Conviction And Volume: Measuring The Information Content Of Hedge Fund Trading by Jonathan Rhinesmith, Scholar At Harward
I provide novel evidence that hedge funds predict and drive the movement of asset prices towards fundamental value. Willingness to move prices, proxied by the share of trading volume consumed, reveals information: the volume consumed by quarterly hedge fund trades strongly predicts future stock returns. The top decile of purchases generates abnormal returns of 5-9% annualized during the following quarter (t-stat 4.4-6.5). Interpreting this phenomenon using the Kyle model of price impact, I test for the empirical patterns one should observe if informed (hedge fund) trades incorporate information into prices. Informed trading impounds earnings news, reducing the reaction to positive earnings announcements by 28%. Informed trading also positively
predicts contemporaneous price movement and future informed trading. These price movements do not reverse. In contrast, mutual fund trades are significantly less informative. Structural and reduced-form estimates imply that consuming 1% of quarterly volume generates 0.3%-0.5% of price impact. Taken together, these results suggest that funds incorporate substantially more information into prices than is apparent from their fund-level returns.
Conviction And Volume: Measuring The Information Content Of Hedge Fund Trading – Introduction
In this paper I study hedge fund trading with two questions in mind. First, are hedge funds informed? Second, if so, how does their information get incorporated into prices? I show that trading volume plays a key role in addressing these questions.
Carlson Capital's Black Diamond Arbitrage Partners fund added 1.3% net fees in the first quarter of 2021, according to a copy of the firm's March 2021 investor update, which ValueWalk has been able to review. Q1 2021 hedge fund letters, conferences and more At the end of the quarter, merger arbitrage investments represented 89% of Read More
I draw on the intuition of the Kyle (1985) model that price impact is a function of volume. An informed fund should trade until the marginal cost of price impact equals the marginal profit of trading an additional share. Willingness to move prices reveals information: if large trades relative to volume cause price impact, then fund managers should only be willing to consume a large share of volume when their private information is particularly compelling. Following this intuition, I study the “volume consumed” – shares traded divided by total volume – by quarterly hedge fund trades.
I demonstrate that the cross section of volume consumed strongly predicts stock returns during the following quarter. The top decile of hedge fund equity purchases by volume consumed generates statistically significant outperformance of 5-9% annualized during the following quarter (t-stat 4.4-6.5). The top five deciles of purchases, representing 79% of purchases by dollar value, display statistically significant outperformance. I focus on purchases because I observe hedge funds’ long portfolios.1 These results suggest that hedge funds are informed.
To study how this information gets into prices, I test for the empirical patterns one should observe if the price impact of hedge fund trades incorporates information. Informed trades prior to the public revelation of earnings should impound earnings information into prices. The associated stocks should then react less when earnings news is revealed. Confirming this reasoning, I find that the reaction to a given positive standardized unexpected earnings surprise (SUE) is reduced by 28% for stocks in the top quintile of volume consumed relative to stocks with no hedge fund activity. I study positive surprises because of my focus on the information content of purchases. Though hedge fund purchases reduce the returns associated with a given earnings surprise, purchases nevertheless predict earnings returns unconditionally (before controlling for the level of the earnings surprise).
I provide three more important pieces of evidence that hedge fund trades incorporate information into prices. First, I show that the prices of high volume-consumed positions increase as hedge funds buy them. This pattern is consistent with price impact. Second, I show that trading is persistent across time. Purchases in quarter t predict purchases in quarter t+1. In quarter t+1, funds buy a greater share of volume in stocks with high quarter t volume consumed than in stocks with low quarter t volume consumed. During quarter t+1, the former positions perform better than do the latter positions. If funds do not cause price impact, then they are leaving money on the table by not building the former positions even faster.
Third, the cumulative outperformance of high volume-consumed positions is significantly positive out to a horizon of 2-4 years. Hedge fund trading is associated with fundamental information, which I define as persistent long-horizon price movements, rather than temporary price pressure, which would revert. This test rules out the possibility that hedge funds merely predict the price impact of their own future trades.
These results are based on trades identified from 13F filings. My hedge fund sample captures $200 billion of equity positions at a given time, on average, and over $500 billion by the end of the sample. The data covers $4.3 trillion of purchases, 1.0% of total volume. In contrast to large hedge fund trades, mutual fund trades that are large relative to volume are significantly less informative. Large mutual fund trades generate strong contemporaneous performance. Trades should cause price impact as they occur, regardless of information content. However, these trades predict at best marginally positive future performance, even after removing funds subject to extreme fund-level flows. This performance tends to revert over long time horizons, which should only occur for non-information-based trades.
Yet there is evidence that a subset of mutual funds are skilled. If informed volume reveals information, then volume consumed within this subset should predict future returns. I confirm this prediction using measures of skill from the literature.
See full PDF below.