n recent years, an ever expanding political, economic, sociological, and even anthropological literature has tried to give convincing answers to each of the following two questions: (i) Why do we have civil wars in some countries and not in others? (ii) Why does post-conflict peace endure in some countries, while it collapses into renewed fighting in others? Building on our earlier work on political institutions, economic development, and civil war as well as the mentioned literature on causes and consequences of intrastate conflicts, their prevention and recurrence, we argue that addressing the second question ignoring the first almost inevitably leads to answers that are affected by selection bias. In fact, we can show that the non-random selection of civil wars biases not only estimates for the error term and the parameters in all directions, but can falsely indicate that duration dependence of post-conflict civil peace is present when it actually is not. Therefore, we first theoretically develop and then empirically test a rational institutionalist two-stage selection duration model that accounts for the non-random sample selection to explain the variation in the endurance of peace in countries with civil war experience. This is done simultaneously estimating the selection outcome (civil war) and the duration outcome (length of post-conflict peace) using full-information maximum likelihood estimation on a data set of more than 100 civil wars during the 1950-2000 period. ..PAT.-Unpublished Manuscript [ABSTRACT FROM AUTHOR]