Trade-Based Analysis of Momentum

Research by Soeren Hvidkjaer

The phenomenon of momentum—that stocks which increase in price one year will tend to continue to increase in price over the next year, and that stocks which decrease in price one year will continue to decrease in price over the next year—has intrigued researchers for a long time. What drives this anomaly? If momentum is driven by a particular kind of investor behavior, then that behavior should be detectable in the trading records, reasoned Soeren Hvidkjaer, assistant professor of finance at the Smith School.

Hvidkjaer studied the trading behavior of small investors to see whether it was consistent with behavioral explanations of momentum. In his paper, “A Trade-Based Analysis of Momentum,” Hvidkjaer examines transactions data for all ordinary common stocks in the New York Stock Exchange and American Stock Exchange between 1983 and 2002. With several billion trades and quotes, it is the largest dataset of its kind. Hvidkjaer distinguished between the trades of large and small investors using the volume of individual trades.

“I studied how investors trade and whether that trading is consistent with behavioral biases. I also looked at investor behavior to see if it could be directly linked to prices,” says Hvidkjaer. “There is limited understanding of how small individual investors actually trade in response to new information.”

Hvidkjaer found that small investors tended to purchase losing stocks during the six-month formation period of the momentum portfolio. The buying pressure became very strong on the formation date, but then a selling pressure gradually set in, which peaked almost a year later. This suggests that small individual investors both underreact and have a delayed reaction to news about low returns on stocks; small investors purchase those stocks when they see the stock price falling, perhaps perceiving it as a bargain. This buying pressure might prevent the stock price from declining further. While small investors do eventually sell losers, they react extremely slowly to the negative formation-period information.

Small investors appeared to have a delayed reaction to winning stocks. Among winners, there was a formation-period buying pressure which largely disappeared on the formation date, and a buying pressure again set in slowly over the following six months. Small individual investors did not appear to immediately purchase stocks with high returns, though buying pressure mounted slowly during the six months following the formation date.

Sorting momentum stocks into portfolios based on formation-period trading pressures revealed another interesting fact: that in the 20 years of transactions in this dataset, Hvidkjaer found that losers with strong small-trade buying pressures underperform losers with selling pressures over the subsequent year. This means that either small investors have a direct effect on prices despite their small size or they tend to buy stocks which are already overvalued, notes Hvidkjaer.

“It seems that small investors are short-term contrarian investors and long-term momentum investors, which is not necessarily an optimal behavior,” says Hvidkjaer. “It may be that individual investors are confusing average past long-term returns with future expected returns.”

Hvidkjaer’s analysis found a completely different trading behavior by large traders: a strong large-trade selling pressure appears among losers during the formation period, followed by a drop on the formation date and a gradual disappearance over the following year. Buying pressure for winners seemed symmetric. Hvidkjaer found little evidence for initial underreaction or delayed reaction among large trades. So small investors appear to trade very differently than larger institutional investors, says Hvidkjaer, and the evidence of this study points toward the trading behavior of smaller investors as a source of the momentum effect.

Hvidkjaer feels that it may become much more difficult to analyze the behavior of small investors in the future, because automated stock trading will make it more difficult to distinguish between institutional and individual investors. “Institutions used to trade in large sizes. If they wanted 50,000 shares, they would often buy 50,000 shares in one trade, and we could tell from the recorded data that it was a large trade. Now a computer can trade for you, and it uses sophisticated algorithms to find where liquidity is best and trades in smaller sizes. An institution will purchase 50,000 shares, but the purchase will be recorded as many smaller trades. So in the future we’ll have to come up with new measures that will distinguish institutional trading from the trades of small, individual investors,” says Hvidkjaer.

Hvidkjaer plans to continue to use this dataset in further research examining how stocks purchased by small traders underperform over the long run. Hvidkjaer’s paper, “A Trade-Based Analysis of Momentum,” was published in the Review of Financial Studies, Summer 2006. For more information, please contact shvidkja@rhsmith.umd.edu.