1. We construct a panel dataset that matches the original on all variables. For conflict onset, the outcome of interest, we use the same conflict dataset that these scholars used and apply the same binary coding to the "main dependent variable [of] conflict onset;Gleditsch;the results of this exercise, we briefly detail our replication and extension approach,2002
2. With respect to control variables, we include all variables used by the authors. Our sources of data differ only when the datasets used by Salehyan and Gleditsch (2006) do not cover the 2002-2013 additional years. Polity scores 5 , indicators for the presence of transborder ethnic groups 6 and, separately, of civil war in neighboring countries, and country distances are extracted from the same data sources and processed as described by;Gleditsch Salehyan;which neighbors are defined by borders falling within a distance of 100 kilometers or less,2006
3. We also use the R package splines to create a cubic smoothing spline with three interior knots for peace years. Results of this analysis are consistent with our sub-national analysis, as shown in S7, potentially suggesting that conflict dynamics between Salehyan and Gleditsch (2006)'s study period and the more contemporary period are different. While an analysis of the differences between those periods is beyond 5 Although we use the same data source, the most current version of that data has been modified since Salehyan and Gleditsch (2006) used it in their study;Cederman Wimmer;We calculate size of the largest ethnic group using ethnic power relations data,2002
4. Refugees and the Spread of Civil War