Investor recognition and stock returns

In related research, Barber and Odean (2008) find that stocks that garner attention are disproportionately purchased by individual investors and subsequently earn poor returns.

We also conducted our tests using an alternative measure of investor recognition that weights each institution by the total dollar amount of that institution’s holdings. The total dollar amount of holdings serves as a proxy for the investable wealth of the institution. Results are qualitatively similar using this alternative measure. The drawback of this alternative measure is that our sample is restricted to large institutional investors who are required to file Form 13F with the SEC, so our measure of investor recognition is already biased in favor of large investors. Assuming that ownership by the smaller institutions in our sample is correlated with ownership by other smaller investors who are not included in our sample, equal weighting should provide a more representative measure of investor recognition.

Note, however, that the investor does not have to actually forget about the security for Merton’s model to apply. The investor simply has to exclude the security from the set of securities in their investable universe. Institutional investors often restrict their investable universe using criteria such as market capitalization, index membership, exchange listing and liquidity. Thus, a security could experience a reduction in investor recognition because it undergoes a change in one or more of these criteria.

Our empirical tests control for the resulting serial correlation in investor recognition using the Newey–West technique with four quarterly lags.

We have received numerous comments to the effect that the magnitude of this return seems too large to be plausible. To put it in perspective, we note that a direct sort on ex post realized returns for the annual period yields a return spread across extreme deciles of over 220%. Thus, the 90.7% spread for change in breadth is certainly large, but not implausibly so.

Merton’s (1987) model predicts that there will be a negative relation between the level of investor recognition and expected returns. Strictly speaking, our changes specification should therefore focus on the relation between change in current period investor recognition and the change in expected returns for the future period relative to the current period. By focusing on levels of future realized returns, our analysis implicitly assumes that the current period expected return is a cross-sectional constant. To the extent that this approach results in a noisy measure of the change in expected return, it should reduce the power of our tests. Note that the alternative approach of using the change in realized returns as a proxy for the change in expected returns is not feasible, because the unexpected component of the current period realized return is related to the current period change in investor recognition through P1. Another approach to testing this prediction is to use a levels specification, but this involves an omitted variables problem (discussed earlier).

We also include a main effect for Rank i-risk in period t. For brevity, we omit main effects for Rank i-risk in periods t − 1 and t + 1, because Rank i-risk is very highly autocorrelated. Including these additional variables has no material effect on the other regression coefficients.

In unreported tests we decompose our financing variable into debt and equity financing and examine how each of these financing components relates to changes in investor recognition. We find that both debt and equity financing exhibit the relations documented for total financing, and that the relations are slightly stronger for debt financing. These results indicate that the relations documented in Table 7 are not mechanically related to the increased shares outstanding arising from equity issuances.

In unreported empirical tests, we find that \(\Updelta\hbox{BREADTH},\) investment and external financing each have incremental explanatory power with respect to future returns.

It is possible that the results in our paper are due to correlated omitted variables associated with the determinants of investor recognition, rather than to investor recognition per se. We view this possibility as remote, because we test four unique implications of the investor recognition hypothesis and find strong support for all four. It would be an extraordinary coincidence if the underlying determinants also happened to explain all four of these predictions.