Mental Models of the Stock Market

The paper asks what subjective reasoning — the “mental model” — underlies economic agents’ stock-return expectations, treating that reasoning as a black box that standard data leave unexplained. The authors use “mental model” to mean “beliefs about how different variables are connected … together with the reasoning about the mechanisms underlying these relationships.” They argue return expectations are pivotal in stock markets — shaping investment, generating trading through their heterogeneity, and potentially driving excess volatility and bubbles — yet conventional data on prices, trades, and even subjective expectations “remain silent on agents’ reasoning.”

The strategy elicits how people reason from “stale” (four-week-old) fundamental earnings news to future returns, a setting that sharply separates three classes of mental model. Respondents compare two scenarios about a company (e.g., Nike) — one neutral, one with good or bad news about future earnings that broke four weeks ago — and predict in which the expected future return is higher, then explain their reasoning in an open-ended question, and separately forecast differences in earnings, uncertainty, and systematic-risk exposure. The three model types are: (i) efficient markets / risk-based pricing, under which stale news is already priced in and irrelevant to expected returns; (ii) temporary mispricing, under which under- or overreaction makes expected returns temporarily differ; and (iii) equilibrium neglect, under which agents directly infer higher expected returns from higher expected earnings. The four-week lag is deliberately conservative — the authors do not take a stance on what counts as “immediate,” so treating public news as still unpriced four weeks later is taken as evidence against believing in efficiency.

Seven samples spanning laypeople to experts: US and German general population, US and German retail investors, US financial advisors, German fund managers at large asset managers, and international academic finance experts. The samples include roughly 2,434 US and 3,852 German general-population respondents, 408 US and 299 German retail investors, 406 US financial advisors, 105 German fund managers (at firms with assets under management in the hundreds of billions of euros), and 116 academic experts (published in leading finance or “top five” journals), who serve as a “state-of-the-art academic benchmark.” Additional experiments with US households probe the origins and consequences of the mental models.

Most non-experts do, and most experts do not. Among academic experts, majorities predict returns will be similar across the good/bad-news and neutral scenarios (67% for good news, 76% for bad news). By contrast, across scenarios 50–80% of general-population respondents, and around 75% of retail investors and financial advisors, make “news-congruent” forecasts (higher returns after stale good news, lower after bad news); among US retail investors specifically the shares are 75% and 81%. Among fund managers the share is lower but substantial (around 51–58% for good news). The predicted return differences are quantitatively meaningful and long-lived — US retail investors on average expect Nike’s return to be 3.7pp higher after stale good news and 7.3pp lower after bad news, and a sizeable minority still make news-congruent forecasts for annual returns four years out. The pattern extends to other firms and to macroeconomic scenarios for the aggregate US market, and replicates in a between-subject study using real news about Siemens Energy with incentivized forecasts (good-news respondents predict a 6.4pp higher return and invest a 26.7pp higher share in the stock).

Experts overwhelmingly reason in terms of market efficiency, whereas non-experts and a substantial share of fund managers neglect equilibrium pricing, directly linking expected earnings to expected returns. Using a pre-designed coding scheme applied to open-ended explanations (double-coded, 73% inter-rater agreement), 77% of experts argue in line with efficient markets / risk-based pricing and only 9% are classified as neglecting equilibrium, while 52% of the US general population, 56–69% of retail investors, and 64% of financial advisors neglect equilibrium pricing (with efficiency/mispricing arguments “almost absent”). Fund managers are heterogeneous: about 33% invoke efficiency, 19% temporary mispricing, and 35% equilibrium neglect. A complementary co-movement analysis confirms this: experts’ return forecasts correlate with predicted uncertainty and systematic-risk exposure, whereas non-experts’ forecasts co-move mainly with predicted company earnings (e.g., a US retail investor expecting higher market earnings is 39pp more likely to predict higher returns). Notably, among non-experts higher measured financial literacy is associated with more, not less, equilibrium neglect — suggesting equilibrium thinking is a facet of financial knowledge not captured by standard literacy measures.

The authors argue it reflects a genuine gap in the mental model rather than mere inattention to trading and price changes. In experiments with US households, an intervention drawing attention to the trading and price changes that occurred over the four weeks does not reduce the share inferring higher returns from stale good news, even though respondents acknowledge that prices changed. A separate experiment that explicitly rules out changes in risk exposure, and another that rules out temporary mispricing, likewise produce no significant shift in forecasts. By contrast, an intervention that explains the concept of equilibrium — how expected future earnings are incorporated into prices — reduces the fraction predicting higher returns after good news by 21 percentage points, and the effect persists in a follow-up survey days later. The authors read this as evidence that “a fundamental unfamiliarity with equilibrium,” not inattention, is central to households’ predictions, and that increasing attention is “futile” when the underlying model is misspecified.

In household-panel data, equilibrium neglect predicts two previously documented “anomalies” in return expectations: pro-cyclicality and return extrapolation. Embedding the survey module in a wave of the Bundesbank Online Panel Households (BOP-HH), the authors show that individuals exhibiting equilibrium neglect are more likely to form pro-cyclical return expectations (linking expected returns to expected economic growth, contrary to standard asset-pricing logic) and to extrapolate past returns. This link is presented as correlational within the panel — equilibrium neglect “predicts”/“is linked to” these anomalies — consistent with, but not established as, a causal driver.

The results support theoretical work that builds non-standard belief formation into macro-finance models and suggest a specific, under-explored mechanism linking earnings expectations to return expectations. Because equilibrium neglect is widespread even among retail investors, financial professionals, and some fund-management teams, the authors argue it is plausibly relevant for market-level outcomes such as momentum, reversals, excess volatility, and high trading volume that arise in models of extrapolation, disagreement, and “partial equilibrium thinking.” They stress that agents’ models “do not necessarily align with prevailing economic theories,” making it necessary to study mental models “in the wild,” and note that the inference-from-stale-news channel “suggests a new mechanism that theory has yet to explore.”

The evidence is survey/experimental, largely built on directional forecasts about mostly hypothetical scenarios, and the market-level and panel implications are argued or correlational rather than causally estimated. The main design uses hypothetical news and directional (higher/similar/lower) predictions to keep questions answerable across groups with different financial knowledge; the authors acknowledge drawbacks (reasoning may be scenario-specific, hypothetical news less credible, forecasts unincentivized, two-scenario framing may heighten sensitivity) and address them with eight alternative scenario conditions and a real-news, incentivized, between-subject robustness study. The paper documents the prevalence and correlates of mental models and their link to expectation anomalies; it does not claim to measure the causal market-level asset-pricing consequences of equilibrium neglect, presenting those as motivated implications for theory.