Though of course one study is unlikely to be definitive. (http://slatestarcodex.com/2014/12/12/beware-the-man-of-one-s...)

Here are 15 more articles on mutual fund persistence: http://www.altruistfa.com/readingroomarticles.htm#Persistenc.... I have not read them all, but my impression of the literature is that:

(1) There is slight positive serial correlation in mutual fund returns

(2) Most of the slight positive serial correlation comes from bad performers staying bad performers (rather than good performers staying good performers)

Theoretically, I find the notion of positive serial correlation far more likely than negative serial correlation. It seems more plausible to me that good fund managers would stay good than that managers would tend to oscillate from good to bad to good on an annual time scale.

In general, negative serial time correlations are generally harder to explain because a negative number squared is positive. Any sort of "rebounding" effect will result in oscillatory behavior, which will only be detectable over a certain time scale. Measuring a positive linear relationship is easier than measuring a negative linear relationship, because the negative solution will have an associated time scale. (That is, solutions that drift/diffuse can be captured over many time scales & sampling rates, whereas solutions that oscillate can only be captured near the rate of oscillation. On time scales that are very zoomed in or very zoomed out, you won't notice that the system is oscillating.) Much of this is handwavy and applies only to linear correlation, but hopefully the point is clear nonetheless.