The question of whether investor attention can predict market returns has long sat at the edge of behavioral finance theory. A working paper1 by Zhi Da, Jian Hua, Tim Chih-Ching Hung, and Lin Peng offers compelling evidence that it can — but with a critical qualifier that changes everything: it depends entirely on who is paying attention.
The distinction the authors draw is not subtle. Aggregate retail attention, measured by abnormal Google search volume for individual stock tickers, negatively predicts one-week-ahead market returns. Aggregate institutional attention, measured by Bloomberg terminal readership activity, positively predicts future market returns — but only in advance of major scheduled news events. Two types of attention, two opposite effects, and two entirely different underlying mechanisms. "Understanding the role of AIA and ARA is important even for well-diversified investors," the authors note, underscoring that these are not idiosyncratic stock-level dynamics that cancel out at the portfolio level.
Retail Attention as a Contrarian Signal
When retail investors collectively focus on stocks — searching tickers at elevated rates relative to recent baselines — the market tends to decline in the week that follows. The effect is economically meaningful. A one standard deviation increase in aggregate retail attention reduces market returns by roughly 22 basis points in the following week, or approximately 11.6 percent annualized.
The mechanism is price pressure. Retail investors rarely short, so heightened attention leads almost mechanically to net buying activity. The authors show that elevated aggregate retail attention is positively associated with contemporaneous retail order imbalances and abnormal equity mutual fund inflows. That buying, when concentrated enough to move prices, creates transitory overvaluation that subsequently reverts. "Aggregate retail attention drives aggregate buying pressure from retail investors," the authors write, "which then results in a transitory marketwide overvaluation that subsequently reverses."
The predictability is asymmetric across market conditions. During periods of high VIX, poor liquidity, and elevated short-sale costs, the negative return predictability of retail attention is significantly stronger. When market liquidity is thin and arbitrage is costly, retail buying pressure distorts prices further and the eventual correction is sharper. This is not noise. It is a structural behavioral pattern that concentrates in exactly the environments where it can cause the most damage.
Institutional Attention as a Leading Indicator
The pattern for institutional attention runs in the opposite direction, but with a conditional structure that rewards careful reading. Unconditionally, aggregate institutional attention does not predict market returns with statistical significance. The signal emerges specifically in the days preceding major scheduled macroeconomic announcements — FOMC decisions, nonfarm payroll releases, PPI reports, and clustered earnings events from large-cap firms.
When institutional attention is elevated heading into those events, subsequent market returns are meaningfully positive. A one standard deviation increase in aggregate institutional attention predicts an 18.77 basis point gain in the following week during news-precedent windows. The authors offer two explanations that are not mutually exclusive: institutions are acquiring information ahead of events that carry genuine risk premium, and elevated institutional attention signals that the upcoming announcement has systematic rather than idiosyncratic importance.
The cross-sectional evidence reinforces this interpretation. Institutional attention's positive return predictability is strongest among high-beta stocks — precisely those with greatest sensitivity to the resolution of marketwide uncertainty. When institutions are paying broad attention to the market before a major announcement, the subsequent risk premium realization flows most forcefully through the stocks most exposed to systematic risk.
A Puzzle Partially Solved
The interaction between the two attention types sheds new light on a documented market anomaly: the pre-announcement return premium observed before clustered after-hours earnings events. Prior research had established that much of the return premium associated with major earnings announcements is realized before the news is released, not after. The mechanism had been unclear.
The authors find that this pre-announcement premium exists only when aggregate retail attention is high on the event day. "The early realization of market returns in clustered after-hours earnings announcement days may be a result of excessive buying triggered by retail investor attention," they write. When retail attention is low, the premium shifts to the following day. The implication is that retail attention-driven price pressure is effectively pulling forward returns that belong to the post-announcement period.
Five Takeaways for Advisors and Investors
First, elevated retail search activity in stocks is a near-term contrarian signal, not a momentum indicator. When retail attention spikes, the evidence suggests restraint rather than participation.
Second, institutional attention concentrated before scheduled macro announcements is a constructive signal. This is not always actionable directly, but it reinforces the historical case for maintaining equity exposure through major announcement windows rather than reducing it out of uncertainty.
Third, both signals are stronger during periods of poor market liquidity and elevated volatility. These are precisely the conditions in which behavioral biases tend to compound. Advisors who recognize the pattern may be better positioned to hold steady or selectively add when retail fear or euphoria is driving price action.
Fourth, the predictive value of these attention signals holds out of sample. The authors report an out-of-sample R-squared of 1.49 percent for aggregate retail attention, with a mean-variance investor willing to pay 226 basis points annually to access the information. That is not a rounding error.
Fifth, a key methodological insight has practical relevance: bottom-up aggregation of individual stock attention consistently outperforms top-down proxies like searches for index tickers. What matters is what retail and institutional investors are doing with individual holdings, not how they characterize the market in aggregate.
To be clear, this research is not a timing system. It is a behavioral lens — one that helps distinguish between attention that reflects genuine information processing and attention that reflects noise-driven enthusiasm. For advisors, that distinction may be among the most practically useful ideas behavioral finance has produced in recent years.
Footnote:
1 Da, Zhi, Jian Hua, Tim Chih-Ching Hung, and Lin Peng. "Market Returns and a Tale of Two Types of Attention." Working Paper, University of Notre Dame and Baruch College, CUNY, 6 Aug. 2024, https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3551662.
