Mutual funds: Is there correlation between past performance and future performance?

Free Exchange and Blodget agree: past performance is no guide to future results, when it comes to mutual-fund returns. Is this true? I posed the question back in 2004, when I wrote this:

When a rather obnoxious man at Citibank tried to sell me some mutual funds once, based on their outperformance, I actually spent quite a bit of time researching this issue. Companies like Morningstar generally group funds into quintiles: the top 20%, the next 20%, and so on. And there is in fact a certain amount of correlation between past performance and future performance. Not a lot, but a little. Funds in the bottom quintile will tend to underperform in the future, funds in the fourth quintile will underperform but not quite as badly, and funds in the top three quintiles are all roughly equally likely to outperform in the future.

This generated a little bit in the way of comments. Said Simian:

Actually the academic evidence points primarily to a random walk when it comes to mutual fund performance. Morningstar ratings - which are currently based on risk-adjusted, peer-relative trailing returns for various horizons (1 year, 3 year, 5 year, etc) - have generally been shown to be non-persistent (Blake & Morley). So prior good performance does not suggest that future performance will be particularly good or bad; it's merely noninformative. The studies which did find some, albeit not very strong, persistence (Warshawsky) suggested that economies of scale played a role. In other words, if a fund did well, assets poured into it, and a large fund is better able to control and allocate costs, to a degree which overcomes the increased market impact of its trades.

Which is actually the opposite of what Blodget is saying, which is that larger funds are less likely to outperform.

The big message, of course, is clear: don't buy mutual funds. And that's good advice even if it is followed by "every economist I have ever met," to quote the Economist's blogger. At the margin, however, is there some kind of correlation between past performance and future performance? I think Simian might have been referring to this paper, by Elton, Gruber, and Blake, but it says that "We find that past performance is predictive of future risk-adjusted performance".

Posted by Felix at 18:34 EST