ABSTRACT
Introduction During theColdWar, the arms race between theUnited States and theUSSR,which dwarfed all previous ones, was the focus of much concern. Since the end of the ColdWar, the focus has shifted to regional antagonisms, but these are still analysed using arms race models developed during the Cold War. The standard framework for empirical study of arms races is the Richardson model described below which explains the time-series pattern of military expenditure between potential enemies in an action-reaction framework. A coupled pair of differential equations explains changes in levels of weapons in each of two nations as a function of the weapons of each side. Once the process has started no country is at fault, the escalation is a consequence of systemic interaction rather than aggression. Despite its popularity the results of the Richardson model has given disappointing results when applied to real data, as Sandler and Hartley (1995: 106-107) note. This is partly because there are real problems inmoving from the abstract model to an empirical one. Any calibration requires decisions about the measurement of the variables, functional form, length of lags and expectation formation that are not specified in the theory. There are also likely to be problems with the quality and reliability of the data which make their use questionable. In addition, the estimation of these models – forward looking, dynamic, simultaneous equation systems – presents its own set of issues and problems.
