Comomentum

We propose a novel measure of the amount of arbitrage capital allocated to the momentum strategy to test whether arbitrageurs can be destabilizing. Our measure, which we dub comomentum, aims to capture the extent to which momentum trades by arbitrageurs become crowded based on high-frequency abnormal return correlations among momentum stocks. Consistent with this aim, we find that comomentum is significantly correlated with existing variables plausibly linked to the size of arbitrage capital. Furthermore, comomentum explains and forecasts time-series variation in important characteristics of momentum returns. First, in the formation period, higher comomentum is associated with higher return spreads between winner and loser stocks. Second, in the holding period, comomentum forecasts relatively-low momentum returns, relatively-high momentum return volatility, and more negative momentum return skewness. Third, in the year after the holding period, high comomentum forecasts a strong reversal to the momentum strategy. Taken together, these patterns indicate that in periods of low comomentum (i.e., low momentum capital), price momentum is an underreaction phenomenon, while in periods of high comomentum (i.e., high momentum capital), price momentum is primarily an overreaction phenomenon due to crowded trading by arbitrageurs which destabilizes stock prices. We argue that such destabilization should be expected for a strategy like momentum where arbitrageur demand is increasing in price and is not a function of some fundamental anchor. Indeed, in a placebo test, we find that a similar measure for the value strategy is consistent with price stabilization. Finally, we examine the trading behavior of long-short equity hedge funds in response to our comomentum measure. We find that the typical hedge fund is able to time the momentum strategy in a way consistent with comomentum. However, this timing ability deteriorates as the fund gets larger. Our main data sources include the CRSP stock return files, Compustat financial data, and Lipper’s Tass database.