| Summary/Abstract |
Three major datasets contain problematic interpretative judgments, arguably biased toward the United States: the Polity dataset; Reiter and Stam's data on war outcomes; and Singh and Way's data on nuclear proliferation. These examples raise the possibility that important datasets in global security studies, and in political science more generally, are systematically affected by an American bias. Bias means that, non-Americans might code the same observations differently, on average. The issue arises because Americans, on average, seem to have certain predispositions that non-Americans, on average, do not have. Other nationalities have their own predispositions. I also demonstrate that each of the three empirical examples has significant implications for causal inferences, altering certain statistical findings based upon them. For instance, I reexamine Haber and Menaldo's study of the resource curse, showing that alternative data coding casts substantial doubt on their inferences.
|