Autocorrelation
Properly computing the risk of a multi-manager portfolio depends not only on how funds returns depend on markets and sectors but also on how past performance influences future performance. Current returns of a fund may depend on its past returns. This property of a return time series is called autocorrelation. Autocorrelation can have a significant impact on the estimation of risk and return statistics. Autocorrelation also contains valuable information about the behavior of fund returns. If there is statistically significant autocorrelation, it has to be taken into account in the risk model to get an accurate picture of the risk of the fund. Autocorrelation can be detected by using the autocorrelation plots in the system. To account for autocorrelation you can include autoregressive terms in the factor model.
Autocorrelation may arise because of several reasons. Funds which trade in illiquid securities tend to exhibit significant levels of autocorrelation. Autocorrelation may also be indicative of returns smoothing practices. For technical strategies the autocorrelation pattern may reveal whether the fund is employing a following approach or is following a mean reverting strategy.