External validity captures the extent to which inferences drawn from a given study's sample apply to a broader population or other target populations. Social scientists frequently invoke external validity as an ideal, but they rarely attempt to make rigorous, credible external validity inferences. In recent years, methodologically oriented scholars have advanced a flurry of work on various components of external validity, and this article reviews and systematizes many of those insights. We first clarify the core conceptual dimensions of external validity and introduce a simple formalization that demonstrates why external validity matters so critically. We then organize disparate arguments about how to address external validity by advancing three evaluative criteria: model utility, scope plausibility, and specification credibility. We conclude with a practical aspiration that scholars supplement existing reporting standards to include routine discussion of external validity. It is our hope that these evaluation and reporting standards help rebalance scientific inquiry, such that the current obsession with causal inference is complemented with an equal interest in generalized knowledge.

Despite growing concerns about the effects of environmental changes, we only have disparate and seemingly contradictory findings about the relationship between natural disasters and intrastate violence. This article addresses that problem by introducing postdisaster reconstruction as a primary explanatory variable for intrastate violence. I extend bargaining theory to predict that postdisaster reconstruction causes a commitment problem, which in turn incentivizes warring parties to fight for the strategic opportunities of reconstruction. Using an instrumental variable approach, I provide an empirical test with a subnational data set for Sri Lanka before and after the 2004 Tsunami. Consistent with my expectations, housing reconstruction increased the number of violent events, while housing destruction had no discernible impact on violence.

Where do armed conflicts occur? In applied studies, we may take ad hoc approaches to answer this question. In some regression studies, for instance, a single conflict event can cause an entire province to be classified as a conflict zone. In this paper, I fill this void of knowledge by developing a machine learning method that is less dependent on the areal-unit assumptions and can flexibly estimate conflict zones. I apply the method to a conflict event dataset and create a new dataset of conflict zones. A replication of Daskin and Pringle (2018, Nature 553, 328–332) with the new dataset indicates that the effect of civil war on mammal populations is much smaller than the original estimate.