The authors find that index funds are run by skilled managers whose deviation from a purely passive strategy attached to the fund’s benchmark contributes to the variability of index fund returns and to the significance of their excess returns. Moreover, although the incremental performance of active funds relative to index funds is evident in the top-performing decile of funds’ excess returns, active funds do not seem capable of beating index funds across the entire return distribution, which raises more questions as to whether the fees charged by active funds are commensurate with the performance gains these funds achieve over the long term.

The authors’ results have implications for asset managers when selecting between index funds and actively managed funds. Further, for active managers, index funds with similar risk characteristics may be nominated as better relative performance benchmarks than existing ones, such as the standard or synthetic indexes used in the asset management industry.

The authors quantify fund overperformance by estimating the alpha of each fund relative to a series of benchmarks. They apply six benchmark models. Some of the models are well recognized (e.g., the market model and the Carhart three-factor model), and some are newer models centered on benchmark indexes (e.g., the seven-factor Cremers–Petajisto–Zitzewitz model).

Because of the large dispersion of index fund returns, performance is also measured by the t-statistic of estimated alpha (i.e., t-alpha) in order to gauge the significance of abnormal performance with respect to the level of risk borne by the index fund manager. The analysis is based on gross rather than net-of-fees fund returns in order to disregard the impact of differences in fees charged between active and index funds.

First, the t-alpha of index funds is found to be higher than the t-alpha of a bootstrapped return distribution (which, by construction, will have a population average “alpha” equal to zero) in more percentiles than the t-alpha for active funds, thereby indicating that skill among passively managed funds is at least the same as that of active funds.

Moreover, the percentage of skilled versus non-skilled funds is estimated by running the false discovery rate methodology of Barras, Scaillet, and Wermers (Journal of Finance 2010), which is higher in the case of index rather than active funds in most benchmark models. Also, the positive association between past fund performance and following net fund flows is traced to the index fund universe. This result indicates that “smart money” monitors performance not only in active but also in index funds in order to be invested in those with a recent positive track record.

The authors show that although active funds do worse than index funds when markets are on a downturn, they do not adequately make up for this cost when markets improve. These results are confirmed using quantile regressions. Finally, tests of second-order stochastic dominance on fund alphas imply that we cannot accept that active funds’ outperformance dominates index funds’ outperformance and we cannot reject that index funds fare better than index funds across the entire alpha distribution.

The number of passively managed funds has grown considerably during the last two decades. Possibly, investors recognize that for the majority of active fund managers, returns across the business cycle do not make up for higher costs. Nevertheless, the selection of a passive fund is not random but is based on past performance relative to its tracking index, its peers and, expectedly, active funds with similar risk exposures. If the growth of index funds continues, active funds may end up an investment choice that appeals only to the not-so-risk-averse investing community.