[gjem97]

There's no discussion here of survivorship bias [1]. Without a discussion of how it is addressed, these results are highly suspect. Specifically, the benchmark takes into account those companies that lost enough value to fall out of the "bottom" of the index and are therefore not in your current list of companies. However its far from clear that your subgroup includes similar companies.

A naive approach to this analysis is as follows. Take all the companies currently in the Fortune 1000, and look to see which of them has had a female CEO in the period under study. Backtest with buys and sells on the dates of hire/termination.

If this is indeed how it was done, the benchmark will almost assuredly underpeform simply because of the underperformance of components that are no longer in the index today.

[1] http://en.wikipedia.org/wiki/Survivorship_bias