Determining Military Expenditures: Arms Races and Spill-Over Effects in Cross-Section and Panel Data

This paper considers the determinants of military spending, building on an emerging literature that estimates military expenditure demand functions in cross-section and panel data, incorporating ‘arms-race’ type effects. It updates Dunne and Perlo-Freeman (2003b) using the SIPRI military expenditure database for the period 1988-2003, finding broadly similar results. It also shows differences in results across panel methods, particularly the within and between estimates and illustrates the importance of recognising and modelling dynamic processes within panel data. Heterogeneity is also found to be an important issue and when countries are broken up into groups on the basis of per capita income there is no obvious systematic pattern in the results. This is seen to imply that the demand for military spending, even between two mutually hostile powers, may depend on the whole nature of the relationship between them (and other countries and events in the region), and not simply Richardsonian action-reaction patterns.