The tell that separates a professional from an amateur forecaster is not accuracy in any given year. It is the sign of the relationship between what ret...
Professionals lean against the past. Individuals and, more damningly, corporate CFOs lean into it. That single difference in slope is the whole ballgame, because the past return is the one input everyone has and almost no one uses correctly.
A new paper by David Thesmar and Emil Verner assembles the longest and cleanest set of expected-return surveys yet gathered, and the charts do something rare: they make a well-worn belief legible. Line up the survey groups side by side and you are not looking at a spread of forecasting talent. You are looking at four distinct extrapolation rules, and only one of them is consistent with how equity markets actually price risk.
The slope is the skill
Start with the mechanism, because the mechanism is where intuition breaks. Expected returns in equities are mean-reverting by construction. When prices fall relative to earnings or cash flow, the forward return you are compensated for rises; when prices run ahead, the forward return compresses. This is not a market anomaly to be traded around. It is the arithmetic sitting underneath every discounted-cash-flow model, every earnings yield, every equity risk premium estimate ever built. Lower past returns, cheaper starting valuation, higher expected forward return. The relationship is negative, and it has been verified more times than almost any other empirical regularity in finance.
Three surveys of professional investors in the Thesmar and Verner collection reproduce that negative slope cleanly. The Value Line survey stretches from 1956 to 2024, the longest run in the set. The IBES professional survey and Robert Shiller's investor survey show the same shape over their own windows. Professionals, in aggregate, expect more from stocks after a bad year and less after a good one. They are behaving as if valuation matters, which is the entire point of valuation.
That is the strongest single observation in the paper, and it is worth stating plainly: the professionals are not forecasting the future well in any absolute sense. Nobody forecasts one-year equity returns well. What the professionals get right is the direction of adjustment. They are calibrated to the sign of the relationship that governs long-run returns, and that calibration is the difference between being roughly right about risk and being systematically wrong.
Where the amateurs invert the sign
The bottom row of the collection is where the damage shows up. The Nagel and Xu survey of individual investors and the Graham and Harvey survey of company CFOs both display the opposite slope. Higher past returns produce higher expected future returns. This is trend extrapolation, and it is the single most expensive cognitive habit in markets.
The problem is not that individuals are optimistic. The problem is that they are optimistic at exactly the wrong time. Trend-following expectations peak when valuations are stretched and forward returns are lowest, and they trough when valuations are cheap and forward returns are highest. The retail investor who expects the most after a two-year rally is expecting the most from the moment that has historically delivered the least. The extrapolation rule is not merely uninformative; it is anti-informative, pointing the forecaster in precisely the direction the evidence says to lean the other way.
The CFO problem is not a curiosity
It would be easy to file the CFO result under behavioral trivia. It is not trivia. A CFO is not a retail investor guessing at the S&P. A CFO controls the capital budget, the timing of buybacks, the issuance calendar, and the guidance the market prices off. If the people setting corporate capital allocation extrapolate past returns into future expectations, the trend-following bias is not confined to household portfolios. It is embedded in the real economy's investment decisions.
Consider the second-order consequence. Buybacks cluster near market tops, when a company's own stock has already run and management, extrapolating, feels richest. Issuance and expansion cluster near the same moments. The CFO expectation error is the microfoundation for the well-documented pattern of firms buying high and, in downturns, cutting investment when forward returns on that investment are best. The same slope that makes the retail investor a poor timer makes the CFO a poor capital allocator, and the CFO is moving far larger sums.
That is the sharpest read the charts support. The retail extrapolation error is a wealth-transfer story. The CFO extrapolation error is a capital-formation story, and it scales with the size of the corporate balance sheet.
The economists who forecast nothing
The most quietly instructive panel is the Livingston Survey of professional economists run by the Philadelphia Fed. The economists show almost no slope at all. Their return forecasts barely move regardless of what stocks did the prior year. They have internalized the disclaimer, past performance is no guide to future results, so thoroughly that they have flatlined.
This looks like discipline and is closer to abdication. A flat forecast is not the same as a mean-reverting one. The economist who ignores past returns entirely is refusing to use the one variable that carries information about forward returns. Zero slope avoids the extrapolation trap, but it also throws away the valuation signal the professionals are capturing. It is the safe answer that happens to be uninformative. The professional investors are not the ones who ignore the past; they are the ones who read it backwards from the amateurs, which turns out to be the right way to read it.
What would break the read
The honest counterargument is that survey slopes are not trading records. A negative slope in expected returns tells you a group is calibrated to valuation; it does not tell you that group makes money. Professionals face fees, career risk, and the temptation to chase flows, and the aggregate slope of their survey answers can be right while their realized performance disappoints. The paper measures the direction of expectations, not the profit-and-loss of acting on them. That gap is real and it is the first place a skeptic should push.
There is a second caveat worth stating. Aggregate slopes hide dispersion. The professional average leans against the past, but that average contains momentum chasers and deep-value contrarians whose errors cancel. The retail and CFO averages could, in principle, mask a well-calibrated minority. The charts show central tendency, not the distribution around it, and a central tendency can be right for a group whose median member is still wrong.
The read holds anyway, because the claim is narrow. It is not that professionals earn more; it is that professionals expect the sign of the relationship that governs equity returns, and the other groups do not. The thesis would break if a longer or cleaner dataset showed the professional slope flattening toward the economists' or flipping toward the retail extrapolators. Until a survey shows professionals chasing trends the way CFOs do, the cleanest reading of this evidence is that forecasting skill in equities is mostly the discipline to lean against your own recent experience, and the expensive error, wherever it appears, is the instinct to lean into it.
The observable condition to watch is corporate behavior at the next market extreme. If buyback authorizations and capital-expenditure guidance surge into the top and contract at the bottom, the CFO slope in these surveys is not a laboratory finding. It is the tape.
