We analyze the puzzling behavior of the volatility of individual stock returns over the past few decades. The literature has provided many different explanations to the trend in volatility and this paper tests the viability of the different explanations. Virtually all current theoretical arguments that are provided for the trend in the average level of volatility over time lend themselves to explanations about the difference in volatility levels between firms in the cross-section. We therefore focus separately on the cross-sectional and time-series explanatory power of the different proxies. We fail to find a proxy that is able to explain both dimensions well. In particular, we find that Cao et al. [Cao, C., Simin, T.T., Zhao, J., 2008. Can growth options explain the trend in idiosyncratic risk? Review of Financial Studies 21, 2599–2633] market-to-book ratio tracks average volatility levels well, but has no cross-sectional explanatory power. On the other hand, the low-price proxy suggested by Brandt et al. [Brandt, M.W., Brav, A., Graham, J.R., Kumar, A., 2010. The idiosyncratic volatility puzzle: time trend or speculative episodes. Review of Financial Studies 23, 863–899] has much cross-sectional explanatory power, but has virtually no time-series explanatory power. We also find that the different proxies do not explain the trend in volatility in the period prior to 1995 (R-squared of virtually zero), but explain rather well the trend in volatility at the turn of the Millennium (1995–2005).

We find that the sign of the correlation between institutional ownership and volatility depends on the firm’s dividend policy: institutional ownership is negatively (positively) related to volatility among non-dividend (dividend) paying stocks. The empirical results are consistent with an interaction between institutional preference for low volatility and the tendency of higher levels of institutional ownership to increase volatility through their trading behavior. This result is robust to many control variables and possible endogeneity concerns. Supporting our conjecture that institutions herd on dividend signals we find that the correlation between turnover and institutional ownership is higher for dividend paying stocks, and that the positive correlation between turnover and institutional ownership is higher on dividend declaration days. Finally, we also find that the level of institutional ownership drops following an increase in volatility for both dividend payers and non-payers, and that volatility rises following increased institutional ownership for dividend paying stocks.