The Persistent Presidential Dummy

Abstract

Whether Republican or Democratic presidents are better for the the stock market has been closely scrutinized for years. Although much of it is discussed only in casual terms, a recent academic study by Santa Clara and Valkanov (2003) documenting that the market does significantly better under Democratic regimes was widely quoted in the financial press. Indeed, a link to the study's results was posted to the Kerry/Edwards website. This paper shows that the statistical tests applied by the authors of the study were wrong, and that, once corrected, the difference in stock market returns under different presidential regimes is not meaningful. The lessons of the paper extend well beyond the presidential effect and emphasize the importance of proper research design in carrying out statistical investigations.

Keywords: spurious regression, persistence, data mining, dummy variable

JEL Classification: G12, C15, C22