Abstract

We reexamine the existence of an upward trend in the idiosyncratic volatility (IV) of US-listed stocks between 1962 and 2000 documented in Campbell et al. (2001). Following the same methodology as in their paper, we confirm the upward trend and the subsequent reversal, as reported by Bekaert et al. (2012). However, taking a closer look, we find that this result is influenced by microstructure biases in realized variances and correlations based on daily returns. Correcting these biases, we find that, depending on whether we use equal- or value-weighted IV and whether we include the pre-NASDAQ period, between 17% and 62% of the trend is removed. Also, the subsequent reversal of IV is almost entirely removed once we correct for these biases. We also highlight the mechanical dependence of the IV measure on industry concentration. From 1962 to 2000, industry concentration also contributes to a higher IV. However, its effect remains quantitatively small.