Using a sample of daily institutional trading data, the authors directly test theories of portfolio pumping and window dressing by institutions. They find evidence consistent with portfolio pumping but discover that abnormally low institutional selling, rather than abnormally high institutional buying, is the cause. They also find evidence inconsistent with window dressing.
What’s Inside?
The authors use a sample of daily institutional trading data to directly test theories of portfolio pumping (i.e., trading at period-ends to suggest better performance) and window dressing (i.e., altering period-end holdings by buying winners and selling losers) by institutions. They find evidence consistent with portfolio pumping but discover that abnormally low institutional selling, rather than abnormally high institutional buying, is the cause. In addition, they find that at year-ends, institutions tend to buy stocks in which they already have large positions. They also find evidence inconsistent with window dressing.
How Is This Research Useful to Practitioners?
Because the authors perform a direct test on actual institutional trading data, their results are more reflective of institutional trading behavior than those of previous research. Prior researchers have used such indirect sources as holdings, mutual fund returns, and stock return data.
In addition, the authors’ finding that portfolio pumping is a result of a larger decline in institutional selling than in institutional buying is new. Investors mostly observe the net effect—that is, that there is more institutional buying than selling at year-end—but knowing the source of the discrepancy allows investors to develop more appropriate investment strategies.
Finally, the evidence that is inconsistent with window dressing suggests that investors can rely on observed institutional holdings data. Window dressing is a problem because it distorts the period-end portfolio composition and could mislead investors into believing the portfolio performed better or had lower risk than it actually did.
How Did the Authors Conduct This Research?
To conduct their study, the authors use transaction-level institutional trading data from Abel Noser Solutions, which collects data directly from institutional investors’ trading systems. The data cover equity trading transactions made by a large sample of institutions from January 1999 through December 2010. Although the Abel Noser data contain transactions for investment managers and plan sponsors, the authors only report results for investment managers because, they argue, the incentive to distort performance is greater for investment managers.
The authors test theories of portfolio pumping and window dressing by institutions. To test the portfolio-pumping hypothesis, they create a monthly abnormal buy (sell) measure based on the percentage deviation in the last day’s institutional dollar volume relative to the average institutional dollar volume during the last five days of the month. They then add a control variable for total turnover in the security because it could be the cause of any observed price inflation. Total turnover is equal to the last day’s total turnover (i.e., shares traded/shares outstanding) as compared with the average total turnover during the last five days of the month. The authors use Center for Research in Security Prices (CRSP) data to calculate total turnover.
To examine the window-dressing hypothesis, they perform two tests. First, they determine whether managers buy stocks that performed well and do not buy stocks that performed poorly toward the end of the quarter or year. Second, they study the trading behavior at the beginning of the year. It is expected that when institutions buy winners to window dress at the end of the year, there will be less buying and more selling of winners at the beginning of the following year. The authors identify winners (losers) as stocks with a past one-year buy-and-hold return that places them in the highest-return (lowest-return) quartile among all stocks in the CRSP database.
Abstractor’s Viewpoint
By using actual institutional trading data, the authors’ results are more reflective of actual institutional trading behavior than are the results of previous research. The data the authors use, however, may not be as widely accessible as the periodic SEC filings made by institutional investors. Therefore, a better test of reliability of institutional holdings data would be to use the data the SEC require and that most investors observe.