The conventional view in asset pricing holds that investors have always grasped, at least intuitively, that risk means the variability of returns. A sweeping new paper by Samuel Hartzmark and David Solomon of Boston College demolishes that assumption with careful precision. Their finding is not merely academic. It has real consequences for how advisors interpret investor behaviour today.
The Curse of Knowledge
Hartzmark and Solomon open with a disarming observation. Finance models assume that the understanding behind them has always existed. Researchers "assume that they are the last to know, relative to everybody else in financial markets." That assumption, the paper argues, is both untested and largely false.
The authors trace the evolution of how equity risk was actually understood across the 20th century, drawing on academic papers, practitioner books, and decades of newspaper data from the Wall Street Journal and New York Times. What they find is a largely forgotten history of ideas that bear almost no resemblance to what modern finance teaches.
Before Markowitz: Stocks Were Gambling
The starting point is stark. Before Edgar Lawrence Smith's 1924 book demonstrated that stocks had historically outperformed bonds, the prevailing view was not that stocks were risky in the modern sense. They were viewed as "speculatons" — essentially gambling — and categorically excluded from the definition of legitimate investment. Chamberlain (1911) put it plainly: "Common stocks as such, are not superior to bonds as long-term investments, because primarily they are not investments at all. They are speculatons."
Irving Fisher offered the more sophisticated alternative: that stock versus bond performance was mostly a function of inflation. Stocks were a hedge against rising prices. Fisher conceded the risks were different but judged them roughly equal: "There is not much to choose between the risks run by investing in stocks and risks run by investing in bonds." This was not negligence. It reflected the absence of any agreed framework for measuring risk quantitatively.
Risk as Probability of Loss: The Bond Analogy
Once stocks were admitted as legitimate investments, investors reached for the nearest familiar framework: bonds. The consequence, Hartzmark and Solomon argue, was a suite of systematic misunderstandings that persisted well into the 1960s.
Risk came to mean the probability of losing money at a long horizon — analogous to bond default. Graham and Dodd's 1934 Security Analysis, one of the most influential investment books ever written, defined a safe investment as one that "holds every prospect of being worth the price paid except under quite unlikely contingencies." The language is explicitly probabilistic and long-horizon. The word "volatility" does not appear in the book at all. Neither does "variance" in any statistical sense. "Standard deviation" is entirely absent.
This framing produced a critical inversion relative to modern thinking. Under probability-of-loss logic, higher expected returns make a stock safer, not riskier, because they reduce the chance of ending up in the red. The authors note that Graham and Dodd's case for value stocks is "the exact opposite of the Fama and French (1993) view that the high returns are compensation for risk."
Nor was there any coherent treatment of intermediate price changes. Following the bond analogy, practitioners focused on cash flows — dividends analogized to coupons — and largely set aside price movements as either speculative noise or a zero-sum game.
The Markowitz Turning Point
Harry Markowitz introduced variance of returns as the definition of risk in 1952. But even Markowitz underestimated the novelty of his contribution. He claimed that substituting "variance of return" for "risk" in earlier writings would produce "little change of apparent meaning." Hartzmark and Solomon show this claim is unsupportable. Markowitz suffered, they argue, from "the curse of knowledge" — the genuine difficulty of imagining what it is like not to know something that now seems obvious.
The empirical adoption of standard deviation as a risk measure lagged further still. The authors report they are "unable to locate a standard deviation of market returns that was published before Lintner (1965)." Neither Jackson (1928) nor Cowles (1938) — both of whom computed standard deviations for other purposes in the same papers — applied the concept to market returns.
A textual analysis of the WSJ and NYT confirms the shift. The word "volatil*" comprised just 0.3% of all risk-related terms in 1915. By 2007, it was 10.5%. A structural break is identified at 1962, a decade after Markowitz, consistent with the CAPM's diffusion into practice.
Model Discovery: What This Means Now
The authors introduce the concept of "model discovery" — the idea that before a model exists, it cannot be used. Investors in 1930 could not have priced equities consistent with consumption-based asset pricing, because that framework would not exist until Lucas (1978) and would not be empirically tested until Grossman and Shiller (1981).
The implication for interpreting historical price data is sobering. "Our conclusion," the authors effectively argue, is that the rational expectations assumption — that investors always behaved as if the current model were true — is not merely an approximation. It is demonstrably wrong.
5 Key Takeaways for Advisors and Investors
- Risk literacy is more recent than it appears. The modern definition of investment risk as return variability only took hold in the 1960s. Many investors, particularly those without formal financial education, still operate on pre-Markowitz intuitions without knowing it.
- Capital-guarantee products exploit old instincts. Structured products with principal protection are not irrational products sold to irrational people. They speak directly to a deeply rooted understanding of risk as "don't lose the principal," which dominated investment thought for decades.
- The dividend-as-income fallacy has long roots. Investors who treat dividend income as bond-like, ignoring price volatility as irrelevant, are reasoning in a way that was mainstream finance as recently as the 1950s. That framing is emotionally coherent but analytically incomplete.
- Academic research changes markets. McLean and Pontiff's finding that anomaly returns decline after academic publication supports the paper's central thesis. The market does not know what researchers have not yet discovered.
- Beware charitable readings of historical wisdom. Graham and Dodd remains a bestseller. Its framework captures important truths. But its conception of risk is, by modern standards, systematically incomplete. Advisors who draw on it should do so with that in mind.
Footnote:
Hartzmark, Samuel M., and David H. Solomon. "A Brief History of Financial Risk." SSRN Working Paper, June 2026. ssrn.com/abstract=6991358.
