The Democratic Peace: Weighing the Evidence and Cautious Inference


  • Dafoe was responsible for the analysis of Mousseau (2013), Oneal for Gartzke & Weisiger (2013); all authors contributed to the writing of the paper. All replication files (data, Stata and R code, and log files) are on the Dataverse Network:

    We thank Charlotte Jandér for suggestions on our figures.


The greater peacefulness of jointly democratic pairs of states is an important finding in research on the causes of war. Here, we outline a set of criteria for evaluating critiques of such robust empirical associations and apply them to two recent articles. Mousseau (2013) claims that contract-intensive economies, as proxied by a measure of life insurance expenditure, account for the democratic peace. However, his research suffers from various problems of analysis and interpretation, including a miscoded dependent variable, a misleading specification for dyadic democracy, a suppression of heterogeneous associations, and a heavy dependence on imputation in which greater than 90% of the values of the central independent variable were (improperly) imputed. We estimate 144 specifications that build from Mousseau's models and control for life insurance expenditures, finding substantial, robust support for the democratic peace. Gartzke & Weisiger (2013) claim that the importance of the democratic peace has declined as the proportion of democracies in the international system has increased; but their tests are misspecified and do not address the issue they raise. There are also serious errors in their data. When these problems are corrected, we find that the peacefulness of democratic pairs has actually increased as the proportion of democracies grew after 1816.