One caveat of current literature on the value of active management is the lack of treatment for the performance measures that can be gamed. We propose to use the performance measure that can’t be manipulated with respect to the underlying distribution, time variation, nor estimation error, (the manipulation-proof performance measure (MPPM, Goetzmann et al. (2007)), to rank all active U.S. domestic equity mutual funds from 1980 to 2013 on a quarterly basis to analyze managerial skills. We find fund managers in the higher ranked persistently outperform lower ranked managers by posting higher gross and net fund returns, higher holding-based returns, and generating positive return gap. Analyzing the holdings of the portfolios indicates higher ranked managers hold stocks with higher information asymmetry, especially the growth companies that are younger, smaller, and with lower liquidity. Our results show that the spread on gross and net fund returns between highest ranked and the lowest ranked fund managers is between 49 and 52 basis points per month. The holding returns are statistically significant for up to six months indicating the stock picking skills exist for those higher ranked managers. Even though MPPM identifies managerial skills, the positive alphas may not be warranted due to their operating expenses.

One dominant approach to evaluate the causal effect of a treatment is through panel data analysis, whereby the behaviors of multiple units are observed over time. The information across time and units motivates two general approaches: (i) horizontal regression (i.e., unconfoundedness), which exploits time series patterns, and (ii) vertical regression (e.g., synthetic controls), which exploits cross‐sectional patterns. Conventional wisdom often considers the two approaches to be different. We establish this position to be partly false for estimation but generally true for inference. In the absence of any assumptions, we show that both approaches yield algebraically equivalent point estimates for several standard estimators. However, the source of randomness assumed by each approach leads to a distinct estimand and quantification of uncertainty even for the same point estimate. This emphasizes that researchers should carefully consider where the randomness stems from in their data, as it has direct implications for the accuracy of inference.