The demand for arms imports

This article presents estimates of demand functions for arms imports for a panel of 52 countries, 1981–99, where there are non-zero observations for both the main measures, WMEAT and SIPRI. In principle, the WMEAT series is a value measure, while the SIPRI series is a volume measure, thus the ratio is a proxy for price. A baseline static log-linear model that makes arms imports a function of this proxy for price, military expenditure and per capita income, is estimated in a variety of ways. This shows significant price effects, with an elasticity around minus one; significant military expenditure effects with an elasticity below one; and no systematic effect of per capita income, though there is some suggestion that richer countries import less for the same level of military expenditure. Thus, there seems to be a relatively well-defined demand function of the form that has been assumed in much of the theoretical work. The article then examines the effects on the results of: measurement error in the proxy for price; choice of estimator; non-linearity; dynamic specification; and possible endogeneity of prices. In general, the results seem robust, though in cross-section there is a non-linearity – arms imports appear to rise and then fall as military expenditure increases – which is not apparent in time-series. The cross-section non-linearity may reflect the long-run effects of the development of domestic arms production capability on imports. However, because good data on arms production capability are not available, this explanation cannot be evaluated. Finally, the article reviews a range of potential criticisms of this approach, and areas for further research are also reviewed.