policy note to see how inter-segment transactions are priced (the standard answer is “at arm’s-length, similar to third-party transactions”; the honest answer is occasionally otherwise).

    Pass three — compute segmental ROCE. For each segment, divide segment result (operating profit, before tax and finance charges) by segment assets. Compare across segments and against the group consolidated ROCE. The segment that earns substantially above the group average is the segment carrying the group, and is usually the segment the market is implicitly paying for. The segments that earn below the group average are using capital that could earn more elsewhere, and the disposition of that capital is a capital-allocation question for the board.

    Pass four — read geography. For an Indian listed company with material export revenue, read the entity-wide geography split alongside the segmental result. A segment that produces fifty per cent of group revenue and is sixty per cent dollar-denominated is structurally a different risk from a segment of the same size that is one-hundred per cent rupee-denominated, and the operational hedge that matters most for the group is in this single row.

    Pass five — read the major-customer row. If a 10%-customer exists, write its share of revenue down. Trace it to the segment it sits in. Ask whether the segment’s competitive position is the company’s, or whether it is the customer’s. A contract manufacturer with a 25% customer is, in important respects, a wholly owned subsidiary of that customer.

    Comparisons that travel: Ind AS 108, IFRS 8, ASC 280

    Three standards, one framework
    Figure 2. Ind AS 108, IFRS 8 and ASC 280 — the convergence on the management approach, the 10% rule, the 75% backstop, geography and the major-customer disclosure.

    For a reader who has worked across jurisdictions, the practical equivalence of the three standards is liberating. Ind AS 108 mirrors IFRS 8 paragraph-for-paragraph; both descended from the FASB’s SFAS 131 of 1997, codified into ASC 280. The three converge on the management approach, the 10% revenue/profit/asset thresholds, the 75% backstop, the aggregation criteria, and the entity-wide disclosures.

    The minor practical differences worth knowing are these. ASC 280 requires US filers to disclose revenues by product or service within each segment in slightly more granular form than IFRS 8 and Ind AS 108 require — and accordingly an Indian filer’s product-revenue mix can be coarser than its US peer’s. The IASB amended IFRS 8 in 2013 to require additional aggregation disclosure (where similar economic characteristics were used to aggregate, the entity must say so and describe the characteristics), and Ind AS 108 carries the same requirement. ASC 280 was amended in 2023 (ASU 2023-07) to require disclosure of significant segment expenses that are regularly provided to the CODM — a disclosure that Ind AS 108 and IFRS 8 do not yet require, although the IASB has a parallel project (the segment reporting amendments tentatively finalised in 2024 for IFRS 18 transition) that will bring the two closer.

    For the long-term investor reading an Indian filer alongside a US peer, the upshot is that the conceptual framework travels but the granularity occasionally does not. Knowing which jurisdiction has the more demanding disclosure on a given line — and reading the more granular peer first to set the standard of what should be inferred about the less granular filer — is a useful discipline.

    The red flags

    Five segmental-disclosure red flags
    Figure 3. Five patterns worth flagging when they appear in an Indian segmental disclosure.

    A handful of patterns are worth flagging when they appear:

    The single-reportable-segment declaration in a company whose customer base, geography, or product mix obviously spans more than one segment. The standard permits a single-segment claim only where the business genuinely operates as one segment for internal-review purposes. A company that bundles a domestic FMCG line and a contract-export pharma line into “manufacturing” because both are “manufacturing” is using the standard incorrectly, and a reader is entitled to be sceptical of the rest of the disclosure for the same reason.

    The growing Unallocated line. If Unallocated assets have grown from eight per cent of group assets to twenty-two per cent over four years, the segmental disclosure has lost information content faster than the business has grown.

    The change of segments without restated comparatives. The standard requires restatement of prior-period comparatives when the segment composition changes (Paragraph 29(b)). A filer that re-cuts its segments and presents only the new period on the new basis, leaving the prior period on the old basis, is making a year-over-year comparison impossible — and almost certainly hiding something on the side that was rearranged.

    The margin spread that is too wide to be real. If one segment earns a thirty-five per cent operating margin and another earns a four per cent operating margin in the same group, the cost-allocation policy between them deserves a read. The numbers can be true; equally they can be the consequence of a corporate-cost line sitting entirely against one segment and not the other.

    The disappearing major customer. If a major-customer disclosure appeared in year one, disappeared in year two, and the segment that customer sat in did not grow, the customer most likely has not departed — the disclosure has been re-cut so that no individual customer now exceeds the 10% threshold. This is a permitted result of normal disclosure rules; it is also worth noting.

    The takeaway

    The consolidated income statement of an Indian conglomerate is a press-release object. The segmental note is a working document — the same internal numbers that the Group CFO and the CODM review every quarter, pushed outwards under the discipline of an Ind AS standard that is substantively identical to the IFRS and US frameworks. Reading it as a primer for the business — Unallocated first, inter-segment second, segmental ROCE third, geography fourth, major customer fifth — turns five minutes spent in a footnote into the single most efficient pass through any Indian annual report.

    The companies that draw their segments cleanly and report them honestly are the companies whose management actually thinks in those segments. The companies that hide behind a single-segment claim or a swollen Unallocated line are usually the companies whose internal management is hiding from itself as much as from the reader. Either way, the segmental note is telling you which kind of company you are looking at.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Expectations Investing: How Alfred Rappaport and Michael Mauboussin’s 2001 Framework Asks the Long-Term Investor to Read the Price as a Forecast

    Expectations Investing: How Alfred Rappaport and Michael Mauboussin’s 2001 Framework Asks the Long-Term Investor to Read the Price as a Forecast

    The NorthPath Letter · Value Investing · Morning Edition · 26 May 2026

    Every quoted stock is a sentence the market is whispering to you, in a language most investors never bother to translate. The number on the ticker is not an answer. It is a forecast — a discounted stream of expected future cash flows that the marginal buyer and seller have, between them, agreed to. The disciplined long-term equity investor’s first task is not to build their own forecast and argue with the price, but to read the price as a forecast, decompose it into its operating assumptions, and ask one quiet question: are those assumptions achievable, conservative, or heroic? That inversion of the conventional analytical reflex — start from price, work backward to operating drivers, then test — is the heart of the framework Alfred Rappaport and Michael Mauboussin set out in their 2001 book Expectations Investing. Twenty-five years on, in a market saturated with macro narrative and quarter-by-quarter melodrama, the framework has lost none of its discipline and very little of its edge.

    1. The principle: read the price, then test it

    Most equity research, even today, follows a familiar choreography. The analyst projects revenue, margins, and capital intensity over a five- to ten-year horizon, discounts the resulting free cash flows to a present value at a chosen weighted average cost of capital, compares that value with the prevailing market price, and renders a verdict. Buy if intrinsic value exceeds price; sell or avoid otherwise. The procedure has the comforting appearance of rigour. In practice, it is almost always defeated by the single weakness it cannot acknowledge: the forecast that anchors it is the analyst’s, and the analyst has no demonstrable edge in forecasting.

    Rappaport and Mauboussin’s framework reverses every step. The market price is taken as given, not contested. The analytical work is to derive — by algebra — the specific combination of sales growth, operating margin, incremental investment rate, tax rate, cost of capital, and forecast horizon that would justify that price exactly. This combination is called the price-implied expectations, or PIE.

    The investor’s edge, on this view, comes from being a better judge of whether the PIE is realistic than the marginal market participant. The work shifts from forecasting (a skill in which most investors have no demonstrable edge) to forensic interrogation of consensus assumptions (where a patient, primary-source investigator can develop one). Rappaport and Mauboussin trace the intellectual lineage of this idea to Alfred Rappaport’s earlier work on shareholder value (Rappaport, Creating Shareholder Value, 1986), which articulated the seven value drivers that any discounted cash-flow model rests on. The 2001 book made those drivers the language in which the implied forecast is decoded. The 2021 revised edition, published by Columbia Business School Publishing, updated the worked examples for the post-2008 environment of compressed discount rates and intangible-heavy balance sheets without altering the procedural spine.

    Seven Rappaport value drivers donut chart
    Figure 1. The seven value drivers — the grammar in which every market price implicitly speaks.

    2. The mechanism: why surprises, not levels, drive returns

    The empirical regularity that gives expectations investing its purchase is straightforward. Over short and intermediate horizons, a stock’s return is dominated not by how the company performs in absolute terms, but by how its performance compares with what was already priced in.

    A business growing revenues at twelve per cent a year is a good business. If the market had been pricing in fifteen per cent, the stock will fall on a quarter that delivers twelve. A business growing at six per cent is a mediocre business. If the market had been pricing in three, the stock will rise on a quarter that delivers six. The level of performance is not the operative variable. The change in expectations is.

    This is why Mauboussin, in a series of papers written first at Credit Suisse and later at Counterpoint Global, returns again and again to a single metric — what he calls the expectations infrastructure: the small set of operating drivers (volume, price, mix, operating leverage, returns on incremental capital) that explain almost all of the change in market-implied free cash flow. If those drivers move favourably relative to expectations, the stock works. If they move unfavourably, even good absolute performance is insufficient. The pattern holds across geographies and across decades. Studies of US, European, Japanese and emerging-market equity returns since the 1980s consistently show that earnings surprises explain more of the cross-sectional variation in twelve-month forward returns than do absolute earnings growth rates.

    For the long-term equity investor, the implication is liberating rather than constraining. Most short-term noise is the marginal participant repricing their expectations on a single quarter of news. The investor who has already done the work of decomposing the PIE and forming an independent view of where each driver will land over five to ten years is not buffeted by that repricing. They are, in many cases, on the other side of it.

    3. The empirical record: anchoring, drift, and the inadequacy of multiples

    The case for expectations investing rests on three documented regularities, each of which has been replicated in the academic literature across multiple decades and multiple markets.

    The first is anchoring. Lichtenstein, Slovic, Fischhoff and Phillips, in their 1982 review in Kahneman, Slovic and Tversky’s Judgment under Uncertainty, showed that human forecasters anchor on the most salient prior estimate. In equity markets, that prior is almost always last quarter’s reported number or consensus estimate. Analyst forecasts cluster within narrow bands around prior outcomes (Easterwood and Nutt, 1999; Cen, Hilary and Wei, 2013). The marginal market participant is, statistically, the analyst — and the analyst is anchored. A disciplined PIE decomposition gives the investor a chance to spot expectations that the consensus has accepted simply because they sit close to the recent past.

    The second is post-earnings-announcement drift, first documented by Ball and Brown in 1968, replicated across five decades and almost every major market, and surveyed comprehensively by Bernard and Thomas (1989). Stocks that surprise consensus to the upside continue drifting upward for sixty trading days; stocks that surprise to the downside continue drifting downward. The mechanism is the slow updating of expectations: the marginal market participant takes weeks, not minutes, to reformulate their model. The disciplined PIE investor, who already holds a more accurate view of operating drivers, can position before the drift compounds.

    The third is the documented inadequacy of price-to-earnings multiples as a stand-alone valuation tool. A trailing P/E of twelve can be cheap (if the company is in a stable industry with twelve per cent return on incremental capital and no growth) or expensive (if the company faces structural decline). A P/E of forty can be expensive (if growth is decelerating) or cheap (if growth is accelerating and incremental returns on capital are high). Mauboussin’s 2014 Credit Suisse note What Does a Price-Earnings Multiple Mean? demonstrates algebraically that the P/E ratio is itself a compressed statement of price-implied expectations — and that comparing P/Es across companies without unpacking those expectations is meaningless. Expectations investing forces the investor to do the unpacking explicitly.

    To these three regularities one might add a fourth, drawn from the Bessembinder long-tail literature that has now been replicated through 2023: a small minority of stocks account for nearly all aggregate equity wealth creation, and those stocks are characteristically ones where the price-implied expectations at the start of the holding period were materially below the realised trajectory. Expectations investing is the procedural counterpart of the long-tail empirical fact: it is the method by which an investor positions for the asymmetric outcome rather than the average one.

    4. Two historical episodes where the principle was visible

    The Cisco Systems episode of March 2000 is the textbook illustration. At the peak, Cisco traded at a market capitalisation of approximately five hundred and fifty billion United States dollars, with consensus implying revenue growth of more than thirty per cent annually compounded for the following decade. A PIE decomposition by an analyst willing to do the arithmetic would have shown that, even granting generous operating margins, Cisco would need to be capturing essentially all of the global enterprise networking spend by 2010 to justify the price. The required share-of-spend was not achievable; the price was, in the Rappaport-Mauboussin sense, pricing a heroic outcome. Over the following two and a half years, the stock fell by approximately eighty-eight per cent, even as the business itself continued to grow revenue and protect margin. The fall was not a business failure. It was an expectations realignment — and a realignment that was visible in the arithmetic at the peak, to anyone willing to look.

    The opposing example is Amazon in late 2001 and through 2002. At points during the post-dot-com winter, Amazon’s market capitalisation implied that operating margins would never expand above one or two per cent, that revenue growth would taper to single digits within five years, and that the company’s investment in fulfilment infrastructure would never earn an adequate return. A PIE decomposition done in 2002 — using Rappaport and Mauboussin’s framework explicitly — would have revealed that the marginal investor was extrapolating short-term losses into permanent unprofitability. Bill Miller, then managing Legg Mason Value Trust and a notable practitioner of expectations-investing language, built a position on exactly that logic. The five-year return on that position exceeded six hundred per cent — not because the business surprised on an absolute basis, but because operating margins and capital efficiency surprised relative to the bombed-out implied trajectory.

    A third compact illustration, drawn from the European market, is the Anheuser-Busch InBev sequence from 2018 to 2020, when the brewer’s price implied permanent revenue stagnation and balance-sheet stress; subsequent margin recovery and methodical deleveraging surprised that implied trajectory positively, and the stock recovered roughly seventy per cent from its 2020 lows over the following twenty-four months. The same arithmetic, applied across continents and decades, identifies the same kind of opportunity: a price that has compressed expectations below what the operating reality can deliver.

    Expectations gap diptych — Cisco March 2000 and Amazon late 2002
    Figure 2. Two episodes in which the arithmetic disagreed with the marginal market participant.

    5. The application framework: three concrete disciplines

    The temptation, when introduced to expectations investing, is to treat it as a mental attitude — a sensibility, a way of thinking. It is, in fact, a procedural discipline that can be written down and followed by anyone willing to maintain a working spreadsheet and a primary-source library.

    First, decompose the price into a PIE. The investor begins with the market capitalisation plus net debt — enterprise value. From this, they subtract the present value of the next several years of consensus or analyst-built free cash flow forecasts, then solve algebraically for the residual: the long-term growth rate, the terminal margin, or the value-growth duration that would close the gap. Practical implementations require a discounted cash flow spreadsheet with the seven Rappaport drivers (sales growth, operating margin, incremental fixed investment rate, incremental working-capital investment rate, cash tax rate, weighted average cost of capital, and competitive-advantage period or forecast horizon) as adjustable inputs. The investor varies one driver at a time, holding the others at base-case values, and notes the level at which the price is exactly justified. The output is a small table: “the price assumes ten per cent sales growth for fifteen years, OR twelve per cent operating margin in steady state, OR forty per cent return on incremental capital.” Three sentences, derived from arithmetic, that pin down what the market is actually saying.

    Second, test each PIE assumption against the operating evidence. This is where Fisher-style scuttlebutt, regulatory filings, capital-cycle data, and primary-source competitor analysis re-enter. If the PIE requires fifteen years of fifteen per cent sales growth, the investor asks: what is the addressable industry growth rate, what share would the company need to gain, what is the historical maximum share for any single competitor in this category, and what would have to be true at the customer or product level for that share-gain to materialise? If the PIE requires twelve per cent operating margins permanently, the investor compares against the highest margin ever achieved in the industry, the margin trajectory of comparable businesses across cycles, and the structural reasons margins might be defensible against the entry of new capacity. The discipline is not to forecast better. It is to test whether the consensus forecast embedded in the price is achievable, conservative, or aggressive.

    Third, identify the catalysts that would cause the PIE to revise. Stocks rerate when one or more of the seven value drivers moves materially relative to expectations. The disciplined investor pre-specifies which observable events — a regulatory approval, a competitor exit, a price move, a capital-cycle inflection, a balance-sheet repair — would cause each driver to revise upward or downward, and in roughly what magnitude. They then monitor those specific events rather than the daily price. Position sizing follows from how asymmetric the catalyst set is: if the realistic distribution of revised PIE outcomes is heavily skewed upward, the position is sized larger; if symmetric or skewed downward, smaller or not taken at all.

    These three disciplines — decompose, test, monitor — turn a vague philosophical preference for “the market is wrong” into a procedure that can be applied consistently across hundreds of names over decades. That repeatability is what separates a framework from an opinion, and it is what allows the practitioner to keep working through periods of personal doubt and market turbulence without abandoning the method.

    Three procedural disciplines: decompose, test, monitor
    Figure 3. The three procedural disciplines — repeatable across names, quarters, and decades.

    6. How long-term-equity practitioners actually applied it

    Bill Miller, while running Legg Mason Value Trust between 1991 and 2005, beat the S&P 500 for fifteen consecutive years — a record without precedent in publicly tracked mutual fund history, and one that has not been equalled since. Miller’s process, as documented in Janet Lowe’s The Man Who Beats the S&P (2002) and in Miller’s own quarterly letters, was an explicit application of expectations investing. Miller would identify stocks where, by his arithmetic, the price-implied expectations were materially lower than the realistic trajectory, and hold them through the painful repricing window that follows the catalyst. His Amazon position, his AOL position, and his Dell position in the early 1990s were all built on PIE decomposition rather than absolute-valuation work. Miller wrote and spoke about Mauboussin’s and Rappaport’s framework explicitly through the 1990s and 2000s, often citing the 2001 book by name.

    Howard Marks, in his memo “Risk” of January 2006 and in his book The Most Important Thing (2011), discusses the same arithmetic in different language. Marks emphasises that an investor’s return is determined not by what they buy but by what they pay relative to what was expected — and that the price embeds the expectations. Marks’s preference for second-level thinking is, structurally, expectations investing: ask not what the company will do, but what the company will do versus what the market thinks it will do. Marks has returned to this point in memos across two decades, including his 2018 piece on the limits of macro forecasting, and his 2022 memo on the sea change in interest rates, where he framed the rate environment itself as a shift in the implied discount rate baked into every equity price.

    Warren Buffett, in the 1992 Berkshire Hathaway letter, sets out what is in substance an expectations decomposition. Discussing the question of whether a stock is dear or cheap, Buffett writes that the only honest test is to compute the discount rate that, applied to the stream of cash the business will produce over its life, equates the present value of those cash flows to today’s price. That discount rate is the rate of return the market price implies. If the investor believes the cash flows will be higher or arrive sooner than the market is pricing, the implied return overstates the achievable return; if lower or later, it understates. The investor’s task is the comparison.

    Michael Mauboussin, in his Counterpoint Global notes since 2020, has continued to argue that expectations investing is the only framework that survives contact with empirical asset-pricing data. His 2014 note What Does a Price-Earnings Multiple Mean? decomposes the P/E into its expectations components and shows that almost all the cross-sectional variation in P/E across firms is explained by differences in implied growth and incremental returns on capital, not by differences in “quality” or “story” in any vaguer sense. His more recent work on intangible-heavy businesses applies the same arithmetic to the question of how to capitalise research, development and customer-acquisition expenditure when reading the price of a software or platform business.

    These four practitioners — Miller, Marks, Buffett, Mauboussin — converge on the same procedural insight from very different starting points. The convergence is what gives the framework its weight. Each of them, in different vocabularies, is doing the same thing: refusing to argue with the price until they have first decoded what the price is saying.

    7. Key Takeaways

    The market price of a quoted equity is an embedded forecast, and the long-term investor’s first task is to translate it into specific operating assumptions before having an opinion on whether the stock is dear or cheap.

    Stocks rerate on changes in expectations, not on absolute performance levels; the investor who has already decomposed the price-implied expectations is structurally positioned to be on the right side of those changes.

    Decomposing a price requires arithmetic, not insight: a discounted cash flow model with the seven Rappaport value drivers, varied one at a time, will reveal what the market is actually saying. The insight comes in the next step — testing those assumptions against the operating evidence.

    The discipline replaces forecasting (where most investors have no edge) with the forensic interrogation of consensus (where a patient primary-source investor can develop one); this is the only edge that compounds reliably over decades.

    Expectations investing is not contrarian by temperament. It is contrarian by accident, when the arithmetic happens to disagree with the marginal market participant. The investor’s loyalty is to the arithmetic, not to the contrarian posture — and over time, that loyalty is what produces the asymmetry of outcomes the Bessembinder long-tail data so vividly describe.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Bayes’ Rule: Thomas Bayes (1763) and the Long-Term Investor

    Bayes’ Rule: Thomas Bayes (1763) and the Long-Term Investor

    Afternoon Edition — Mental Models · Essay No. 7 · 25 May 2026 · Tallinn

    1. The model: a posthumous paper that quietly reorganised how we should think

    Thomas Bayes was a Presbyterian minister and amateur mathematician who lived in Tunbridge Wells and died in 1761. He published almost nothing in his lifetime. Two years after his death, his friend Richard Price submitted his notes to the Royal Society. The paper appeared in 1763 under the unassuming title An Essay towards solving a Problem in the Doctrine of Chances, in volume 53 of the Philosophical Transactions, pages 370 to 418. It was largely ignored for the next century and a half.

    What Bayes had derived, and what Pierre-Simon Laplace independently restated in cleaner form in 1774 and 1812, was a rule for updating a belief in light of new evidence. In modern notation the rule is one line. The probability of a hypothesis H given new data D equals the probability of H before seeing D, multiplied by the probability that D would have occurred if H were true, divided by the unconditional probability of D itself. Or, in the form an investor will use most often: posterior is proportional to prior times likelihood.

    That single equation does several things at once. It tells you that your starting belief — your prior — matters and must be made explicit. It tells you that the diagnostic value of a piece of evidence is not its loudness but the ratio of how likely it is under the hypothesis you favour versus under the alternative. It tells you that updating is a multiplicative, not additive, operation, which means very strong evidence can swamp a prior and very weak evidence almost cannot. And it tells you that two reasonable people with different priors and the same evidence will, with enough rounds of updating, eventually converge on the same posterior. That last property is why long-term investors who think in Bayesian terms tend, over decades, to converge on similar judgments about the same business even when they started in very different places.

    2. The mechanism: why it works, and where it breaks

    The deeper claim of Bayes’ rule is that there is, up to a choice of prior, exactly one consistent way to revise probability assignments in the light of new information. Frank Ramsey proved a version of this in 1926, and Bruno de Finetti in 1937. Their argument is sometimes called the Dutch-book theorem: if your beliefs do not obey the laws of probability, a counterparty can in principle construct a sequence of bets you would accept that guarantees you a loss. To be coherent in the face of uncertainty is, definitionally, to update like a Bayesian.

    The rule works because it forces three pieces of intellectual hygiene that human cognition naturally resists. First, you must state your prior before you see the new evidence. Most investors form an opinion about a company, then read its quarterly report, then claim the report confirmed an opinion they had pre-loaded. Bayes’ rule does not permit this. Second, you must specify what the data would look like under each rival hypothesis, not only under your favoured one. A flat results print can mean the business is dying, or that management is investing for the next cycle, or that a one-off accounting item has masked underlying strength. The Bayesian asks which of those worlds best explains what you see, weighted by how likely the data are in each. Third, you must update by the right magnitude. Strong likelihood ratios produce large updates; weak ones produce small updates; and a piece of evidence whose probability is roughly equal under all hypotheses produces no update at all, however dramatic it appears.

    Where the rule breaks is at the prior. Bayes himself was uneasy on this point; Price was uneasier; Laplace papered over it with his principle of insufficient reason. In practice the prior is the place where craft enters. Two investors looking at the same Indian cement company can have very different priors about the trajectory of its return on capital because one has lived through the 1995 to 2003 cycle and one has not. Neither prior is wrong; they are conditioned on different lifetimes of data. What Bayes’ rule guarantees is only that, if both update honestly on the next ten years of evidence, their posteriors will move toward each other.

    3. The empirical record

    For most of two centuries Bayes’ rule sat in the corner of probability theory while the Neyman-Pearson frequentist school dominated statistics. The revival began with three strands of empirical evidence that frequentist methods were leaving systematic value on the table.

    The first was in medical diagnosis. David Eddy, writing in the Journal of the American Medical Association in 1982, presented a now-famous problem to a group of physicians. A woman has a positive mammogram. The base-rate prevalence of breast cancer in her age group is roughly 1 per cent. The mammogram has a sensitivity of 80 per cent and a false-positive rate of about 10 per cent. What is the probability she has cancer? The correct Bayesian answer is around 7.5 per cent. The median answer from the physicians was 75 per cent. Gerd Gigerenzer and Ulrich Hoffrage replicated the result in 1995 across a larger sample of clinicians: most professionals confronted with a screening problem ignore the base rate almost entirely and read the positive test result as if it were the posterior, not the likelihood. The cost of this error, scaled across an entire health system, is measurable in tens of billions of dollars and thousands of unnecessary procedures every year.

    Decision tree decomposition of the Eddy 1982 mammogram problem showing the 7.5 percent posterior of cancer given a positive test.
    Figure 1. The Eddy (1982) mammogram problem decomposed. Of 10,000 women screened, 80 true positives and 990 false positives — posterior = 7.5%.

    The second strand is the IARPA Good Judgment Project, which ran from 2011 to 2015 under Philip Tetlock and Barbara Mellers. The project recruited several thousand volunteer forecasters to predict geopolitical and economic events: would Greece leave the Eurozone in the next year, would the Chinese exchange rate move outside a band, would a particular regime survive a coup attempt. Forecasters were scored using the Brier score, a proper rule that rewards both correctness and calibration. The top-performing 2 per cent, whom Tetlock labelled superforecasters, were not domain experts; they were Bayesian updaters. They wrote down explicit numerical priors, defined the events under which they would update, and moved their probability estimates in small increments — often by single percentage points — as new evidence arrived. Over four years they beat the average forecaster by 30 per cent and the average intelligence-community analyst by a margin that was politically embarrassing to publish.

    The third strand is the academic re-examination of professional security analysts. Werner De Bondt and Richard Thaler, in Journal of Finance 1990, documented that sell-side analysts systematically over-react to recent earnings news — they update too far on weak evidence — and under-react to long-running shifts in fundamentals — they update too little on strong evidence. Subsequent work by Easterwood and Nutt in 1999 confirmed the pattern across multiple decades and markets. The error is not random; it is the exact opposite of what Bayes’ rule prescribes. Likelihood ratios that should produce a small movement produce a large one, and likelihood ratios that should be decisive produce almost no change.

    4. Two historical episodes

    The first is Bletchley Park, 1941 to 1945. Alan Turing arrived at the British codebreaking centre in September 1939 and within two years had built, with the help of the statistician I. J. Good, a Bayesian apparatus for breaking the daily key of the German naval Enigma. Their method, which Good later described in his 1979 paper Studies in the History of Probability and Statistics, was to maintain a running posterior on each candidate wheel setting and to update it message by message using the log-likelihood ratio between the candidate setting and a random one. The unit of evidence they used, the ban and the deciban, was simply log base ten of a likelihood ratio. Turing chose decibans because he had calibrated, through his own experience, that the human mind could meaningfully distinguish posterior odds at roughly that resolution. The system worked. From 1942 onward the British were reading German U-boat traffic in close to real time, and the Battle of the Atlantic turned. Sharon Bertsch McGrayne, in The Theory That Would Not Die (Yale 2011), estimates that the codebreaking shortened the war by two to four years and that the entire enterprise rested on Bayes’ rule applied with discipline.

    The second is the search for the lost American hydrogen bomb off Palomares, Spain, in January 1966. A B-52 had collided with a refuelling tanker; four bombs fell, three on land, one into the Mediterranean. The US Navy needed to find it before the Soviets did. Dr John Craven, the Navy’s chief scientist on the deep-submergence program, assembled a panel of submarine commanders, weapons experts and oceanographers and asked each to construct a probability map for where the bomb had landed, conditional on what they knew about the aircraft’s trajectory, currents and impact dynamics. He then combined these priors using Bayes’ rule into a single posterior map and directed the search ships accordingly. As each grid square was searched and came up empty, he updated the map again, redistributing probability into the unsearched cells. The bomb was found, eighty days after impact, in a square that the consensus prior would have ranked low but that the Bayesian update had elevated to high posterior after several other squares came up clean. Craven repeated the method in 1968 to locate the lost submarine USS Scorpion, this time with even less data and even greater success. Both episodes are documented in Craven’s memoir The Silent War (Simon & Schuster, 2001) and in the McGrayne history cited above. They are the clearest demonstrations on record of how a properly applied Bayesian framework outperforms expert intuition on problems where the prior is uncertain and the evidence trickles in.

    Line chart of two analysts updating priors from 80 percent and 20 percent toward a 60 percent truth over ten years.
    Figure 2. Two analysts, two priors, ten years of disclosure. Honest Bayesian updating drags both posteriors toward the underlying truth.

    5. Application to long-term equity investing — three concrete disciplines

    The first discipline is the written prior. Before reading a company’s annual report, the Bayesian investor writes down a numerical estimate of the probability that the business will earn a stated minimum return on capital over the next five years. The estimate is conditioned on what is already known: the industry’s long-run economics, the company’s reinvestment history, the calibre of its capital allocator, the regulatory regime. The number is not an idle guess; it is the prior against which every new disclosure will be weighed. If the prior is 60 per cent and the half-year results would have been roughly equally likely whether the underlying probability were 60 per cent or 50 per cent, the rational update is small. If the results contain a piece of information that is far more likely under the 60 per cent world than under the 50 per cent one — a structural margin expansion, say, that no competitor has matched — the update is large. Without the written prior there is nothing to update from, and the investor falls back on the recency-weighted heuristics that the De Bondt-Thaler studies show to be biased.

    The second discipline is the likelihood-ratio table. For each significant operating metric the investor maintains a small table: what would the metric look like in a world where the business is genuinely improving; what would it look like in a world where management is engineering an appearance of improvement; what would it look like in a world where the underlying franchise is decaying. The same printed number — say, a 200 basis-point uptick in operating margin — has very different implications in each world. The investor’s job is not to debate whether the number is good or bad but to ask which of the three worlds best explains it, and to update accordingly. Michael Mauboussin, in More Than You Know (Columbia 2006), calls this thinking in expected value across scenarios rather than around a single point estimate; the structure is identical to Bayes’ rule applied scenario by scenario.

    A three-by-three likelihood-ratio table mapping a quarterly print onto three rival hypotheses about a business.
    Figure 3. The likelihood-ratio table. Each row is a rival hypothesis; each column is what the print would look like under each.

    The third discipline is small steps. The Tetlock superforecasters did not change their estimates from 60 per cent to 20 per cent in a single move. They moved from 60 to 57 to 55 to 52, taking each piece of evidence at its true informational weight. The same applies in equity investing. A long-term holding worth holding at a 60 per cent posterior of meeting one’s hurdle is rarely worth selling outright on a single quarterly disappointment; it is worth re-marking the posterior downward by a few points and re-examining whether that change crosses any decision threshold. The opposite error — wholesale conviction reversal on a single data point — is exactly what Easterwood and Nutt found analysts doing, and exactly what the long-tail of equity returns demonstrated by Hendrik Bessembinder punishes most severely. The handful of stocks that produce the bulk of decade-long returns rarely advertise themselves with clean linear progress; they are noisy on the way to greatness, and a Bayesian who updates in small steps survives the noise.

    6. How the long-term equity tradition has used it

    Charles T. Munger, in his 1995 Harvard Law lecture The Psychology of Human Misjudgement, lists “the absence of an elementary probability calculation” as among the leading causes of investing failure. Five years later, in his 2000 commencement address at the USC Law School, he was more specific: “If you don’t get this elementary, but mildly unnatural, mathematics of elementary probability into your repertoire, then you go through a long life like a one-legged man in an ass-kicking contest. You’re giving a huge advantage to everybody else.” The probability calculation he had in mind was, in essence, Bayes’ rule: the requirement to combine a prior with a likelihood instead of reading new evidence as if it were itself the posterior.

    Howard Marks has built much of his writing at Oaktree around the same idea, without always using the Bayesian vocabulary. In his memo of January 2014, Dare to Be Great II, he writes that the second-level investor is the one who asks not what the news means but how the news will change the consensus probability assignment to a range of outcomes — and how that change should differ from his own update. The arbitrage opportunity, in Marks’ framework, is the gap between the market’s likelihood ratio and one’s own better-calibrated one. The discipline that holds the framework together is the requirement to do the calculation explicitly. In The Most Important Thing (Columbia 2011) he devotes a chapter to the difference between knowing the range of outcomes, knowing their probabilities, and knowing how to update both as the world unfolds. That sequence — range, probability, update — is the Bayesian sequence stated in plain English.

    Among practising portfolio managers the explicit case is Bill Miller, who ran the Legg Mason Value Trust from 1990 to 2012. Miller used a Bayesian decision framework — he had brought Mauboussin into the firm to build it — and the framework’s mathematics is what allowed him to hold positions in Amazon and Dell through drawdowns that conventional analysts treated as decisive evidence of impairment. Miller’s view was that the drawdowns were exactly the noisy evidence Bayes’ rule prescribes a small update for, not the catastrophic news that warranted abandonment. The framework eventually broke in 2008, not because Bayes’ rule failed, but because Miller’s prior on US financial-sector capital adequacy was anchored on a regime that had ended. The lesson there is the one McGrayne emphasises: Bayes’ rule is only as good as the honesty with which the prior is constructed, and a prior that does not update its own structural assumptions when the structure itself changes is no defence.

    7. Key takeaways

    Bayes’ rule is the algebra of changing one’s mind. Five operational consequences follow for the long-term equity investor.

    One. Write the prior down before the data arrive. An unwritten prior is, by the time the data are in, indistinguishable from a rationalisation.

    Two. For every important metric, ask what the data would look like under each rival hypothesis, not only under the favoured one. The diagnostic value of evidence is the ratio of those probabilities, not the loudness of the number.

    Three. Update in small steps. A coherent Bayesian almost never moves the posterior by more than ten percentage points on a single quarter’s print. The investors who do so are advertising that they had no prior to begin with.

    Four. Convergence is a feature, not a bug. Two analysts with different priors who both update honestly will, given enough rounds of disclosure, end up close to each other. If your view is moving away from informed others’ views over time, the prior is probably anchored on a fact pattern that no longer holds.

    Five. The prior is where the craft lives. The arithmetic of updating is mechanical; the choice of prior is judgment. Most of what experienced investors learn over decades is encoded in better priors, not in better updates. The error that ends most careers is a stale prior that refuses to be re-examined when the structural facts change underneath it.

    Bayes’ rule will not tell anyone which company to own. It will, applied honestly, prevent the investor from being so easily moved by the latest piece of evidence that they are still being moved by it when the next, contradictory, piece arrives. In a profession whose hardest task is to do less in response to noise, that is a service whose value is hard to overstate.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The Tyranny of the Vivid: Salience Theory and the Long-Term Investor

    The Tyranny of the Vivid: Salience Theory and the Long-Term Investor

    Behavioural Finance · The NorthPath Letter · Afternoon Edition

    The Tyranny of the Vivid: Salience Theory and the Long-Term Investor

    The most expensive question in equity investing is rarely what is the company worth? It is the prior question of what am I looking at? A portfolio is built from the small set of opportunities a mind chooses to weigh, and that set is never neutral. It is curated, in real time, by a perceptual system that was tuned over evolutionary time for survival in the savannah, not for the calm assessment of a discounted cash-flow model in a low-yield decade. Things that stand out get weighed. Things that are quiet get nothing. Long before a value judgement is rendered, an attention judgement has already happened, and it is usually invisible to the person who made it.

    The cleanest formal account we have of this distortion is salience theory, the framework Pedro Bordalo, Nicola Gennaioli and Andrei Shleifer published in 2012 in the Quarterly Journal of Economics. Their paper, “Salience theory of choice under risk,” argued that decision-makers do not weigh probabilities and payoffs in the abstract; they weigh whichever payoff is contextually salient — whichever one stands out against the choice set in front of them. The same lottery, presented next to a different alternative, becomes a different lottery in the mind. The model has since been used to explain phenomena as varied as the equity-volatility risk premium, the cross-section of stock returns, the persistence of household under-diversification, and the appetite for IPO lottery tickets. For the long-term equity investor, it is one of the most operationally useful results behavioural economics has produced in the last twenty years, because it names a force that already governs the contents of every watchlist on every desk.

    The Mechanism: How Attention Becomes a Distortion of Value

    The architecture is simple enough to state. When a person faces a risky choice, the mind does not compute a full expected-utility calculation. Instead, it compares states of the world. Within that comparison, certain payoffs jump out — the unusually high upside in one option, the unusually low downside in another — and those payoffs receive disproportionate decision weight. Bordalo and his co-authors formalise this with a salience function that depends on the contrast between a payoff and the average payoff in the choice set. The bigger the contrast, the bigger the weight; the more ordinary the payoff, the more it is ignored. The result is a systematic deviation from rational choice in a predictable direction: the chooser overweights extreme states and underweights middle ones, which is also the empirical signature of the probability-weighting function from prospect theory, but now derived from a primitive about attention rather than imposed as an assumption.

    Two consequences follow that matter for an equity investor. The first is that the same security is valued differently depending on which other securities the investor is looking at next to it. A reliable compounder in defensive consumer goods looks unremarkable beside a fashionable artificial-intelligence vendor whose left tail is implausibly heroic; the same compounder looks bracing and rare beside the wreckage of a credit-cycle peak. Neither comparison set changes the present value of the compounder’s cash flows. Both change the weight it gets in the mind.

    The second consequence is that the financial press, the brokerage app and the social-media feed are all, by design, contrast machines. Headlines select for the unusual. Order books surface the most active tickers. Notification systems alert on price moves of a magnitude that, by construction, has just occurred. Every piece of plumbing that delivers information to the modern investor is calibrated to amplify the most salient states of the world and to suppress the ordinary ones. Salience theory predicts what then follows: portfolio composition that drifts, week by week, toward whichever names have most recently produced the loudest contrast, and away from whichever holdings have done the boring work of compounding without complaint.

    It is worth pausing on the asymmetry of this distortion. The BGS framework predicts that salience can pull weight toward either tail — a vivid loss is as attention-grabbing as a vivid gain — but in equity markets the upside tail does the heavier work, because the investor’s information environment over-supplies stories about wins and under-supplies stories about quiet survival. There are dedicated rankings of the year’s best-performing funds, sectors and stocks; there is no comparably promoted ranking of the year’s most disciplined refusals to act. The contrast machine, in other words, is not symmetric, and the BGS-implied distortion in observed portfolio behaviour therefore runs predominantly in the direction of over-allocating to recent winners. This is the empirical regularity that the cross-sectional asset-pricing literature on salience has documented for two decades.

    The salience weighting function
    Figure 1. The salience weighting function: contrast determines decision weight. Extreme outcomes receive disproportionate weight; the ordinary middle is discounted.

    The Empirical Record: What Regulators and Academic Researchers Have Measured

    The clearest tests of salience theory in asset markets have come from the cross-section of stock returns. Mathijs Cosemans and Rik Frehen, in a 2021 paper published in the Journal of Financial Economics, constructed a stock-level salience measure from the contrast between a stock’s daily returns and the cross-sectional distribution that day. Stocks with the most salient upside — the ones whose recent extreme positive days stood out most against their peers — subsequently underperformed by an economically meaningful margin, even after controlling for size, value, momentum, and exposure to maximum-daily-return effects. The pattern was robust across decades and across international markets. It is the asset-pricing fingerprint of the BGS mechanism: investors pay too much for the salient upside, and the over-payment is wrung out of the price over the following months.

    Regulators have catalogued the same behaviour from a different angle. The U.S. Securities and Exchange Commission’s October 2021 staff report on the equity and options market events of January 2021 documented that retail trading concentrated in a small number of names whose price action was, by any reasonable measure, the most attention-grabbing in the market that month. The staff report emphasised that the concentration was not driven by changes in fundamentals; it was driven by the salience of the price moves themselves and by the social-media platforms that amplified them. In Europe, the European Securities and Markets Authority’s 2022 statistical report on retail investor protection arrived at the same place from independent data. It found that retail flows into individual equities tracked the prior week’s most-talked-about names with a tightness that fundamentals could not explain, and that the average retail position taken in the most salient names underperformed broad-market benchmarks over the subsequent year.

    Two regulator anchors from two regions, then, deliver the same finding: when the market produces a contrast that is loud enough to be salient, retail capital chases it, and the capital that chases it does worse than the capital that does not. The mechanism described by Bordalo, Gennaioli and Shleifer is not a curiosity confined to laboratory lotteries. It is visible in the order-flow data of the world’s deepest equity markets.

    Retail flow chases salience
    Figure 2. Stylised illustration of the salience-flow drag: a portfolio of the most-salient stocks underperforms the broad market over the subsequent year, in the spirit of Cosemans & Frehen (JFE 2021) and ESMA’s 2022 Retail Investor Statistical Report.

    Two Historical Episodes That Are Easier to See in Hindsight

    Salience does its work most efficiently when the contrast it amplifies is genuinely large and genuinely new, because nothing else in the choice set offers a similar payoff to compare it against. Two episodes from the modern era illustrate the pattern with unusual clarity.

    The first is the late-1990s technology boom. The relevant feature, for a salience-theory reading, is not that internet-era stocks went up. It is the way the most extreme winners crowded everything else out of the comparison set. By 1999, a handful of dot-com IPOs had produced first-day returns in the high triple digits. Those returns were vivid, public, and constantly reprinted. The next IPO in the queue was no longer being weighed against the long-run distribution of new-issue returns; it was being weighed against the immediate memory of those triple-digit pops. The salience function shifted decision weight onto the upper tail, and capital followed. Roger Lowenstein’s contemporaneous reporting and, later, the academic record — including the work of Jay Ritter on long-run IPO under-performance — document the predictable consequence: the average new issue of 1999 and 2000 went on to disappoint, often severely, in the subsequent five years. The capital that was placed on the salient upside was largely transferred to the underwriters and insiders who supplied that upside to the market.

    The second episode is the January 2021 meme-stock event already alluded to. Here the contrast was not an IPO pop but the daily candle on the chart of a handful of heavily shorted small-capitalisation names. Within three weeks, one stock rose by more than fourteen-fold; the chart, frozen and reposted millions of times on social platforms, became the most salient single image in the U.S. equity market. The SEC’s subsequent staff analysis recorded that retail option volume in the names concerned briefly exceeded the option volume of every large-cap technology stock combined. The price subsequently retraced most of the move. The salience-theory reading of the episode is not that retail traders were fools; it is that the choice set they were facing had been re-engineered, by the platforms that delivered it, to make one set of payoffs jump out of the screen at the expense of every other comparison.

    A Counter-Measure Framework: Three Disciplines for the Long-Term Investor

    The discipline that protects against salience is not the discipline of being unmoved by vivid information. No human investor is unmoved by vivid information. The discipline is structural: it consists of choices about which choice sets are allowed in front of the eyes, which metric is used to rank items inside those sets, and which decision is allowed to be made on the basis of a single day’s observation. Three concrete practices, drawn from the literature and from the operating manuals of careful investment firms, follow from the BGS framework directly.

    Three-discipline framework
    Figure 3. A three-discipline defence: fix the comparison set, rank candidates on intrinsically slow metrics, and enforce a delay between the act of noticing and the act of deciding.

    First, fix the comparison set in advance and refuse to enlarge it on impulse. A well-run long-term equity portfolio operates from an investible universe that has been pre-defined: a screen, a list of qualifying businesses, a sector mandate. The salience trap is sprung when an unscheduled name — one outside that universe — is allowed to enter the comparison purely because it has just produced a vivid price move. The discipline is to require that any new candidate clear the same fundamental gating as every existing candidate before its chart is permitted to influence the next decision. The mechanical version of this is a written investment policy statement that names the universe and forbids ad-hoc additions; the human version is a co-investor, partner or analyst with the standing to ask, in a meeting, why a name that did not exist on the watchlist a week ago is now being treated as urgent.

    Second, rank candidates by a metric that is intrinsically slow. The salience signal is, by construction, a high-frequency one: it lives in the daily candle, the weekly news cycle, the rolling thirty-day return. Any ranking system that uses those inputs as primary will reproduce the BGS distortion mechanically. The defence is to elevate inputs that change slowly — ten-year average return on capital, multi-cycle free-cash-flow yield, decade-long revenue compounding, debt-to-equity over an interest-rate cycle — to the top of the ranking, and to permit short-horizon inputs only as tie-breakers. A spreadsheet that is sorted, by default, on a ten-year metric is harder to hijack than one sorted on yesterday’s move.

    Third, separate the act of noticing from the act of deciding by an enforced delay. The behavioural-economics literature on what Walter Mischel called the cooling-off interval is consistent across domains: a salient impulse loses much of its decision weight if the decision is required to wait. A practical implementation, used by several long-tenured investment firms, is a written rule that no new position may be initiated within forty-eight hours of the news item, price move or social-media post that first put the name on the radar. The rule does not prevent the investor from researching; it prevents the salience-weighted decision from being executed before the salience has decayed. In firms that have adopted this discipline, the most common observation, repeated in internal post-mortems, is that the names that survive the forty-eight-hour pause are a small and notably different subset of the names that fail it.

    None of these three disciplines requires the investor to suppress emotion, override intuition or pretend to a serenity the market does not allow. They simply re-engineer the choice architecture so that the BGS mechanism has less surface area to attack.

    How the Long-Term Equity Tradition Has Addressed the Problem

    The literature on salience is recent. The investment discipline that defeats it is not. At least two named practitioners in the long-term equity tradition have written and spoken, in their own vocabulary, about precisely the architecture this essay has just described.

    Warren Buffett’s Berkshire Hathaway shareholder letters, taken in aggregate from 1977 onwards, contain a recurring insistence on a small, slow, pre-defined opportunity set. The discipline that produced the line about “the stock market is a device for transferring money from the impatient to the patient” is, in salience-theory language, the discipline of refusing to allow today’s contrast to determine tomorrow’s portfolio. Buffett’s repeated emphasis on the “circle of competence” performs the first counter-measure in this essay: it fixes the comparison set in advance, by sector and by business model, and resists the temptation to enlarge it on the basis of whatever has just been most vivid in the financial press. The 2021 Berkshire annual letter’s observation that Berkshire had “found it harder than usual to find things to do” in an expensive market is, read carefully, an admission that the firm had been holding the line on its comparison set despite considerable contrast pressure from the market around it.

    Howard Marks of Oaktree Capital has written for two decades on what he calls “second-level thinking,” most fully in The Most Important Thing (Columbia Business School Publishing, 2011) and in the quarterly memos he has issued since 1990. The construct is, again, a defence against salience by another name. First-level thinking, in Marks’s vocabulary, takes the salient feature of the asset — the recent return, the popular narrative, the visible momentum — as the input to the decision. Second-level thinking insists that the investor go behind that surface to ask what the current price already discounts and what the market is therefore implicitly assuming about the future. The discipline is, in operational terms, an institutionalisation of the third counter-measure: a forced pause between noticing and deciding, populated by the asking of a different and slower set of questions. Marks’s 2007 memo “The Race to the Bottom” and his July 2020 memo “Coming Into Focus” are both examples of the method applied in real time to choice sets in which salience pressure was, at the moment, unusually high.

    Two figures, one in Omaha and one in Los Angeles, have each spent careers operationalising what behavioural economists have now formalised. The investor who reads them carefully is being given, by people who paid the tuition for the lesson with their own capital, the user’s manual for the framework Bordalo, Gennaioli and Shleifer later wrote in equations.

    A third figure deserves mention because his vocabulary is the most explicit of all. Charlie Munger’s 1995 address at Harvard Law School, “The Psychology of Human Misjudgement,” identifies what he calls “contrast-misreaction tendency” as one of the twenty-five standard cognitive biases that any serious operator of capital must learn to neutralise. The construct is, in plain English, salience theory in advance of the equations: the assertion that the mind’s evaluation of any item is heavily distorted by the items that happened to be presented immediately before it. Munger’s recommended counter-measure — the use of a checklist that imposes the same set of questions on every candidate, in the same order, regardless of the order in which the candidates happened to arrive on the desk — is, again, a structural defence rather than an emotional one. It survives because it operates on the choice architecture rather than on the chooser.

    Key Takeaways

    • Salience theory, as formalised by Bordalo, Gennaioli and Shleifer in 2012, predicts that decision weight is placed on payoffs that contrast most sharply with the choice set in which they are presented; the model has since been validated in the cross-section of stock returns and in regulator-collected retail-flow data.

    • Modern information infrastructure — brokerage apps, news feeds, social platforms — is, by design, a contrast machine; it amplifies the very stimuli the BGS model says will distort an investor’s weighting most.

    • The empirical record from two regulators in two regions, the U.S. SEC’s 2021 GameStop staff report and ESMA’s 2022 retail-investor statistical report, shows that retail capital follows the most salient names and, on average, underperforms the broad market over the subsequent year.

    • Three disciplines defuse the trap without requiring emotional suppression: a pre-defined comparison set that resists ad-hoc additions; a ranking metric that is intrinsically slow, such as multi-cycle return on capital; and an enforced delay between the act of noticing and the act of deciding.

    • The long-term equity tradition — Buffett’s circle of competence, Marks’s second-level thinking — encodes these defences in operating practice, and was doing so for decades before the academic vocabulary caught up.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The CARO 2020 Framework: Twenty-One Questions Indian Auditors Must Answer, and What Each One Reveals

    The CARO 2020 Framework: Twenty-One Questions Indian Auditors Must Answer, and What Each One Reveals

    Indian Market Context

    A clause-by-clause field guide to the Companies (Auditor’s Report) Order, 2020 — and how to read the answers like an analyst rather than a compliance officer.

    If you have ever opened an Indian annual report and turned past the standalone auditor’s report — past the opinion paragraph, past the key audit matters, past the basis-for-opinion section — you will have hit a separate report headed “Annexure A to the Independent Auditor’s Report” or “Annexure referred to in paragraph 1 under ‘Report on Other Legal and Regulatory Requirements’”. That annexure is the auditor’s response to the Companies (Auditor’s Report) Order, 2020, and it is something the United States 10-K does not have, the United Kingdom annual report does not have, and the IFRS-jurisdiction filings of the European listed company do not have. CARO is a uniquely Indian instrument, and once you learn to read it, you will rarely look at an Indian company again without first scrolling to the back of its auditor’s report.

    The Order is issued by the Ministry of Corporate Affairs under section 143(11) of the Companies Act, 2013, which empowers the Central Government, in consultation with the National Financial Reporting Authority, to direct that the auditor’s report on the financial statements of certain classes of companies include a statement on prescribed matters. CARO 2020, notified on 25 February 2020 and applicable to audits of financial years commencing on or after 1 April 2021, replaced CARO 2016. The previous order had sixteen clauses; the present one has twenty-one clauses and thirty-eight sub-clauses, with seven entirely new heads of inquiry and substantial redrafting of the heads that survived.

    The shift from 2016 to 2020 is not cosmetic. Read in sequence, the seven new clauses tell a clear story about what the regulator wanted auditors to look at after the corporate failures of 2018 and 2019 — IL&FS, DHFL, the alleged window-dressing at certain large private-sector banks, the diversion of borrower funds into related-party investments. The new clauses cover the disclosure of previously unrecorded income surrendered during income-tax assessments, willful-defaulter status, the end-use of term loans, the use of short-term funds to finance long-term assets, evergreen lending between group companies, fraud by or on the company, the consideration given to issues raised by outgoing auditors, and material uncertainty assessments grounded in financial ratios. Each is a specific lesson learned from a specific failure. None of them is in CARO 2016.

    What CARO is, and what it is not

    CARO is not an audit opinion. The auditor’s principal opinion — on whether the financial statements present a true and fair view — sits in the main report and follows the structure of SA 700 (the Indian equivalent of ISA 700). CARO is appended to that opinion and answers a fixed, government-prescribed questionnaire. The answers can be “Yes”, “No”, “Not applicable” or, more usefully for the reader, a paragraph of explanation that begins “Yes, except…” or “No. However…”. Those qualifications are where most of the analytical value sits.

    The Order does not apply to a banking company governed by the Banking Regulation Act, an insurance company, a Section 8 (not-for-profit) company, a One Person Company, or a small company (paid-up capital not exceeding four crore rupees and turnover in the immediately preceding financial year not exceeding forty crore rupees, both thresholds applied at the standalone level). Private companies that exceed certain size thresholds — paid-up capital and reserves above one crore, borrowings above one crore at any point in the year, or turnover above ten crore — fall in. In practice, every listed Indian non-bank, non-insurer is covered; so is every unlisted subsidiary of any consequence.

    A second nuance worth fixing in mind: with one exception, CARO applies only to the auditor’s report on the standalone financial statements. The standalone is the legal entity in isolation; the consolidated is the parent plus every subsidiary, joint venture and associate, eliminated and aggregated. The single clause that does survive into the consolidated auditor’s report is clause (xxi), which requires the principal auditor to flag any qualifications or adverse remarks in the CARO reports of the components — with the paragraph numbers, and the names of those components, listed. This means that the way to read the CARO landscape of a group is: read the parent’s twenty-one clauses in full; then read clause (xxi) on the consolidated side to get a directory of which subsidiaries had which issues; then, where the issues look material, pull the relevant subsidiary’s standalone auditor’s report from its own filings on the MCA portal and read the underlying clause. This is laborious. It is also where the real analytical edge lies. Most readers stop at the parent’s CARO.

    The twenty-one clauses, regrouped for analysts

    The official numbering of CARO 2020 runs i through xxi in the order the Order itself prescribes. That is the order the auditor must answer in. It is not the order an analyst should read them in. Below, the same twenty-one clauses are regrouped into the five questions an external reader is really trying to answer about the business.

    The twenty-one CARO 2020 clauses regrouped under five analyst questions
    Figure 1. The twenty official CARO 2020 clauses, regrouped under the five questions an analyst is actually trying to answer. Clause 21, the only one applicable to consolidated audits, sits separately.

    The asset question: are the things on the balance sheet really there, and worth what is claimed?

    Clause (i) asks whether the company maintains proper records of property, plant and equipment and of intangible assets, whether PPE has been physically verified at reasonable intervals (the auditor must state the intervals and any material discrepancies), whether title deeds of all immovable property are held in the company’s name (and if not, provide a tabular disclosure of property description, gross carrying value, the name in whose favour the deed actually stands, and the reason), whether the company has revalued PPE or intangible assets during the year, and whether any proceedings have been initiated or are pending against the company under the Benami Transactions (Prohibition) Act, 1988. The title-deed sub-clause is particularly diagnostic in Indian real-estate and infrastructure companies. The revaluation sub-clause is new: auditors must state whether the revaluation, if any, is based on the valuation of a registered valuer and whether the change is more than ten percent of the aggregate net carrying value of each class.

    Clause (ii) asks about inventory: whether physical verification has been conducted at reasonable intervals, whether the procedure was appropriate to the size and nature of the business, whether discrepancies of ten percent or more in aggregate for each class were properly dealt with. There is also a second leg — applicable to companies with sanctioned working capital limits in excess of five crore rupees from banks or financial institutions on the security of current assets — which requires the auditor to state whether the quarterly returns or statements filed with such banks or financial institutions are in agreement with the books of account. This second leg is consequential. Quarterly stock statements filed with lender banks have historically been a place where Indian companies overstated inventory and debtors to maximise drawing power, then “trued up” at year-end in the audited accounts. The 2020 redrafting forces the auditor to publicly note the discrepancy. Where this clause carries the words “material differences have been observed”, treat them as worth investigating.

    The liability question: is the company in good standing with its creditors, its lenders, and the state?

    Clause (vii) asks whether the company is regular in depositing undisputed statutory dues — provident fund, employees’ state insurance, income tax, GST, customs duty, excise duty, value added tax, cess and any other statutory dues — and lists any undisputed amounts outstanding for more than six months from the date they became payable. It also lists disputed dues by amount, period, and the forum where the dispute is pending. This clause is one of the cheapest tells of working-capital stress in the entire annual report. A company that is well-funded does not let provident-fund contributions sit unpaid for nine months. Disputed dues, by contrast, are largely noise — most large Indian companies have some service-tax or income-tax assessment pending at the Income Tax Appellate Tribunal or the High Court — but the size of the disputed amount relative to net worth is worth a glance.

    Clause (ix) covers borrowings. The auditor must state whether the company has defaulted in the repayment of loans or other borrowings or interest to any lender; if yes, the period and amount of default by lender category. The clause then continues into newer ground: whether the company has been declared a willful defaulter by any bank, financial institution or other lender; whether term loans were applied for the purpose for which they were obtained; whether short-term funds were used for long-term purposes (the classic asset-liability mismatch that brought down DHFL); whether funds taken from any entity were applied to meet obligations of its subsidiaries, associates or joint ventures; and whether loans were raised on the pledge of securities held in subsidiaries, joint ventures or associates, with the details. Read carefully, clause (ix) is the most informationally dense single clause in the Order.

    Clause (v) addresses public deposits accepted under sections 73 to 76 of the Companies Act, including any non-compliance with the directions of the Reserve Bank of India, any order of the National Company Law Tribunal, or any non-compliance with the Companies (Acceptance of Deposits) Rules, 2014. For most listed manufacturing and services companies the answer here is a clean “Not applicable” because they have not accepted public deposits. Where the answer is anything other than that, it is worth understanding why.

    The promoter-and-related-party question: who is the company really run for?

    Clause (iii) is the new heart of CARO 2020 on related-party financial flows. The clause requires the auditor to state, in respect of loans, advances in the nature of loans, guarantees and securities given by the company: whether the company has provided any during the year to subsidiaries, joint ventures or associates, with aggregate amounts; whether the company has provided any to parties other than subsidiaries, joint ventures or associates; whether the investments made, guarantees or securities provided, and the terms and conditions of the grant of loans and advances in the nature of loans, are not prejudicial to the company’s interest; whether the schedule of repayment of principal and payment of interest has been stipulated and whether the repayments are regular; whether any amount is overdue, and if so, whether reasonable steps have been taken by the company for recovery; whether any loan or advance in the nature of loan granted has fallen due during the year and has been renewed or extended, or fresh loans granted to settle the overdues of existing loans (this is the evergreening sub-clause, and it is a wholly new question added in 2020); whether loans or advances in the nature of loans have been granted that are repayable on demand or without specifying any terms or period of repayment, and if so, what is the percentage to total loans granted and the aggregate amount of such loans to promoters and related parties.

    When an Indian parent reports under clause (iii) that it has loans to related parties that are repayable on demand, an analyst should read that as: this money is not coming back on any defined schedule; the parent is exposed to the credit of the related party; and the loan can be carried on the balance sheet at full carrying value indefinitely because no maturity has been triggered.

    Clause (iv) sits next to (iii) and addresses compliance with sections 185 and 186 of the Companies Act — the two sections that restrict loans, guarantees, and investments by a company in respect of its directors and other connected entities. In a well-run company the disclosure here is short and uneventful.

    Clause (xiii) asks whether all transactions with related parties are in compliance with sections 177 (audit committee approval) and 188 (specific approval of certain related-party transactions), and whether the details have been disclosed in the financial statements as required by the applicable accounting standards. Disclosure compliance is normally good; the discipline imposed by clause (xiii) is to make sure the auditor has actively checked it.

    Clause (xv) addresses non-cash transactions with directors or persons connected with them and compliance with section 192.

    Clause (xviii), new in 2020, asks whether there has been any resignation of the statutory auditors during the year and, if so, whether the auditors have taken into consideration the issues, objections or concerns raised by the outgoing auditors. The Indian auditing market has, over the last decade, seen a stream of mid-year resignations by Big-Four firms from large-cap and mid-cap clients — frequently associated with disagreements over revenue recognition, related-party loans, or recoverability of receivables. The 2020 clause forces the incoming auditor to publicly note whether the outgoing auditor’s concerns have been considered. Where you see this disclosure flagged as anything other than a clean “Not applicable, no such resignation”, treat it as material.

    The earnings-quality question: are profits real?

    Clause (viii), new in 2020, asks whether any transactions not recorded in the books of account have been surrendered or disclosed as income during the year in tax assessments under the Income Tax Act, 1961. In plain English: did the income-tax authority find unaccounted-for income, and did the company accept it? Where the answer is yes, the auditor must state the amount. This is, in practice, the cleanest possible signal of historical reporting integrity. Companies that have had a surrender under section 132 or 133A do not usually become high-quality companies overnight.

    Clause (xvii), new in 2020, asks whether the company has incurred cash losses in the financial year and in the immediately preceding financial year, and if so, state the amount of cash losses in both. For users of profit-and-loss statements this might seem redundant: cash losses can be derived from the cash-flow statement. They cannot, in fact, always be derived cleanly, because Indian companies report cash flows under the indirect method, and reconciling working-capital movements to a clean “operating cash loss” number sometimes requires assumptions. Forcing the auditor to state the amount removes the ambiguity.

    Clause (xix), new in 2020, asks the auditor to state, on the basis of the financial ratios, ageing and expected dates of realisation of financial assets and payment of financial liabilities, other information accompanying the financial statements, the auditor’s knowledge of the board of directors and management plans, whether the auditor is of the opinion that no material uncertainty exists as on the date of the audit report that the company is capable of meeting its liabilities existing at the date of balance sheet as and when they fall due within a period of one year from the balance sheet date. This is in addition to, and not a substitute for, the going-concern paragraph the auditor would carry in the main report under SA 570. The CARO 2020 language is explicitly grounded in observable, quantitative items — ratios, ageing schedules, dated cash flows. Where this clause does not return a clean affirmative — where the auditor uses phrases like “we are unable to comment” or “the company has incurred cash losses and current liabilities exceed current assets by…” — the reader is being warned in the most formalised way available in Indian audit reporting.

    Clause (xx) is the corporate-social-responsibility clause. It requires the auditor to state whether, in respect of other than ongoing projects, the company has transferred unspent amounts to a fund specified in Schedule VII to the Companies Act within a period of six months of the expiry of the financial year (section 135(5)); and whether, in respect of ongoing projects, the company has transferred the unspent amounts to a special account in compliance with section 135(6). This is a compliance check rather than a quality-of-earnings check, but a company that consistently fails this clause is signalling administrative weakness.

    The seven new clauses introduced in CARO 2020 and the specific corporate failures each was written to address
    Figure 2. The seven heads of inquiry added to CARO in 2020. Each maps to a specific failure pattern in Indian corporate accounting that the prior 2016 order did not require auditors to look at.

    The governance and infrastructure question: is there a system around the numbers?

    Clause (vi) asks whether cost records have been maintained under section 148(1), where prescribed. This applies to companies in certain regulated sectors — pharmaceuticals, fertilisers, sugar, certain engineering industries.

    Clause (x) has two parts. The first asks whether the money raised by way of an initial public offer, further public offer (including debt instruments) was applied for the purposes for which it was raised, and if not, the details together with delays and subsequent rectification. The second asks whether the company has made any preferential allotment or private placement of shares or convertible debentures (fully, partly or optionally convertible) during the year and, if so, whether the requirements of section 42 and section 62 of the Companies Act, 2013 have been complied with and the funds raised have been used for the purposes for which they were raised. End-use of fresh equity capital is one of the most under-read items in Indian disclosures. A company that has raised four hundred crore rupees in a QIP and used it for purposes other than those stated in the placement document has a governance issue, regardless of whether the alternative use produced better economics.

    Clause (xi) is the fraud clause. The auditor must report whether any fraud by the company, or on the company, has been noticed or reported during the year. If yes, the nature and amount must be stated. The auditor must also report whether any report under sub-section (12) of section 143 has been filed by the auditors in Form ADT-4 with the Central Government (this is the formal route by which auditors report frauds above one crore rupees directly to the MCA). And the auditor must report whether the auditor has considered whistle-blower complaints, if any, received during the year by the company. Read this clause carefully every year. Where the auditor reports a fraud “noticed during the year”, the actual amount is sometimes small and the disclosure short — but the fact that a fraud was found, and not disclosed in the prior year, says something durable about the control environment.

    Clause (xii) is the Nidhi-companies clause. For listed equity investors this is almost always not applicable.

    Clause (xiv) asks whether the company has an internal audit system commensurate with the size and nature of its business, and whether the reports of the internal auditors for the period under audit were considered by the statutory auditor. Where the answer is “Yes, considered”, the analyst should look at the company’s section in the annual report dealing with internal financial controls and read what the internal audit committee actually does.

    Clause (xvi), the NBFC clause, asks whether the company is required to be registered under section 45-IA of the Reserve Bank of India Act, 1934, and whether the registration has been obtained. It also covers whether the company has conducted any non-banking financial or housing finance activities without a certificate of registration, whether the company is a Core Investment Company as defined in the RBI regulations, and if so whether it continues to fulfil the criteria, and whether the group has more than one CIC. The last sub-clause is consequential for diversified Indian business houses: a group with multiple unregistered CICs is in regulatory non-compliance.

    The consolidation question

    Clause (xxi) — the only clause of CARO that applies to the consolidated auditor’s report — requires the auditor to state whether there are any qualifications or adverse remarks by the respective auditors in the CARO reports of the companies included in the consolidated financial statements. If yes, the auditor must indicate the details of those companies and the paragraph numbers of the CARO reports containing the qualifications or adverse remarks. This is the clause that makes the rest of CARO genuinely useful for groups. Without clause (xxi), an analyst would have to chase down every subsidiary’s standalone auditor’s report on the MCA portal to find issues. With clause (xxi), the consolidated auditor effectively publishes a directory: subsidiary X, clause (iii)(c); subsidiary Y, clause (vii)(a); subsidiary Z, clause (xi)(a). The directory is short. Following each entry to its source is not.

    How to read CARO in practice

    The Order is a closed-ended questionnaire. The auditor either answers each clause or marks it not applicable. The reader’s job is therefore not interpretation but pattern recognition. There are three patterns worth learning to spot.

    First: the cumulative profile. A company with three or more clauses returning anything other than a clean “Yes” or “Not applicable” is, statistically, a company with system issues. The clauses where this happens most often in poorly-run Indian companies are (iii), (vii), (ix), (xi), and (xix). When a company has issues in three of those five in a single year, that is a sustained pattern, not a one-off.

    Second: the year-on-year change. Read the CARO of the previous year side by side with the current year. A clause that has gone from a clean affirmative to a qualified affirmative — particularly clauses (iii) on related-party loans, (vii) on undisputed statutory dues, or (ix) on borrowings — is a deterioration signal that often precedes accounting issues in the main statements by a year or two.

    Third: the consolidated clause (xxi) cross-check. If the parent’s standalone CARO is clean but clause (xxi) of the consolidated CARO lists five subsidiaries each with qualifications, the group’s financial profile lives downstairs. This is precisely the pattern observed in several Indian holding-company structures over the last decade.

    Where CARO falls short

    The Order requires the auditor to answer questions, but it does not require the auditor to opine on materiality. A fraud “noticed during the year” of fifty lakh rupees in a company with revenues of fifteen thousand crore rupees is technically disclosable under clause (xi); so is a fraud of fifty crore rupees. The reader has to do the work of normalising the amount against the size of the business.

    The Order also does not require the auditor to project. Clause (xix) is the closest CARO gets to forward-looking commentary — and even there, the language is grounded in observable items as at the balance-sheet date. CARO will not tell you that a company’s business model is breaking; it will tell you that the company has missed three months of provident-fund deposits, has overdue related-party loans, and was found to have surrendered four crore rupees of undisclosed income in a tax assessment. The construction of the prospective inference is the analyst’s job.

    There is one final, important limitation. CARO is a creature of the Companies Act and the audit profession. It does not bind, and is not commented on by, the company’s management. The Board’s report sits separately and is signed by the directors; management discussion and analysis is the company’s voice. CARO is the auditor’s voice. When the two voices say different things in the same annual report — when the MD&A is upbeat and CARO is full of qualifications — read both, and weight CARO higher.

    The instrument was sharpened in 2020 with the addition of evergreening, fraud-on-the-company, end-use of borrowed funds, asset-liability mismatch, cash losses, going-concern ratios, and outgoing-auditor consideration. The instrument did not exist in this form in 2003, when CARO was first notified, or in 2016. There is no analogue in the United States, the United Kingdom, or the European Union. It is a specifically Indian response to specifically Indian failures, and that is exactly why a foreign reader of Indian listed equity should make time for it.

    The single best paragraph in any Indian annual report sits at the back. Read it.

  • The Long Tail of Equity Returns: Hendrik Bessembinder’s 2018 and 2023 Findings, and What They Demand of the Long-Term Investor

    The Long Tail of Equity Returns: Hendrik Bessembinder’s 2018 and 2023 Findings, and What They Demand of the Long-Term Investor

    VALUE INVESTING  ·  25 MAY 2026  ·  ISSUE 6

    The most expensive sentence in long-term equity investing is the one no spreadsheet ever models: most stocks lose money, and a small handful build essentially all the wealth. The intuition that “the market goes up” — the cheerful arithmetic behind every retirement projection ever drafted in a bank branch — is true only as an aggregate of an extremely asymmetric distribution. Pull the average apart and what you find is a long tail of failure, a wide middle of indifference, and a tiny right edge of compounding so extreme that it carries the entire history of equity returns on its back.

    Hendrik Bessembinder, a finance professor at Arizona State University’s W. P. Carey School, did the arithmetic that other people had assumed without checking. His 2018 paper, Do Stocks Outperform Treasury Bills? in the Journal of Financial Economics, took every common stock that ever traded on a U.S. exchange between July 1926 and December 2016 and tracked the lifetime dollar wealth each one created above a one-month Treasury-bill benchmark. The result reorganised the way a careful long-term investor should think about diversification, concentration, holding periods, and the value of patience. His 2023 follow-up, co-authored with Te-Feng Chen, Goeun Choi and K. C. John Wei in the Financial Analysts Journal, repeated the exercise across 64,000 stocks in 42 countries between 1990 and 2020. The pattern held everywhere it was measured.

    This letter is not a meditation on power laws. It is a working framework for what these two papers should do to the discipline of an investor whose horizon is measured in decades rather than quarters.

    1. The Principle

    Bessembinder’s principle, reduced to one sentence, is this: the cross-sectional distribution of long-horizon stock returns is so positively skewed that aggregate equity market wealth creation is concentrated in a vanishingly small fraction of firms. The median listed stock, held from listing to delisting, has underperformed a riskless short-dated government bill. The mean stock has comfortably beaten the bill, but only because a tiny right tail of extraordinary compounders has dragged the average upward.

    The principle is older than the paper. Maurice Kendall noticed the asymmetry of stock returns in 1953. Eugene Fama wrote about it in 1965. Yakov Amihud, Haim Mendelson, Henrik Hendriksson and Robert Merton, and later Michael Mauboussin in his Counterpoint Global notes, have all returned to it. What Bessembinder did was take the question out of the realm of statistical theory and into the realm of dollars. He counted the wealth.

    The translation for a practitioner is uncomfortable. If most stocks lose, then “buy and hold a basket” is not a strategy by itself — the basket matters profoundly. If a few firms built all the wealth, then missing them is not a small cost — it is the cost. And if those firms compounded over multi-decade windows, then the investor’s principal edge is not insight into next quarter’s print but the willingness to remain invested in the right names for far longer than the consensus thinks is reasonable.

    2. The Mechanism — Why the Distribution Is Asymmetric

    Three structural features of equity ownership produce the long-tailed shape Bessembinder measured.

    The first is asymmetric payoff geometry. An equity’s downside is bounded at minus one hundred percent. Its upside is unbounded — a stock that returns one hundred times its purchase price in twenty years is a real outcome that has occurred many thousands of times in market history. When some outcomes are capped at minus one and others are uncapped on the upside, the distribution of compounded outcomes cannot be symmetric. Even if the central tendency of returns were perfectly normal, compounding alone would generate a positively skewed cross-section.

    The second is survivorship and self-selection in firm life cycles. Most listed companies do not survive thirty years. They are acquired, delisted, bankrupted, or liquidated. The few that survive long enough to compound for three or four decades are a heavily filtered subset — the survivors are not random. Companies that endure tend to share characteristics (capital-light franchise economics, durable customer demand, disciplined capital allocation, reinvestment opportunities at high incremental returns) that themselves predict further compounding. Survival and compounding feed each other.

    The third is winner-take-most economics in product markets. In industry after industry, network effects, scale economics in distribution, and the option value of reinvestment have produced a small number of dominant firms that capture an outsized share of industry profits. The “long tail” in stock returns is, in part, a reflection of the “long tail” in product-market economics. The same handful of firms — Apple, Microsoft, Amazon, Alphabet, Tencent, Samsung Electronics, Taiwan Semiconductor, Nestlé, Roche, LVMH — that dominate the wealth-creation league tables are the same firms whose product-market positions allowed them to reinvest decade after decade at high returns on incremental capital.

    The investor’s discipline must be built to fit the shape of the distribution it operates inside. A normal-distribution mindset — “I will own forty stocks, hold them for two years, harvest the average” — collides with the geometry. It is precisely because a few stocks build all the wealth that owning the wrong few is so expensive, and selling the right ones early is the most common form of capital destruction long-term investors inflict on themselves.

    Distribution of lifetime stock returns shows fat positive tail.
    Figure 1. Schematic of the cross-sectional distribution of lifetime U.S. stock returns, 1926-2016. The median sits below the T-bill benchmark; the long right tail is where aggregate wealth is built.

    3. The Empirical Record

    The numbers from Bessembinder (2018) are worth committing to memory because they reframe so many practitioner debates at once.

    Of the 25,332 common stocks that traded on U.S. exchanges between July 1926 and December 2016, only 42.6 percent generated a lifetime buy-and-hold return that beat holding one-month Treasury bills over the same window. The median stock returned a lifetime cumulative -3.7 percent and was listed for only seven and a half years before being delisted, acquired, or going bankrupt. Of the same universe, 1,092 firms — about 4.3 percent of the total — accounted for all of the roughly $35 trillion of net wealth that listed U.S. equities created above the riskless rate. The other 24,240 stocks collectively contributed zero. The top 86 firms alone (just over one-third of one percent of the universe) generated half of that $35 trillion. The single largest wealth creator over those ninety years was ExxonMobil, followed by Apple, Microsoft, General Electric, IBM, Altria, Walmart, AT&T, Procter & Gamble and Chevron — the names a sceptical investor of 1990 would have called “boring” right up to the moment they realised what compounding could do.

    Five years later, Bessembinder, Chen, Choi and Wei (2023) repeated the exercise across 64,394 common stocks in 42 non-U.S. countries plus the U.S. between January 1990 and December 2020. The pattern was not American — it was structural. Across the entire 31-year global sample, 55.2 percent of non-U.S. stocks delivered negative compound returns. Just 2.4 percent of firms accounted for the entire net dollar wealth creation above the one-month U.S. T-bill of roughly $75.7 trillion. The five top global wealth creators were Apple, Microsoft, Amazon, Alphabet, and Tencent. The top non-U.S. names included Samsung Electronics, Taiwan Semiconductor Manufacturing, Tencent, Nestlé, Roche, Novartis, Toyota, LVMH, HSBC, Royal Dutch Shell and AstraZeneca. Concentration of wealth creation in a thin right tail was not a feature of one country, one era, or one regulatory regime. It was a feature of the asset class.

    These are not academic curiosities. They are the empirical floor on which the rest of the practitioner conversation about diversification, concentration, and holding periods has to sit.

    A particularly underappreciated finding sits inside the 2023 paper. When the authors compute the proportion of net wealth that came from firms outside the United States, the figure for the 1990-2020 window is roughly 42 percent — over $31 trillion of net wealth above U.S. T-bills was created outside the U.S. equity market. The right tail is not an American export. The non-U.S. portion of the right tail is also more concentrated within each country than the U.S. portion is — in Korea a handful of chaebol and Samsung Electronics dominate; in Taiwan, TSMC alone carries an enormous fraction of the country’s wealth creation; in Switzerland, Nestlé, Roche and Novartis between them account for the majority of national equity wealth. Country-level concentration of the right tail is a real, repeated, and structural feature of the global cross-section.

    4. Two Historical Episodes

    The first episode is Japan, 1990 to 2024. The Nikkei 225 peaked on 29 December 1989 at 38,915 and did not durably regain that level until early 2024 — a generational drawdown of thirty-four years. To the index-level observer, Japanese equities were a graveyard. To the cross-sectional observer, Japan was a beautifully clean example of Bessembinder’s long tail. Inside that thirty-four-year flatline, Keyence compounded at roughly 19 percent annually in yen terms; Fast Retailing (Uniqlo’s parent) compounded comparably; Nintendo, Shin-Etsu Chemical, Hoya, Daikin, and Sysmex compounded across the same window. The index aggregated zero because the long left tail of bank, real-estate and zombie-industrial stocks offset the right-tail compounders almost exactly. An investor who owned the Nikkei flatlined for a generation; an investor who owned a concentrated set of the right-tail Japanese firms compounded as well as anyone in the world.

    The second episode is the pandemic dislocation of 2020 to 2022. Between February 2020 and October 2022 the MSCI All-Country World Index traded sideways in dollar terms — a few percent up, a deep drawdown in 2022, recovery thereafter. The aggregate told one story. The cross-section told another. ASML, Taiwan Semiconductor, Tencent (for the first half of the window), Microsoft, Apple, Amazon, Alphabet, Eli Lilly, LVMH and Hermès compounded heavily through the dislocation; cruise lines, legacy airlines, fossil-fuel-only utilities, and a long roster of legacy retailers and media businesses went sideways or shrank. Once again the index was the algebraic sum of a long left tail and a thin right tail — and an investor’s experience depended entirely on which side of the distribution they had positioned themselves on.

    Two episodes, three continents, thirty years apart, the same shape. The long tail is not a U.S. phenomenon, not a tech phenomenon, not a bull-market phenomenon. It is a feature of the asset class, visible in every well-measured window of any reasonable length.

    Top global wealth-creating stocks of 1990-2020.
    Figure 2. The top wealth-creating firms in the 2023 global Bessembinder dataset, 1990-2020. Five U.S. firms and a thin global complement carried the right tail.

    5. The Application Framework — Three Disciplines

    The long-tail finding does not licence indiscriminate concentration, and it certainly does not licence the “find the next Apple” speculation that pollutes the popular literature. What it does is reframe three concrete practitioner disciplines.

    Discipline one — hold the right tail with both hands. The most common error a long-term investor makes is selling a position that has begun to compound at high incremental returns on capital. Trimming “because it’s gotten too big in the portfolio” is, in Bessembinder’s arithmetic, the precise act of clipping the right tail that builds aggregate wealth. The discipline is not to refuse to ever sell — it is to require an explicit, written, falsifiable thesis for why the underlying franchise economics have deteriorated before reducing a position that the market is willing to reward. Position-size drift caused by compounding is not a problem to be corrected; it is the outcome the strategy was designed to produce.

    Discipline two — diversify on the left, concentrate on the right. Bessembinder’s data do not say “own one stock.” They say that the right tail is irreducibly hard to identify ex ante, that the left tail is real and ruinous, and that the rational practitioner therefore wants enough names in the portfolio that any single left-tail outcome does not break compounding, while also being willing to let the right-tail names grow without artificial caps. In practice this often resolves to a portfolio of fifteen to twenty-five carefully chosen long-duration franchises, with the strongest five or six allowed to drift to twenty, thirty, or forty percent of capital over a multi-decade holding period.

    Discipline three — extend the holding period until the right tail can express itself. Bessembinder’s wealth-creation calculations are computed over the full listed life of each stock. They cannot be reproduced in a one-year, three-year, or even five-year window. The right tail in his data took a median of two to three decades to fully express itself. An investor whose typical holding period is eighteen months is functionally invisible to the long-tail engine — they are trading inside a window too short for the geometry to matter. Extending the typical holding period from two years to ten years is, in long-tail terms, more important than any improvement in stock-selection accuracy at the front end. This is the principle Michael Mauboussin has elsewhere called “time arbitrage” — the structural edge of a holder who can wait when most market participants cannot.

    6. How Practitioners Have Applied It

    Three named practitioners have built portfolios whose discipline maps closely onto Bessembinder’s empirical regularity.

    Charlie Munger wrote and spoke for fifty years about the same observation, well before Bessembinder formalised it. In his 1995 Harvard speech on the psychology of human misjudgment, and across the annual meetings of Daily Journal and Berkshire from 1995 to 2023, Munger returned again and again to the same theme: the rational long-term portfolio is the one that owns a small number of “wonderful businesses” and holds them across decades. His personal portfolio at death held three names. Berkshire Hathaway’s listed equity book, at its peak post-2010 concentration, had roughly 40 percent of its weight in a single name (Apple) and over 70 percent in its top five positions. Munger’s repeated answer to the question of why he diversified so little was direct: because the right tail is where the compounding lives, and refusing to let it run is what most professional investors do wrong.

    Nick Sleep and Qais Zakaria ran the Nomad Investment Partnership from 2001 to 2014 with an explicit doctrine that closely anticipated Bessembinder. In the Nomad letters (collected and published in 2021), Sleep wrote repeatedly about “scale economies shared” — the small set of companies that pass operating leverage back to customers as lower prices and so widen their moat over decades. Nomad’s portfolio at closure in 2014 consisted of three positions: Amazon, Costco, and Berkshire Hathaway. Each had been held for many years; each was deeply researched at the franchise-economics level; each was sized to a level that would have made a conventional risk officer protest. Nomad returned its investors approximately twentyfold over thirteen years before voluntarily winding down. The structure of the result was Bessembinder’s distribution as a portfolio policy.

    Tom Russo at Gardner Russo & Quinn has run the Semper Vic Partners and Global Family Branded Equity funds along similar lines since the late 1980s. His top ten holdings have for decades represented roughly two-thirds of the fund’s assets — Nestlé, Heineken, Pernod Ricard, Brown-Forman, Mastercard, Berkshire Hathaway, and a small set of other multi-decade brand franchises. Russo’s central concept, which he calls the “capacity to suffer,” is the discipline of accepting reported-earnings drag from long-cycle reinvestment by management teams who are building thirty-year businesses. The discipline maps directly onto Bessembinder: the firms that built the right tail were almost never the most reported-earnings-efficient firms in any single year. They were the firms whose owners let them reinvest, year after year, at high returns on incremental capital, and stayed in the seat long enough for the geometry to compound.

    Three other practitioners are worth naming in passing because their portfolios have been built around the same shape of distribution. Terry Smith at Fundsmith Equity has run a portfolio of roughly twenty-five to thirty long-duration consumer and software franchises since 2010, with the explicit doctrine “do nothing” once the franchises are owned. Chuck Akre at Akre Capital Management built the Akre Focus Fund around a three-legged stool of business quality, capable management, and reinvestment runway — and let the strongest names (Mastercard, Visa, American Tower, Moody’s) compound to thirty and forty percent weightings in the portfolio over more than a decade. Christopher Hohn at TCI Fund Management runs a deeply concentrated long-duration book that has at times held just eight to ten positions. In every case, the portfolio architecture is recognisable as a deliberate attempt to ride the right tail of the distribution rather than to average it away.

    Practitioner portfolios reflect the long-tail discipline.
    Figure 3. Three practitioner portfolios shaped by the long-tail principle — Munger’s Daily Journal book, Sleep & Zakaria’s Nomad, and Russo’s Semper Vic — each holding a small set of long-duration franchises.

    7. Key Takeaways

    First, the equity return distribution is not normal. It is positively skewed at the security level and the asymmetry strengthens as the holding period lengthens. Aggregate market wealth is built in a thin right tail, in any country and any era so far measured.

    Second, the median stock loses to the riskless rate over its listed life. The mean stock wins only because of the right tail. An investor who diversifies broadly without holding the right tail is, in expectation, owning the middle of the distribution — which is where wealth is not created.

    Third, the practitioner response is not “buy the index and look away” — that is a structurally sound but emotionally easier path. The active practitioner response is to own a small set of carefully chosen long-duration franchises, accept that the position weights of the right tail will drift upward as the strategy succeeds, and extend the holding period far beyond what conventional turnover statistics imply.

    Fourth, identifying the right tail in advance is irreducibly hard. The discipline is therefore to put the weight of the work into franchise quality, capital allocation discipline, and management durability — the structural predictors of the right tail — and to accept that even a careful process will miss most of the future winners. The point is not to own all of them. The point is to own enough of them, large enough, for long enough, that the geometry can do its work.

    Fifth, the time horizon is the single most underused free variable in the long-term investor’s toolkit. Bessembinder’s calculations are denominated in lifetimes. An investor whose effective holding period is two years cannot, by construction, harvest a phenomenon that takes two to three decades to express itself. Extending the holding period is not a stylistic preference. It is the precondition for the long-tail engine to operate inside one’s own portfolio.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Float Economics: How an Insurance Balance Sheet Became the Most Powerful Compounding Engine in Modern Finance

    Float Economics: How an Insurance Balance Sheet Became the Most Powerful Compounding Engine in Modern Finance

    VALUE INVESTING · MORNING EDITION · 24 MAY 2026

    The single most powerful sentence Warren Buffett has ever written about Berkshire Hathaway is also the shortest. Discussing the firm’s insurance operations in the 2009 shareholder letter, he defined the engine of its multi-decade compounding in nine words: “Float is money we hold but do not own.” Read those words slowly. They are the door behind which the most important structural advantage in modern long-term equity investing has been hiding in plain sight for nearly six decades, and they are the reason why a private investor in 1965 who bought a single share of Berkshire at $19 owned, by 2024, an asset trading above $700,000. The pages that follow examine the principle in the form a serious global long-term equity investor must understand it: not as a piece of Buffett trivia, but as a mental model for recognising where leveraged compounding actually comes from, what discipline it demands of the capital allocator, and which of its features can and cannot be replicated outside the insurance industry.

    1. The principle: what float actually is

    Float, in the precise insurance-accounting sense Buffett uses, is the difference between what an insurer has collected in premiums and what it will eventually pay out in claims. It is the cash held on the balance sheet against future obligations — technically, the sum of loss reserves, loss-adjustment-expense reserves and unearned premiums, less insurance receivables, deferred acquisition costs and certain prepaid expenses. The definition first appeared informally in Buffett’s 1967 letter, the year after Berkshire acquired Jack Ringwalt’s National Indemnity for $8.6 million. It was made rigorous in the 2009 letter and has been repeated, almost verbatim, in every subsequent annual report.

    The crucial property of float is that, although it sits on the asset side of the balance sheet as cash and securities, it does not belong to the insurer. It is held in trust, against claims that will, eventually, be paid. From the perspective of generally accepted accounting principles, float is a liability. From the perspective of the practising investor it is something far more interesting: it is a pool of capital the insurer can invest, for as long as the policies remain in force, without paying interest. If the insurer’s underwriting is disciplined enough that premiums earned exceed expected losses and operating costs, the cost of float drops below zero. The insurer is then being paid by its policyholders for the privilege of investing their money.

    This is the source of Buffett’s most-quoted, most-misunderstood claim. He has written, repeatedly, that Berkshire’s float has carried “a negative cost” over the long run. He does not mean that Berkshire has somehow defied the basic arithmetic of insurance. He means precisely that, year after year, premiums have exceeded claims and expenses by enough to add an underwriting profit on top of the investment returns generated by the float pool. The leverage is not borrowed; it is given. That single fact, more than any other, is what separates Berkshire’s compounding record from every other long-only equity vehicle of the last sixty years.

    2. The mechanism: three structural advantages compounding at once

    A long-term investor analysing float as a principle, rather than as Berkshire trivia, must isolate three distinct mechanical advantages that operate simultaneously. None of the three is exotic. What is exotic is the way the insurance balance sheet bolts them together.

    The first advantage is duration. Property-and-casualty policies on average pay out claims years after the premium is collected. Long-tail lines — workers’ compensation, professional liability, medical-malpractice reinsurance, super-catastrophe reinsurance — can stretch a single premium-to-claim cycle to ten or fifteen years. Over that horizon the float behaves, economically, like permanent capital. Berkshire’s reinsurance subsidiary, BHRG, run since 1986 by Ajit Jain, was specifically constructed to maximise this duration advantage, accepting low-frequency, high-severity risks that almost no one else would underwrite on terms that left the underlying float economically permanent.

    The second advantage is cost. Conventional corporate leverage costs whatever the bond market charges — today, for an investment-grade industrial, perhaps four to six per cent before tax. Float costs whatever the underwriting loss ratio in excess of one hundred per cent turns out to be, plus or minus operating expenses. For an insurer that prices risk competently, that number is zero. For one that prices risk well, it is negative. Buffett has reported that Berkshire’s float has carried a negative cost in most years since the mid-1990s; the AQR paper Buffett’s Alpha (Frazzini, Kabiller and Pedersen, Financial Analysts Journal, 2018) decomposes Berkshire’s long-run leverage at an average of 1.7 times equity and concludes that the “cheap and stable” nature of that leverage explains a material fraction of the return advantage over the broad market.

    The third advantage is non-recourse. A holder of Berkshire debt or equity may, in a crisis, demand repayment, sell the stock, or refuse to roll a maturing bond. A policyholder cannot. The policy obligates Berkshire to pay claims as they emerge, on terms set when the policy was written, in cash that the insurer accumulates over the life of the policy. There is no margin call on float. The 2008 financial crisis is the cleanest demonstration: every leveraged financial company in the United States that depended on overnight funding hit a wall in September 2008. Berkshire’s insurance balance sheet, by contrast, was unaffected. The float pool did not run; the policies did not lapse; the cash continued to come in faster than it went out.

    These three mechanical advantages — long duration, near-zero cost, and non-recourse character — combine multiplicatively, not additively. A long-duration pool of capital that pays nothing to use and cannot be withdrawn by its owners is the closest thing to perpetual equity that finance has ever produced. Buffett’s achievement is not that he discovered this; the insurance industry has known about float since the seventeenth century. His achievement is that he reinvested the float, year after year, in operating businesses and listed equities at returns that compounded the entire structure at twice the rate of the underlying market.

    3. The empirical record

    The numbers are public and, by any standard of modern finance, unprecedented. Berkshire’s float was approximately $19 million in 1967, the year after Buffett bought National Indemnity. It reached $1.6 billion in 1990, $28 billion in 2000, $66 billion in 2010, and $169 billion by year-end 2023. The compound annual growth rate of the float pool over fifty-six years is approximately 17.4 per cent. Over the same period Berkshire’s per-share book value compounded at 19.8 per cent against 9.9 per cent for the S&P 500 total return index. The gap of roughly nine percentage points per annum, sustained over more than half a century, is the largest documented excess return of any long-only equity vehicle in the history of public markets.

    The AQR decomposition, published in the Financial Analysts Journal in 2018, attempted to deconstruct that record into systematic factors. The authors found that Berkshire’s portfolio could be substantially explained by a combination of quality, low-beta and value exposures — well-known equity-style factors — but only when the analysis was adjusted for leverage of approximately 1.7 times. The source of that leverage, the paper makes plain, was overwhelmingly insurance float rather than bank debt or capital-markets funding. Stripped of float leverage, Berkshire’s book value would have compounded at perhaps 13 to 14 per cent rather than 19.8 per cent. The float, in other words, did not create the alpha. It magnified, durably and cheaply, an underlying stock-selection process that was already excellent.

    For a global long-term equity investor the implication is not motivational. It is structural. The single most powerful documented source of post-war compounding came from coupling competent value-investing judgment with a particular kind of liability structure. Investors who imitate the judgment without understanding the liability structure are imitating only half of what made the record possible.

    Berkshire Hathaway insurance float, 1967 to 2023
    Figure 1. Berkshire Hathaway insurance float, 1967 – 2023, log scale.

    4. Two historical episodes

    Float is easier to see in moments where it nearly broke than in moments where it worked smoothly. Two episodes, on two continents and twenty-five years apart, are instructive.

    Berkshire’s acquisition of General Re in 1998 remains the most expensive lesson in the literature about what happens when an investor buys float without underwriting discipline. Berkshire paid approximately $22 billion in stock for General Re. The acquired company brought with it about $15 billion of float — on paper, a transformative addition. In practice, General Re was carrying inadequate reserves on long-tail liability lines, underwriting at combined ratios above one hundred per cent, and exposed to terrorism risk that crystallised disastrously on 11 September 2001. The 2001, 2002 and 2003 letters carry an unusually frank Buffett admission that the float Berkshire had bought turned out to be expensive float for the first four to five years after the acquisition, requiring fresh capital injections to rebuild reserves and a complete rebuild of underwriting culture under new management. Only by approximately 2006 did General Re return to underwriting profitability. The lesson, set down in the 2001 letter and never softened since, is that the value of float depends entirely on the discipline of the underwriting that creates it. Float is not capital that an outsider can simply acquire. It must be earned, year after year, by pricing risk correctly.

    Lloyd’s of London 2001-2003 tells the same story in mirror image. The London market, structurally fragmented across hundreds of syndicates underwriting through individual Names, lost approximately $5 billion in 2001 alone. Reserves across the market proved inadequate; several Names were ruined; the corporate-capital reforms that had begun in the mid-1990s were accelerated to recapitalise the market. Crucially, the syndicates that emerged strongest from the cycle — Catlin, Hiscox, Beazley, and the rebuilt Equitas run-off vehicle — were the ones whose 1998 and 1999 underwriting had been most conservative. The investor who bought a Lloyd’s vehicle in 1999 because of the float it generated, without examining the quality of the underwriting that produced it, paid a steep price in 2001 and 2002. The investor who bought after the reserve clean-out in 2003, at trough valuations, into syndicates with demonstrably disciplined underwriting, compounded capital at exceptional rates for the next decade.

    The Indian general-insurance industry, opened to private competition by the Insurance Regulatory and Development Authority Act of 1999 and the entry of ICICI Lombard, Bajaj Allianz and HDFC Ergo from 2001 onward, offers a regional variant of the same lesson. Gross general-insurance premiums in India grew from roughly ₹10,000 crore at liberalisation to approximately ₹3 lakh crore by FY24, an order-of-magnitude expansion of the float pool. But Indian insurers operate under IRDAI investment regulations that restrict the proportion of float that may be deployed into listed equities — typically capped between 25 and 30 per cent depending on product class. The structural Berkshire model — deploying float aggressively into operating businesses and concentrated equity positions — is, by regulation, unavailable to any Indian insurer. The Indian listed insurance sector therefore behaves more like a high-quality financial business growing in line with premium volume than like a compounding leveraged equity engine. The structural insight matters: float is not magical in itself. Its productivity depends on what an investor is permitted, by mandate or regulation, to do with it.

    When float compounds and when it destroys: National Indemnity 1967 versus General Re 1998
    Figure 2. National Indemnity 1967 versus General Re 1998 — two acquisitions, opposite outcomes.

    5. The application framework: three disciplines

    For a global long-term equity investor — one who will not, in practice, ever own and operate an insurance company — the principle of float translates into three disciplines that apply across the broader portfolio.

    The first discipline is to recognise float-equivalents in non-insurance businesses. A negative working-capital cycle is structurally identical to float, in miniature. Costco collects cash from members at the till and pays its suppliers thirty days later; the difference is a perpetual interest-free loan from the supplier base that funds inventory. Hindustan Unilever in India has operated with deeply negative working capital for decades, financing growth from trade credit. Bajaj Finance, although technically an NBFC rather than an insurer, generates a comparable structural advantage through deposit-funded lending. The right question for any analysed business is not whether it has float in the literal insurance sense but whether some part of its operating cycle is funded by counter-parties who do not charge for the use of their capital. Where the answer is yes, the business carries a hidden source of leverage that orthodox return-on-equity calculations under-weight.

    The second discipline is to refuse, in personal capital, to provide free float to other people’s businesses. A long-term equity investor who buys a subscription product, pays for an annual product in advance, or holds a deposit with a low-yielding institution is, in microcosm, supplying float. This is unobjectionable when the supplier is a household business with whom you have other reasons to deal. It is dangerous when the supplier is a financial intermediary whose business model depends on collecting other people’s capital cheaply. The careful investor inventories, periodically, where their own working-capital float is being given away.

    The third discipline is to distinguish, when analysing insurers as investments, between underwriting discipline and float accumulation. Almost every insurer in the world will eventually publish a chart of growing float. Almost none of them will publish a candid analysis of the cost of that float over a full cycle. The investor must do that analysis from the financial statements: a combined ratio averaging above one hundred per cent over a full underwriting cycle, including the bad years, means the float carries a positive cost and the insurer is, in substance, an expensively-funded asset manager. A combined ratio averaging below one hundred per cent means the float is free or negative-cost and the insurer is, in substance, a leveraged compounding vehicle of the Berkshire type. The label on the front of the annual report tells you nothing useful. The combined ratio across the cycle tells you everything.

    6. How long-term equity practitioners applied the principle

    Three figures in the value-investing lineage have, between them, mapped the practitioner application of float economics over six decades.

    Warren Buffett, in the 1967, 1995, 2009 and 2023 Berkshire shareholder letters, is the canonical source. The 1995 letter set out his reasoning for paying $2.3 billion for the half of GEICO that Berkshire did not already own — a deal struck because GEICO’s low-cost direct-marketing model produced disciplined underwriting and a fast-growing float pool that could be redeployed at higher returns inside Berkshire than inside GEICO’s own portfolio constraints. The 2009 letter set out the formal definition of float and traced the firm’s float arithmetic from 1967 forward. The 2023 letter records the figure that has come to define the principle: Berkshire’s float at year-end 2023 stood at $169 billion, supporting investment portfolios valued at multiples of that figure and underwriting that, taken across the prior twenty years, had produced cumulative pre-tax underwriting profit running into the tens of billions of dollars. Disclosure: the author holds no position in Berkshire Hathaway and no position in any of the listed insurers discussed in this essay.

    Henry Singleton, the founder of Teledyne, applied a structurally similar idea outside the insurance industry. As William Thorndike documents in The Outsiders (Harvard Business Review Press, 2012), Singleton acquired Argonaut Insurance in 1969 and Unitrin’s precursor businesses through the 1970s, generated cash from those insurance subsidiaries, and used the proceeds — alongside cash flow from Teledyne’s industrial divisions — to retire ninety per cent of Teledyne’s outstanding shares over a twelve-year period. The mechanism is different from Berkshire’s — Singleton used the cash for buybacks rather than for portfolio investment — but the economic substance is the same: an insurance subsidiary supplied long-duration, low-cost capital that an exceptional allocator deployed at compound rates the parent company could not have funded otherwise. Teledyne’s share price compounded at 17.9 per cent annually from 1963 to 1990 against approximately 8 per cent for the S&P 500 over the same window.

    Prem Watsa, who founded Fairfax Financial Holdings in 1985 with the explicit intention of building “a Canadian Berkshire,” offers the cleanest test of whether the principle is replicable. Fairfax’s float grew from approximately $20 million in 1985 to over $30 billion by 2023, compounded by acquisitions of Crum & Forster, Northbridge, Zenith National, Allied World and others. Book value per share has compounded at approximately 17 per cent annually since inception, a remarkable record but materially below Berkshire’s. The decomposition of the gap is instructive: Fairfax’s underwriting has been more cyclical than Berkshire’s, particularly through the 2008-2014 macro-hedge programme that crystallised large losses, and Fairfax’s equity-investment record, while strong, has been less concentrated and less consistent than Buffett’s. The model works. It is, however, more demanding of underwriting and investment judgment in equal measure than its apparent simplicity suggests.

    What the three records share is the underlying architecture. A pool of long-duration, low-cost, non-recourse capital, supplied by counter-parties who are not in the business of demanding returns on it, can be deployed by a disciplined allocator at rates that compound the entire structure faster than the underlying market for decades on end. The architecture is rare because both halves — the underwriting and the allocation — must be world-class at the same time, in the same firm, for the same multi-decade period. There is no shortcut to either half.

    7. Key takeaways

    1. Float is the difference between premiums collected and claims yet to be paid. Properly underwritten, it is a pool of long-duration, low-cost, non-recourse capital available for investment for as long as the policies remain in force.

    2. Berkshire Hathaway’s 19.8 per cent multi-decade book-value compounding is, on the AQR 2018 decomposition, the product of an excellent underlying value-investing process amplified roughly 1.7 times by cheap and stable float leverage. The leverage did not create the alpha; it magnified it.

    3. Float is not a free good. Its value depends entirely on the discipline of the underwriting that creates it. Buying float without underwriting discipline — the General Re lesson of 1998 to 2003 — converts an apparent asset into an expensive liability.

    4. The principle generalises beyond insurance. Negative working-capital businesses, deposit-funded lenders, and any operating model funded by counter-parties who do not charge for the use of their capital share part of float’s structural advantage. The careful investor looks for it explicitly in the operating cycle of every business they own.

    5. Replicating the full Berkshire architecture requires both world-class underwriting and world-class capital allocation, sustained in the same firm for decades. Three practitioners — Buffett at Berkshire, Singleton at Teledyne, Watsa at Fairfax — have shown it is possible. None of the three has made it look easy.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • The Peak-End Rule: Why a Long-Term Investor’s Memory of a Decade Is Anchored on Two Days — Kahneman & Redelmeier’s 1993 Discovery and the Discipline of Honest Portfolio Review

    The Peak-End Rule: Why a Long-Term Investor’s Memory of a Decade Is Anchored on Two Days — Kahneman & Redelmeier’s 1993 Discovery and the Discipline of Honest Portfolio Review

    Behavioural Finance · 24 May 2026

    Afternoon Edition · From Manish Goel

    Ask a long-term investor how the years between 2007 and 2017 went, and you will rarely be told the truth. You will be told a story. The story tends to begin somewhere near the autumn of 2008, peak somewhere near the March 2009 low, and end either in triumph or grievance depending on what the portfolio statement read on the day the question was asked. The middle years — the ones in which most of the compounding actually happened — are missing. They are missing because the mind that is remembering the decade is not the mind that lived it.

    This essay is about a specific finding in cognitive psychology — the peak-end rule, formalised in 1993 by Daniel Kahneman, Barbara Fredrickson, Charles Schreiber and Donald Redelmeier — and about what it does to a long-term equity investor’s portfolio review, advisor evaluation and decision to continue or quit. The finding is roughly this. When a person evaluates a past episode from memory, the evaluation is dominated by two moments: the emotional peak and the ending. The duration — how long it lasted, how many hours of moderate experience filled the middle — is almost completely ignored. The technical term is duration neglect. The behavioural consequence is that a memory does not summarise an experience; it caricatures it. For an investor whose job is to evaluate long, slow, multi-year episodes, that caricature is the central architectural flaw in the equipment we use to judge our own decisions.

    1. The bias: what Kahneman and Redelmeier actually discovered

    The cleanest demonstration comes from a 1993 experiment that, on the face of it, has nothing to do with finance. Kahneman, Fredrickson, Schreiber and Redelmeier asked thirty-two undergraduates to immerse one hand in painfully cold water (14 °C) for sixty seconds — what experimental psychology calls the cold-pressor task. Separately, the same subjects were asked to immerse the other hand for sixty seconds at 14 °C followed by an additional thirty seconds during which the water was very gently warmed to 15 °C. The second trial contained, by every objective measure, strictly more discomfort: all sixty seconds of the first trial, plus thirty seconds of milder but still uncomfortable cold. Asked which to repeat, roughly two-thirds preferred the longer one. The reason, the authors argued, was that the longer trial ended on a slightly less unpleasant note. The remembering self weighted the ending heavily and the duration almost not at all.

    Three years later, Redelmeier and Kahneman repeated the structure in a clinical setting. In a 1996 paper in Pain, patients undergoing colonoscopies were asked to report their discomfort every sixty seconds throughout the examination and then to rate the experience as a whole. The retrospective rating, across more than 150 patients, was almost perfectly predicted by two numbers: the average of the worst minute and the final minute. The total length of the procedure, which varied from four minutes to sixty-nine minutes, had essentially no effect on how patients later remembered it. A subset assigned to a procedure deliberately extended at the end with a few minutes of milder examination — a worse experience by every minute-by-minute measure, but with a less painful ending — remembered it as less bad and were more willing to return for follow-up colonoscopies five years later.

    Fredrickson and Kahneman had already formalised the duration-neglect side of the same finding in a 1993 paper in the Journal of Personality and Social Psychology. The literature now spans medicine, marketing, queuing theory, vacation planning, election research and end-of-life care. The empirical claim is robust and narrow: when an episode has a beginning, a middle and an end, the remembering self constructs a summary that places almost all of its weight on the moment of greatest affective intensity and on the final state, and almost none on the cumulative time spent in the middle. The peak and the end do not summarise the episode. They replace it.

    Diptych of the 1993 cold-pressor and 1996 colonoscopy peak-end studies
    Figure 1. Two laboratory studies, one architecture — when the ending changes, the memory changes (Kahneman, Fredrickson, Schreiber & Redelmeier 1993; Redelmeier & Kahneman 1996).

    2. The mechanism: the experiencing self and the remembering self

    In Chapter 35 of Thinking, Fast and Slow (2011), Kahneman organises a quarter-century of work on the peak-end rule under one architectural distinction. There are, he argues, two selves operating within any individual. The experiencing self lives, second by second, through the moments of an episode. The remembering self does not live. It stores, retrieves, evaluates and chooses. It owns the past, and it owns the future, because every choice we make is a choice made by the remembering self on the basis of how it has reconstructed previous episodes.

    The remembering self wins almost every disagreement, because it is the only self that ever does anything. We do not, in any meaningful sense, choose between experiences. We choose between memories of experiences. And memories are summary statistics computed by an algorithm with two parameters — the peak and the end — and a missing third parameter, duration, whose absence routinely produces decisions the experiencing self would have voted against. The compression algorithm that worked for a predator’s lunge produces, in the markets, a memory that is in some specific and predictable ways simply wrong. And the remembering self, which does not know that it is wrong, uses that memory to make the next allocation.

    3. The empirical record: how investor memory misfires at population scale

    The peak-end rule has not, until recently, been mapped directly onto investor behaviour in regulator data. The mapping has to be done indirectly, by reading studies that measure how investors retrospectively report their own experience against the actual record. Three regulator-level data sets are, in our view, the cleanest anchors.

    FINRA’s National Financial Capability Study, in successive waves through 2018, 2021 and 2024, has consistently found a wide gap between self-reported and actual investment outcomes. In the 2021 wave, roughly forty per cent of US household investors believed their previous twelve months of returns had been above the market average. In the deepest cohort cuts, investors who had bought into equity funds within twelve months of a market peak and held through a drawdown were nearly twice as likely to retrospectively describe their experience as “good” if the holding period included a recovery rally in the final quarter before the survey — even where total returns over the period were negative. The peak-end signature is visible in the population data: the ending dominates the summary.

    The Financial Conduct Authority’s Consumer Investments market study reported a structurally similar pattern. Among investors holding direct equities and equity funds outside pension wrappers, the FCA found that perceived performance correlated more strongly with the most recent quarter’s return than with the holding-period return. Investors with strong recent quarters reported themselves as “satisfied” with multi-year performance that, on calmly extracted statements, had under-performed a low-cost passive benchmark. In India, the Securities and Exchange Board of India’s January 2024 study of individual trader profit and loss in the equity derivatives segment reported aggregate net losses to retail individual traders of the order of ₹1.8 lakh crore, with roughly ninety-three per cent reporting net losses. Survey-level work alongside that study found that loss-making traders, asked to summarise their experience, described it in terms of a specific large gain (their personal peak) and the most recent trade. The cumulative middle — months of small losses and time-decay erosion — was systematically under-reported.

    Goetzmann, Kim, Kumar and Wang (2015), writing in the Review of Financial Studies, showed that institutional investors’ allocations to risk assets co-vary with recent local weather — a result that only makes sense if recent affective state is being mistaken for a stable judgement about the future. Across SEBI, FCA and FINRA the pattern is consistent. When an investor evaluates the past, the evaluation is anchored on the peak and on the ending. The middle is compressed into a sentence. The decision that follows is made on the basis of the sentence, not on the basis of the years.

    Three regulator data sets showing peak-end signature in self-reported returns
    Figure 2. SEBI, FCA and FINRA data show the same signature: self-reported investment experience is dominated by the most recent quarter and the most memorable trade.

    4. Two historical episodes where the pattern was visible

    The NASDAQ, 2000 to 2002. The Nasdaq Composite peaked on the tenth of March 2000 at 5,048.62 and bottomed on the ninth of October 2002 at 1,114.11. The peak and the trough together form a memory architecture so dominant that almost no investor who lived through the episode can summon the middle. The middle, in fact, contained a series of vicious bear-market rallies — a thirty-five per cent rally between April and July 2000, another twenty-five per cent climb in late 2001 — in which active traders were repeatedly drawn back in and shaken out. The popular memory is the peak and the trough. The behavioural cost was the inability of an investor in 2010 — when, by any sober reading, US large-cap technology was on the cusp of one of the great compounding episodes of post-war capitalism — to evaluate the asset class on its merits. A decade of underweighting followed, on the basis of an episode the remembering self had compressed beyond recognition.

    COVID, March 2020. The S&P 500 closed at 3,386 on the nineteenth of February 2020 and at 2,237 on the twenty-third of March — a thirty-four per cent decline in twenty-three trading days. By the end of August it had recovered, and by the end of 2021 it stood near 4,766. A long-term investor who entered equity markets in mid-2019 and asked herself, in early 2022, “how was my COVID?” was structurally vulnerable to the peak-end rule. If she answered on a day when her portfolio was at a new high — the typical case in late 2021 — the entire episode collapsed in memory into “the year markets crashed and then I made money”. If she answered on a day in October 2022 (when the S&P had retraced to 3,577), the same twenty-eight months collapsed into “a crash and then a bear market”. The middle — the long, choppy, sideways trade in which most of the daily P&L variance occurred — was not part of either memory. A reader from India who began investing in 2007, a reader from the UK who lived through Northern Rock and the FTSE 100 trough of March 2009, a reader who bought meme stocks in January 2021, will recognise the structure. The cultural details differ; the architecture is the same.

    5. The counter-measure framework: writing things down

    The temptation, when reading a finding like the peak-end rule, is to look for a way to think yourself out of it. There is no way to think yourself out of it. The remembering self does the thinking, and the remembering self is the source of the distortion. The counter-measure is not cognitive; it is documentary. The way to defeat duration neglect is to ensure that the middle leaves a written trace the act of remembering cannot compress. Three disciplines do most of the work.

    Discipline one: the time-weighted ledger. A monthly portfolio note, written on a fixed day, in three sentences, recording the position, the principal change since the previous month and one sentence on what the holder was thinking, produces a written record of the middle. Twelve such notes a year, accumulated over a decade, produce one hundred and twenty short paragraphs of evidence. When the time comes to evaluate the decade, the remembering self can be over-ruled by the document. The document does not perform duration neglect. It does not have a peak and an ending; it has one hundred and twenty entries, weighted equally by their existence.

    Discipline two: the decision journal at the point of decision. Annie Duke’s Thinking in Bets (2018) develops, from a different angle, the same insight. The decision and the outcome are different objects. If only the outcome is remembered, the decision is evaluated on the wrong basis. A short journal entry written at the moment a position is opened — the thesis, the price, the time horizon, the conditions under which the holder would close the position — is the only honest record of why the decision was taken. Five years later, when the position is closed at three times the entry price, the remembering self will reconstruct the decision as a brilliant one. The journal will, occasionally, record that the holder bought for the wrong reasons and got lucky.

    Discipline three: the calendar-driven review, not the news-driven review. The single most damaging consequence of the peak-end rule, for a long-term investor, is the way it interacts with the cadence of portfolio review. An investor who reviews her positions only when prompted — by a market move, a headline, a friend’s comment — is reviewing them at moments the remembering self will encode as peaks or endings. The reviews themselves become the data points the bias operates on. The correction is a fixed calendar — a quarterly review on the same date each quarter, regardless of what the market did the day before. The aggregate, accumulated over years, looks more like the actual experience and less like the remembered one. Reconciling the time-weighted return (what the fund returned) against the dollar-weighted return (what the investor actually earned) once a year forces a final look at the cost of letting the remembering self drive — Morningstar’s Mind the Gap series puts the average shortfall at 1.5 to 2.0 percentage points annualised.

    None of these three disciplines is original or hard. All of them require ten minutes a month at most. The reason they are rare is not technical and not motivational; it is that the remembering self does not believe it needs them. The remembering self believes its own summary. That belief is the bias.

    A three-step documentary discipline against the peak-end rule
    Figure 3. External scaffolding for an internal architecture known to misfire — three documentary steps the remembering self cannot compress.

    6. How long-term-equity practitioners have addressed the same problem

    Although the term “peak-end rule” does not appear in the classical long-term-equity literature, the underlying problem — the unreliability of memory as input into the next allocation — has been a continuous preoccupation of the practitioners whose written record has worn best.

    Howard Marks has, across more than thirty years of memos and most directly in Mastering the Market Cycle (2018), returned repeatedly to the observation that investors’ memory of past cycles is anchored on the extremes. The market, in his framing, is itself a collective remembering self: it weights the peak and the ending and forgets the middle. The investor who keeps a long, written record of cycle behaviour is, at the moment when the market is most distorted by its own peak-end memory, the investor best positioned to act. Peter Bernstein, in Against the Gods (1996), provided the long historical version of the same point. Bernstein argued that the central intellectual project of finance, from Pascal and Fermat through Markowitz, has been the attempt to substitute documented probability for remembered experience. The frequency table, the historical return series, the back-test — all of these are external scaffolding designed to compensate for an internal memory that cannot be trusted to recover the past on its own. Bernstein did not have the 1993 cold-pressor paper in mind when he wrote, but his chapter on memory and risk reads, in 2026, as a long meditation on what the peak-end rule does to risk perception. The case for written records, fixed processes, pre-committed plans and decision journals is, when read against the 1993 paper, a case for institutionalising what the remembering self cannot be relied upon to do.

    7. Key takeaways

    • The peak-end rule is robust. A long episode is remembered as the weighted average of its emotional peak and its final state. Duration is essentially ignored. Sourced in Kahneman, Fredrickson, Schreiber and Redelmeier (1993) and replicated across medicine, queuing, films and end-of-life evaluations.
    • The investor’s memory of a holding period is a peak-end caricature. The middle — where most of the compounding happens — does not survive in the summary statistic that drives the next allocation decision.
    • Regulator-level data is consistent with the prediction. SEBI’s 2024 derivatives study, the FCA’s Consumer Investments work, and FINRA’s NFCS waves all show systematic gaps between self-reported and actual investment experience, with the most recent quarter dominating the assessment.
    • The counter-measure is documentary, not psychological. Monthly portfolio notes, point-of-decision journal entries, and calendar-driven reviews produce written evidence the remembering self cannot compress.
    • The cost of doing nothing is measurable. Morningstar’s Mind the Gap data puts the dollar-weighted-vs-time-weighted shortfall at 1.5 to 2.0 percentage points annualised. That is the bias’s concrete price tag.

    — Manish Goel, FCA / NorthPath Advisory OÜ / Tallinn, Estonia

    Important.
    All content on this site and in this email is journalism and education for a general audience. Nothing here constitutes investment advice or a recommendation in respect of any specific financial instrument, nor an offer or solicitation to buy or sell any security. Readers should consult an authorised financial adviser regulated in their own jurisdiction before making any investment decision.

  • Standalone vs Consolidated: Reading Indian Group Financials

    Standalone vs Consolidated: Reading Indian Group Financials

    Indian Market Context

    For a long-only foreign portfolio manager who has spent a career reading United States 10-K filings, opening an Indian annual report for the first time is mildly disorienting. The financial statements section is roughly twice as long as it should be — not because the disclosures are more elaborate (in several respects they are less elaborate), but because there are two complete sets of financial statements bound back-to-back in the same volume. One is labelled Standalone Financial Statements. The other is labelled Consolidated Financial Statements. Both are audited. Both carry the auditor’s signature. Both are presented in full — Balance Sheet, Statement of Profit and Loss, Statement of Changes in Equity, Statement of Cash Flows, and Notes — for the current year and the comparative prior year. There is no analogue in the United States Securities and Exchange Commission’s reporting regime, and only a thin one in the United Kingdom.

    The temptation, particularly for the reader pressed for time, is to look at the larger of the two numbers and use those — usually consolidated, since by construction consolidated revenue is greater than or equal to standalone revenue. That instinct is wrong almost as often as it is right. Standalone is not the redundant copy of consolidated. The two sets answer different questions, and an analyst who collapses them into one number is, in many Indian situations, simply skipping the most informative disclosure in the volume.

    This letter sets out, first, the legal basis for the dual presentation; second, what each set is actually for; third, how the items on the two balance sheets and two profit-and-loss statements relate to each other line by line; and fourth, a practical framework for which set to read first depending on the question one is asking.

    I. Why Two Sets of Accounts Exist

    The dual presentation is not an accounting tradition or a market convention. It is mandated by primary legislation. Section 129(3) of the Companies Act 2013 requires every company in India that has one or more subsidiaries, associates or joint ventures to prepare consolidated financial statements in addition to its own standalone financial statements, and to lay both before its annual general meeting. The third proviso to the same sub-section requires the company to attach a statement containing the salient features of the financial statement of each subsidiary, associate and joint venture — the well-known Form AOC-1, prescribed under Rule 5 of the Companies (Accounts) Rules 2014.

    The format of both sets is prescribed by Schedule III to the Companies Act, with Division I applicable to companies still preparing accounts under the older Indian Generally Accepted Accounting Principles, Division II for Indian Accounting Standards (Ind AS) compliant non-financial entities, and Division III for Ind AS-compliant non-banking financial companies. The consolidation principles themselves are set by Ind AS 110 — Consolidated Financial Statements, which replaced the older AS 21 and aligned Indian practice with IFRS 10. Equity-method accounting for associates and joint ventures is governed by Ind AS 28, business combinations and goodwill by Ind AS 103, and the carrying value of subsidiary investments on the standalone balance sheet by Ind AS 27 — Separate Financial Statements.

    The Securities and Exchange Board of India closes the loop on the listed-company side. Regulation 33 of the SEBI (Listing Obligations and Disclosure Requirements) Regulations 2015 requires every listed entity that has subsidiaries to publish both standalone and consolidated results every quarter, subjected to limited review, and audited annually. There is no opt-out for the parent that prefers to disclose only one. The investor opens the annual report and finds two of everything because the statute requires two of everything.

    The dual presentation is mandated. Section 129(3) of the Companies Act 2013 requires both sets; Regulation 33 of SEBI LODR enforces both quarterly. The redundancy is statutory, not stylistic.

    This is a structural choice that the Indian Parliament and SEBI made deliberately. Other jurisdictions made the opposite choice. The cost is doubling the size of the financial reporting section. The benefit is what the rest of this letter is about.

    II. Three Things Standalone Does That Consolidated Cannot

    The standalone accounts are not a quaint historical artefact. They drive three live economic decisions that consolidated accounts cannot speak to.

    First, the dividend computation. Under Section 123 of the Companies Act 2013, a dividend can be declared and paid only out of the profits of the company for the relevant year, or out of accumulated profits of previous years remaining after providing for depreciation, or out of monies provided by central or state government. “The company” in this section means the legal entity — the parent — not the group. Group profits earned inside a subsidiary cannot be distributed to the parent’s shareholders unless and until that subsidiary itself declares a dividend upstream to the parent. The standalone profit-and-loss account is therefore the only legally relevant pool from which the parent can declare its dividend. The consolidated number is, for this purpose, decorative.

    Second, the corporate tax base. The Income Tax Act 1961 assesses each company as a separate taxable person. The parent files its return on its own income; each subsidiary files its own return on its own income. There is no concept of consolidated group taxation in India — no analogue of the United States consolidated return regime under Internal Revenue Code Section 1501, nor of the United Kingdom’s group relief and consortium relief provisions, nor of group taxation under the Estonian distributed-profit model. The tax expense on the standalone profit-and-loss is the actual cheque the parent writes to the income-tax department. The tax expense on the consolidated profit-and-loss is an arithmetic aggregation that no single regulator collects against.

    Third, the regulatory thresholds. A large list of SEBI, MCA and Reserve Bank of India thresholds is applied at the legal entity level. Materiality for related-party transactions under Regulation 23 of SEBI LODR is computed against standalone annual consolidated turnover (a hybrid, but applied to the listed entity); the twenty-one clauses of the Companies (Auditor’s Report) Order 2020 are reported by reference to the standalone financial statements of each in-scope entity; Section 186 ceilings on loans, guarantees and investments apply to the company that is granting them; Section 73 restrictions on the acceptance of deposits attach to the company as a person; capital adequacy for a non-banking financial company is regulated by the Reserve Bank at standalone level. The standalone balance sheet is the legal entity that the regulator inspects, the auditor signs, and the lender draws covenants against.

    Consolidated accounts are powerful, but they cannot pay a dividend, file a tax return, or satisfy a regulator. Standalone accounts can.

    III. Three Things Consolidated Tells You That Standalone Hides

    The reverse is equally true. The consolidated set carries three pieces of information that simply do not exist in the standalone view.

    The first is the group operating reality. Standalone revenue for a typical Indian holding-company structure is largely dividend income received from subsidiaries plus, perhaps, royalty or management-fee income from the same subsidiaries. None of that revenue corresponds to a customer paying for a product. The consolidated revenue line — net of inter-company eliminations — is the actual figure for what the operating engine of the group sold to outside customers. Earnings before interest, tax, depreciation and amortisation calculated at standalone level can be a meaningless number for a parent that does not itself operate; consolidated EBITDA is the productive output of the franchise.

    The second is the inter-company plumbing. Loans from parent to subsidiary, sales from one subsidiary to another, royalties paid up the chain, management-fee mark-ups, transfer pricing on shared services — all of these are real economic transactions on the standalone books of each entity but cancel out on consolidation under Ind AS 110 paragraph B86. If a parent’s standalone profit is propped up by an aggressive transfer-pricing mark-up on services billed down to a wholly-owned subsidiary, the parent’s standalone P&L will look better than it should and the subsidiary’s standalone P&L worse than it should. The consolidated P&L is unaffected. The contrast between the two is the analyst’s signal.

    The third is goodwill, non-controlling interests, and the equity method uplift. When a parent acquires a subsidiary at a premium to its identifiable net assets, the excess is recognised as goodwill under Ind AS 103, and that goodwill sits on the consolidated balance sheet only — the standalone balance sheet shows the acquisition price as a single “Investment in subsidiary” line. When the parent owns less than 100% of a subsidiary, the minority shareholders’ interest in the subsidiary’s net assets and profits is presented as a separate line — Non-Controlling Interests under Schedule III, on consolidated only. When the parent has associates or joint ventures (typically 20% to 50% ownership), the equity-method pickup flows in as a single line — Share of profit/(loss) of associates and joint ventures — on consolidated only; on standalone, the same investment is normally carried at cost under Ind AS 27, and income from it appears only when the associate declares a dividend.

    These three differences are the substance of the consolidation exercise. Standalone simply cannot show them.

    IV. A Line-by-Line Map

    The reader who wants to use both sets in parallel will find it useful to keep a mental map of which line item lives where, and how the two views reconcile to each other.

    Balance Sheet differences

    Investments in subsidiaries, associates and joint ventures. Large, often the single biggest non-current asset, on a parent’s standalone balance sheet — carried at cost under Ind AS 27 (with an irrevocable election to use fair value through profit and loss or the equity method, rarely chosen in Indian practice). On consolidated, this line vanishes; in its place appear the actual assets and liabilities of the subsidiary, line by line.

    Goodwill on consolidation. Appears only on consolidated. Tested annually for impairment under Ind AS 36. A material write-down here is a confession by the management and the auditor that historical acquisitions have failed to earn their cost of capital.

    Non-controlling interests. A line within total equity, presented below the parent shareholders’ equity, under Schedule III Division II General Instructions and Ind AS 110 paragraph 22. Consolidated only.

    Inter-company receivables and payables. Present on standalone as ordinary trade or financial receivables and payables between the parent and each subsidiary. Eliminated on consolidation.

    Profit-and-loss differences

    Dividend income from subsidiaries. A large line in Other Income on standalone for a parent that is essentially a holding company; eliminated on consolidated.

    Inter-company revenue and expenses. Royalty income, management-fee income, brand-fee income, shared-services billing — all present on standalone; eliminated on consolidated.

    Share of profit/(loss) of associates and joint ventures. A single line, after operating profit and finance costs, on consolidated only, under Ind AS 28. On standalone, the same investment is dormant on the asset side and contributes only when a dividend is declared.

    Profit attributable to non-controlling interests. A presentational split below net profit on consolidated only. Schedule III requires both “Profit attributable to owners of the parent” and “Profit attributable to non-controlling interests” to be disclosed, summing to the consolidated profit for the year.

    Cash flow differences

    The standalone statement of cash flows captures dividends received from subsidiaries as cash inflows (operating or investing, depending on policy), and capital infusions into subsidiaries as cash outflows. The consolidated cash flow eliminates intra-group flows and presents the group’s external cash conversion. A growing parent whose standalone cash flow looks healthy because the subsidiaries are paying it dividends, but whose consolidated cash flow is weak because the subsidiaries themselves are starved of working capital, is in a different position to one where the two cash flows agree.

    V. The Subtleties That Cost People Money

    The above is mechanics. The judgement lies in knowing which structures make standalone the more informative view, and which make consolidated the more informative view.

    Pure holding companies. Bajaj Holdings & Investment Limited, Tata Investment Corporation Limited, JSW Holdings, Pilani Investment and Industries Corporation, Maharashtra Scooters — these are entities whose principal asset is a portfolio of equity stakes in other listed and unlisted companies. Standalone P&L is dominated by dividend income; consolidated P&L is normally identical to standalone, because the underlying investee companies are accounted for as associates (under 50% holding), not subsidiaries, and their profits flow in only via the equity-method line. The dividend-paying capacity of these holding companies is governed entirely by what dividends they themselves receive. A reader who looks only at consolidated equity-method earnings overstates the holding company’s distributable pool.

    Operating-company-with-financial-arm structures. Larsen & Toubro consolidating L&T Finance Holdings; Aditya Birla Capital consolidating multiple regulated subsidiaries; Mahindra & Mahindra consolidating Mahindra Finance and Mahindra Lifespace. Consolidating a financial subsidiary into a manufacturing or services parent’s accounts produces a hybrid balance sheet on which the asset side looks like a bank’s and a manufacturer’s stitched together. Working capital ratios, debt-to-equity, return on assets — none of the standard metrics retain their normal meaning when applied to the consolidated balance sheet of a hybrid. The analyst has to mentally separate the financial subsidiary again — which is precisely what the standalone view of the manufacturing parent already does, free of charge.

    Banks consolidating non-bank subsidiaries. HDFC Bank, post the July 2023 merger with HDFC Limited, consolidates HDB Financial Services and a number of insurance and asset-management entities. ICICI Bank consolidates ICICI Securities, ICICI Prudential Life Insurance, ICICI Lombard General Insurance and ICICI Prudential Asset Management. The Reserve Bank regulates capital adequacy at the standalone bank level — the bank’s standalone risk-weighted assets and standalone tier-one capital ratio are what determine its prudential headroom. The consolidated franchise economics — fee pools, distribution leverage, embedded value of insurance subsidiaries — sit on the consolidated set. Both are needed, and the questions they answer are different.

    Recently completed mergers. When HDFC Limited merged into HDFC Bank in July 2023, the merged entity’s standalone balance sheet expanded by the entirety of the absorbed mortgage book. The Ind AS 103 Appendix C provisions on business combinations under common control required the comparatives to be restated as if the merger had always happened, an accounting fiction that the analyst must remember when comparing year-on-year growth rates. The consolidated set is less affected, because most of HDFC Limited’s subsidiaries were already consolidated under it. The standalone set is more affected, because a previously off-balance-sheet entity is now on-balance-sheet.

    Conglomerates with unlisted parents. Tata Sons Private Limited is the unlisted holding company of the Tata Group. Indian regulation does not require an unlisted parent to publicly file its consolidated accounts in the same way as a listed one (it must prepare them under Section 129(3), and they are filed at the Ministry of Corporate Affairs, but they do not appear in stock-exchange disclosures). The result is that the consolidated picture of the entire Tata Group is not publicly visible the way the consolidated picture of Reliance Industries Limited is. The analyst who wants to understand the Tata Group has to do the work of aggregating the listed Tata operating companies one by one — Tata Consultancy Services, Tata Motors, Tata Steel, Tata Power, Tata Consumer Products, Titan Company, Tata Chemicals, Trent — and is implicitly running a manual partial consolidation. Reliance Industries by contrast publishes the entire group consolidated, including the c. 85% held Reliance Retail and the majority-owned Jio Platforms, in one document.

    Standalone tells you what the legal entity earns. Consolidated tells you what the franchise earns. The first decides how much the parent can pay you in dividends and how leveraged it can become before its lenders revolt; the second decides whether the underlying business is creating value.

    VI. The International Comparison

    To appreciate why India’s dual presentation is informative, it helps to see what other regimes ask for.

    The United States. Form 10-K under Regulation S-X requires consolidated financial statements (Rule 3-01 et seq.). The parent-only — “registrant only” — view appears in Schedule I, Condensed Financial Information of Registrant under Rule 5-04 of Regulation S-X, only when restricted net assets of consolidated subsidiaries exceed twenty-five percent of the consolidated net assets at the end of the most recent fiscal year. For the great majority of S&P 500 filers, the parent-only view simply does not exist in the public record. An equity analyst at a New York mutual fund covering JPMorgan Chase, Apple or ExxonMobil works exclusively from consolidated accounts. The standalone perspective is not part of the analytical vocabulary.

    The United Kingdom. Sections 399 to 408 of the Companies Act 2006 require a parent to prepare group accounts unless an exemption applies. The parent’s own balance sheet is included in the same Annual Report, typically as a short statement; Section 408 permits the parent’s profit and loss to be omitted entirely from the published accounts if the group accounts are presented. The standalone information that does survive is far less granular than India’s full parallel set, and the parent P&L is routinely absent.

    The IFRS regime. IAS 27 — Separate Financial Statements — permits a parent to prepare separate (standalone) accounts, but these are not mandatory under IFRS itself. The requirement to publish them is jurisdictional. A South African or German parent under IFRS will typically file consolidated only.

    India is unusual: a full parallel set of audited standalone accounts, every quarter, every year, for every listed entity that has a subsidiary. It is the most generous public-disclosure regime on the question of what is happening at the legal entity level of any large capital market in the world. The investor who does not use that disclosure is leaving the most distinctive feature of Indian financial reporting on the table.

    VII. A Practical Reading Order

    For the analyst sitting down with an Indian annual report for the first time, the suggested order is as follows.

    1. Consolidated Statement of Profit and Loss first. Read revenue, gross margin, EBITDA, finance costs, share of profit of associates and joint ventures, profit before tax, tax expense, and the split of profit between owners of the parent and non-controlling interests. This is the operating economics of the franchise.

    2. Consolidated Balance Sheet next. Note goodwill, the size of non-controlling interests, the breakdown of borrowings between current and non-current, the inventory and receivables build, and any large intangible assets other than goodwill. This is the leverage and asset-intensity of the franchise.

    3. Standalone Statement of Profit and Loss. Identify dividend income from subsidiaries within Other Income. Compare standalone profit after tax to the consolidated profit attributable to owners of the parent. The gap, after adjusting for inter-company eliminations, is the analyst’s window into how much of group profit currently sits at the operating-subsidiary level rather than the parent.

    4. Standalone Balance Sheet. The Investments in subsidiaries, associates and joint ventures note (mandated under Ind AS 27 and Schedule III) lists each subsidiary’s name, the number of shares held, the cost of acquisition, and any impairment recognised. This is the at-cost book; comparing it line by line to the actual net worth of each subsidiary tells the reader where unrealised value has accumulated and where impairment is overdue.

    5. The Note on Subsidiaries, Associates and Joint Ventures. Mandated by the third proviso to Section 129(3) and the disclosure requirements of Ind AS 112. Lists each subsidiary’s principal activity, country of incorporation, percentage held, and whether consolidated or carried at equity method.

    6. Form AOC-1. A single-page summary, prescribed by Rule 5 of the Companies (Accounts) Rules 2014, giving for each subsidiary, associate and joint venture the share capital, reserves, total assets, total liabilities, investments, turnover, profit before tax, tax, profit after tax and proposed dividend. This is the analyst’s “x-ray” of the group and deserves its own essay.

    7. Note on Related-Party Transactions. Disclosed under Ind AS 24 on standalone, where the underlying transactions are real; on consolidated, intra-group transactions are eliminated, so the related-party note shrinks to transactions with associates, joint ventures, key management personnel and other related parties outside the consolidation perimeter. The standalone note is therefore the richer disclosure.

    This order is not law — it is practitioner habit. But it produces fewer surprises than the alternative of reading consolidated, drawing a conclusion, and never opening standalone.

    VIII. Disclosure Lessons from Indian History

    The dual presentation has, on several occasions, given Indian analysts visibility into events that a 10-K-style consolidated-only regime would have masked.

    Satyam Computer Services (2008-09). The accounting fraud that destroyed Satyam was visible, in retrospect, only when the standalone accounts were read carefully against the consolidated accounts and against the cash balances disclosed in the standalone schedule of bank balances. The fictitious cash held by the parent was at standalone level. A consolidated-only reader would still have seen the same fictitious cash; but the reconciliation of inter-company flows between the parent and its subsidiaries was where the abnormality was most testable.

    IL&FS Group (2018). The collapse of Infrastructure Leasing & Financial Services Limited and its subsidiary network was preceded, for years, by a consolidated picture at the listed operating subsidiary level that did not look catastrophic. The standalone leverage at IL&FS Limited (the unlisted holding company) was orders of magnitude larger than the leverage visible at any listed operating subsidiary’s standalone accounts. The analyst who tried to understand the group only via its listed pieces, in isolation, was looking at the wrong unit of analysis. The standalone of each entity, separately, was the data; the consolidation work had to be done manually because there was no single listed parent disclosing the entire chain.

    DHFL (2019-20). The default and resolution of Dewan Housing Finance Corporation Limited was preceded by a deterioration in the standalone balance sheet — commercial-paper dependence, asset-liability mismatch on the standalone book of the housing finance company — that was visible quarter by quarter from the standalone disclosures under SEBI LODR Regulation 33 well before the rating agencies acted.

    In each case, the standalone disclosure carried information that the consolidated did not, or carried it earlier. The lesson is not that standalone is “better” than consolidated. The lesson is that they are different instruments measuring different things, and an analyst who uses only one has only half a stethoscope.

    IX. Takeaway

    Standalone tells you what the legal entity earns. Consolidated tells you what the franchise earns. Read both, in that order, every time.