As is well known, guns are heavily involved in violence in America, especially homicide. This has led many to the logical conclusion that America’s high rate of gun ownership must be at least partially responsible for the nation's high rates of violence, or at least its high homicide rate. But does the positive correlation between gun prevalence and homicide rates reflect a causal relationship? It is also possible that higher crime rates increase gun levels by stimulating people to acquire guns, especially handguns, for self-protection; and holdings of guns for self protection can deter crime. Causality in the guns-crime relationship may therefore run in either or both directions, and estimates of the impact of guns prevalence on crime must account for this simultaneity. This paper estimates the causal impact of gun levels on violence rates using a county-level cross-sectional dataset on every U.S. county with a population of at least 25,000 in 1990. We motivate our cross-sectional approach by showing that three recent attempts to address this question using panel data and time-series variation in gun levels and crime fall foul of a proxy problem: all relied on proxies for gun levels that are essentially uncorrelated over time with direct, survey-based measures of gun prevalence. In our analysis, gun ownership levels are proxied using the percent of suicides committed with guns, which recent research indicates is the best measure of gun levels for cross-sectional research. We instrument gun levels with three plausibly exogenous instruments that are correlated with gun prevalence in the US: subscriptions to outdoor sports magazines, voting patterns, and numbers of veterans. We report robust tests of instrument exogeneity and relevance, and find strong evidence of the existence of endogeneity problems in the guns-homicide relationship. When the problem is ignored, gun levels are associated with higher rates of gun homicide; when gun levels are treated as endogenous, this association disappears or reverses. Our results indicate that gun prevalence in the US has no significant net positive effect on homicide rates. We show that our instruments satisfy formal tests of instrument exogeneity, and we discuss the econometric interpretation and power of these tests to detect invalid instruments. We also show how to use priors about the violation of the assumption of instrument exogeneity to improve inferences about the guns-homicide link. |
