Working Paper: NBER ID: w2640
Authors: Alex Kane; Stephen Gary Marks
Abstract: Previous investigators have shown that the Sharpe measure of the performance of a managed portfolio may be flawed when the portfolio manager has market timing ability. We develop the exact conditions under which the Sharpe measure will completely and correctly order market timers according to ability. The derived conditions are necessary, sufficient, and observable. We compare them to empirical estimates of actual market conditions, and find that the circumstances which can lead to a failure of the Sharpe measure do in fact occur. We show, however, that such failures can be greatly reduced by more frequent sampling.
Keywords: Performance Evaluation; Market Timing; Sharpe Measure
JEL Codes: G11
Edges that are evidenced by causal inference methods are in orange, and the rest are in light blue.
| Cause | Effect |
|---|---|
| Sharpe measure (C29) | misclassification of market timers (G14) |
| unobserved shifts in portfolio composition (G11) | non-normal distribution of returns (C46) |
| greater ability (D29) | higher Sharpe measure (G19) |
| frequency of data sampling (C83) | accuracy of Sharpe measure (C58) |
| market conditions (P42) | performance evaluation of market timers (G14) |
| Sharpe measure's ability to order performance (C52) | distribution of returns (D39) |