We use minute-by-minute Bloomberg online status data to characterize two important dimensions of sell-side equity analysts' work habits: we estimate the average workday length (AWL) to proxy for analysts' general e ort provision and we use the percentage away day (PAD) to proxy for their soft information production. Both AWL and PAD vary much more across analysts than across time. Controlling for coverage, AWL is positively related to the quantity and the timeliness of analyst forecasts, while PAD is negatively related to quantity but positively related to market reaction to recommendation change and the probability of becoming a star analyst. Both are positively related to forecast accuracy, even after controlling for analyst fixed effects. COVID lockdown provides further causal evidence. Traveling analysts (with high pre-COVID PAD) experience a significant reduction in forecast accuracy during the lockdown, especially among faraway firms. Using pre-COVID analyst commute time to instrument increased AWL during the lockdown, we find a higher AWL to significantly increase output and improve the accuracy of the forecasts.

This paper documents that equity analysts revise their forecasts about long-term earnings—earnings beyond two years—to rationalize observed stock price movements, even when the price movements are driven by shocks that are identified by economists as unrelated to fundamentals. These forecast revisions initially overreact to price movements but subsequently revert. The forecast revisions are also often delayed relative to price movements, further indicating that they are responding to, rather than driving, price movements. Our findings indicate that the observed high correlation between stock or stock market returns and changes in subjective valuation implied by analysts’ earnings growth forecasts reflects, in part, reverse causality from prices to beliefs. We estimate this channel and find that it accounts for approximately one-third of the stock-level correlation and for approximately half of the market-level correlation. Assuming that analyst forecasts are a good proxy of investor beliefs, this channel provides a possible microfoundation for inelastic demand.

We present causal evidence that non-fundamental correlated demand exerts a first-order impact on systematic returns. Mutual fund investors chase fund performance via Morningstar ratings, regardless of the rating methodology. Until June 2002, ratings depended on fund returns without any style adjustment, and thus mutual funds with the same investment style had highly correlated ratings. This methodology led rating-chasing investors to direct capital into winning styles, generating correlated demand. Capital flows exerted non-fundamental price pressure on the underlying stocks, creating style-momentum that reverted over time. In June 2002, Morningstar reformed its rating methodology so that ratings became equalized across styles. The reform demonstrates the causal impact of rating chasing: once the reform was implemented, style-level price pressures via the mutual fund channel immediately became muted. Furthermore, the dispersion in style performance declined sharply, and style momentum and reversal disappeared. We estimate that Morningstar rating chasing explains a substantial part of the size and value factors’ time-series variation.