Not So Smart Alphas

This note critically examines common methodologies for estimating alpha in actively managed funds and their reliability as measures of manager skill. Standard approaches often conflate genuine skill with statistical artifacts due to oversimplified risk assumptions, imperfect benchmarks, omitted factors, and misattribution of beta timing. Using two hedge fund examples, a market-and factor-neutral fund and a trend-following fund, we show that alpha estimates are highly sensitive to model choice and specification, with regression intercepts sometimes misleadingly positive and/or accompagnied with large confidence intervals. The analysis highlights the need for rigorous, multi-model evaluation and careful interpretation, emphasizing that alpha is a conditional metric, challenging to measure and persistently uncertain.