Introduction

One of the eternal problems of behavioral social science is the creation of indicators that satisfy two often contradictory conditions, local validity and global relevance. To the extent that we create general theories we also by implication need indicators that can be applied to a relatively wide variety of situations. But as this variety increases so do the risks that the connection between indicator and concept weaken. For example, the manifestations of power may change over time; what is perhaps suitable in the nineteenth century is no longer so in the nuclear era. Since we would like to apply the concept of power to various circumstances, this may mean using indicators that reflect its changing nature (Stoll and Ward 1989). Indicator construction is a balancing act between idiosyncratic models designed for particular cases and models of general applicability. The research design may englobe dozens of countries over many decades as the research from the Correlates of War Project typifies, frequently covering all nations from 1816 to 1980; others such as Levy (1983) and Thompson (1988) use the period from the end of the fifteenth century to the present. If an indicator must be valid over the period from the musket to the atomic bomb this places extraordinary demands on indicator construction.