Reposted from Craig Newmark's blog, an article that claims that dividend paying stocks beat the market. I won't argue with the overall finding. I am sure it is true for the 1993-2007 time period. However, this doesn't mean that it has predictive ability for the future. The article smells strongly of data mining.
In the middle of the study period, the tech bubble burst, resulting in a massive loss in value for tech stocks, who tend to be non-dividend payers.
The moral of the story is that you can pretty much show any trading rule if you pick the right data and the right time period.
As a side note, my students should note that alpha is incorrectly estimated here because the study uses raw and not excess returns.