The Time-Series Behavior and Pricing of Idiosyncratic Volatility: Evidence from 1926 to 1962

The recent literature on idiosyncratic volatility has documented three main empirical findings. First, Campbell, Lettau, Malkiel, and Xu (2001) show that idiosyncratic volatility exhibits an upward trend between 1962 and 1997. Second, Goyal and Santa-Clara (2003) find that aggregate measures of idiosyncratic volatility predict one-month-ahead excess market returns from 1962 to 1999. Third, Ang, Hodrick, Xing, and Zhang (2006a) report a negative and significant relation between idiosyncratic volatility and cross-sectional stock returns from 1963 to 2000. The first result has led to a number of papers attempting to explain the causes of this upward trend, while the second and third results remain highly controversial. We re-examine these three findings using a 37-year holdout sample of daily returns from 1926 to 1962. We find robust empirical evidence of (1) a statistically significant downward trend in idiosyncratic volatility, (2) an insignificant relation between average idiosyncratic volatility and one-monthahead excess market returns, and (3) a highly significant inverse relation between idiosyncratic volatility and cross-sectional stock returns. These results shed new light on the time-series behavior and pricing of idiosyncratic volatility.