Random coe¢cients models of arms imports  Ron P. Smith Birkbeck College, University of London Ali Tasiran Middlesex University and Gteborg University November 2009 Abstract Prepared for Special Issue of Economic Modelling in honor of P.A.V.B. Swamy. This paper uses a large panel of data with up to 19 time-series ob- servations for almost 150 countries to estimate models of arms imports. Qualitative evidence suggest a non-linear relationship. As income and military expenditure grow, the propensity to import rst rises and then falls as a domestic arms industry develops. We face the di¢- culty that there is virtually no data on domestic arms procurement or production capability. We try to avoid this di¢culty by adopt- ing a random coe¢cient approach in order to identify any systematic inuences on import propensity, through the impact of military ex- penditure, size of the armed forces or income on unobserved domestic production capability. While a clear non-linear pattern is apparent in the cross-section relationship, once one allows for parameter hetero- geneity such a pattern is not apparent in the time-series. J.E.L. Classication: C23, C24, D74 Keywords: Sample selection, arms imports, random parameters.  We are very grateful to sta/ of SIPRI for help with the data and for comments on earlier versions at the September 2006 Turkish Economic Association Meeting Ankara, at the Department of Economics, Statistics and Information Technology, rebro University, in April 2007 and the Annual Conferences on Economics and Security in Bristol in 2007 and Thessaloniki in 2009. 1