Supplemental Materials for “Presidents and Parties: How Presidential Elections Shape Coordination in Legislative Elections”

1.1 Case Selection Criteria and Presidentialism We elaborate on each of our case selection criteria in turn. First, the explanation for our chosen time period is straightforward. We saw no reason to not broaden our coverage both forwards and backwards in time relative to the literature, provided the appropriate data were available. This ensured that our presidential regime-only models would have a somewhat more adequate number of cases. Because data became difficult to obtain around the turn of the twentieth century, we somewhat arbitrarily chose to draw the lower bound at 1900; data also became difficult to obtain for very recent elections, which led us to the 2005 upper bound. As noted in the main paper and as demonstrated below, confining the analysis to the post-World War II period does not alter our conclusions. Second, as is conventional, we employ the minimalist, procedural definition and operationalization of democracy developed by Alvarez, Cheibub, Limongi and Przeworski (1996, 1999). This means that we study only those legislative elections in countries where the chief executive is elected; the legislature is elected; more than one party competes; and incumbents have actually lost elections. As noted in the main paper and as demonstrated below, neither controlling for pre-1990 OECD membership; eliminating elections in African countries; nor eliminating single country-elections, all relatively unconsolidated democracies, alters our conclusions. Third, also straightforward is our decision to confine the analysis to countries with a population of at least one million. Comparing elections in tiny Nauru (population approximately thirteen thousand) to elections in the United States (population approximately three hundred million) seems akin to comparing apples and oranges—especially when one is concerned, as we are, about the challenges of cooperating across districts. The literature implicitly takes a similar approach: the effective number of ethnic groups, one of its control variables, is only available for larger countries (see, for example, Fearon 2003), which means that small countries would otherwise need to be list-wise deleted from the analysis. Hence, it makes sense from a theoretical standpoint; for reasons of data availability; and in the interests of comparability to restrict our analysis to elections in larger countries. Fourth and finally, we exclude elections conducted under electoral systems with two relatively unusual features: fused elections and single, nation wide electoral districts. In fused elections such as post-1980s Bolivia, voters cast a single ballot for the presidency and the legislature, but distinct