Studies attempting to predict public expenditures using political variables have generally incorrectly assumed that political, like socioeconomic, variables function as determinants of policy levels and types. If one assumes, however, that the function of the political process in representative government is to translate citizens' demands for various levels of services as accurately as possible, then political variables must be conceptualized as mediating in character. In the absence of knowledge of whether public services are being oversupplied or undersupplied in relation to citizen demands, the effects of political variables can therefore, be better tested by moving beyond single-equation regression models.
An alternative method for examining impacts of political variables is presented by integrating them with conventional theories of political representation. This leads to a concept of “translation error,” and ways of examining relationships between this concept and political variables are explicated. Finally, the greater theoretical utility of this reformulation for the study of public policy is examined, and empirical studies which indicate its validity are cited.