Flow Of Funds Studies
Any analysis of mutual fund cash ?ows will uncover interesting information about how and when mutual fund investors make decisions. It provides a chronology of trading decisions made before the fact (ex-ante), which can then be compared to market outcomes (ex-post).
One use for fund ?ow data is a measurement of market timing skill among fund investors. This is done by studying changes in in?ows and out?ows among asset classes and then comparing this data to future moves in the markets. The overwhelming evidence shows from these analyses that people don’t have timing skill. In fact, frequent changes in asset selection hurt portfolio performance by a signi?cant amount.
Flow of funds studies go back several decades, with the early studies ?nding that investors chase top-performing funds. One study from 1978 titled “Is Fund Growth Related to Fund Performance?” found that investors disproportionally added to top-performing funds over a 10-year period from 1966 to 1975.2 A similar study conducted in 1992 concluded that investors responded more strongly to high performance in aggressive actively managed funds by purchasing more of them than less aggressive funds.3
Fund ratings were also a factor in fund-chasing decisions. The Boston Globe and the Wall Street Journal both reported in 1995 that about 97 percent of new investments that year went into mutual funds that had previously been awarded four or ?ve stars by Morningstar. A 2001 study found that an initial Morningstar ?ve-star rating results, on average, in six months of abnormal ?ows (53 percent above the normal expected ?ow). The authors of that study also found signi?cant abnormal ?ow in the case of rating changes, with positive ?ow for rating upgrades and negative ?ow for downgrades.4
The Federal Reserve Bank of Atlanta conducted its own study and found that “mutual fund investors use raw return performance and ?ock disproportionately to recent winners but do not withdraw assets from recent losers.” The Federal Reserve report noted that because of this behavior, “mutual fund managers have an implicit incentive to alter the risk of their portfolios to increase the chances that they are among the winners.”5
ETFs tend to be used by people who are more active traders than traditional mutual fund investors, and this leads to more mistakes. Cash ?ow studies show extremely poor market-timing results by active ETF investors. TrimTabs Investment Research, a consolidator of mutual fund ?ow data, concluded that equity prices tend to fall after equity ETFs rake in large sums of money and rise after equity ETFs post heavy out?ows. Regression analysis suggests the probability that equity ETF ?ows are a contrary leading indicator of equity prices is more than 99 percent. This means the ?ow of ETF money predicts market changes with high accuracy—in the opposite direction!6
One mutual fund cash ?ow study after another has consistently shown the same performance-chasing phenomenon. Fund styles with superior performance and high fund ratings raked in the most money, and this usually occurs close to the time when these investment styles peak in performance.
Institutions Are Also Trend Followers
Performance chasing isn’t limited to individual investors. Pension fund committees exhibit similar behavior, although not to the same degree. Amit Goyal and Sunil Wahal examined the selection and termination of private investment managers by 3,400 pension plans between 1994 and 2003. Plan trustees showed a tendency to hire investment managers after they delivered positive excess returns. However, these new managers failed to deliver returns better than the managers who were terminated for poor performance.7
In a more recent study, Jeffrey Busse, Amit Goyal, and Sunil Wahalu used a new, survivorship bias-free database to examine the performance and persistence in performance of 4,617 active domestic equity institutional products managed by 1,448 investment management ?rms between 1991 and 2008. Controlling for the Fama-French three factors and momentum, the trio found no distinguishable alpha in the data.8