Berry, William. “Testing Budgetary Theories with Budgetary Data: Assessing the Risks,” American Journal of Political Science, 30 (August 1986): 597-627.

Empirical tests of theories of budgeting and the expenditure choices of governments typically involve quantitative models in which the dependent and independent variables share common components. This can result in strong positive or negative statistical coefficients even in the absence of substantively meaningful relationships among theoretical concepts. Indeed, this methodological problem together with "incrementalism in budgeting" often poses a potential alternative explanation for empirical findings normally taken as confirmation of other theories of budgeting and expenditure choice. In this paper, I present a simulation approach for testing empirical results against the null hypothesis that the results are merely a consequence of incrementalism combined with "built-in" structural relationships among variables due to common "components" shared by the variables. The approach is illustrated with analyses of the methodologies of two recent studies: one of the welfare versus defense tradeoff in Western industrialized nations, and one of the state budgetary process.