There are many models of fiscal policy in the economic literature and each has been based on a particular set of assumptions concerning the interaction of policy variables. However, even though these assumptions are critical to the behaviour of the models, there has as yet been no systematic attempt to test their validity or relative importance. Thus, in undertaking the research presented here, we were motivated by a desire to provide such a systematic study and to establish a general set of guidelines around which specific models of fiscal policy could be built.

Our methodology involved the identification of any empirical regularities during the past two decades that could be used to characterize the conduct and assess the impact of fiscal policy in Canada. For our purposes, fiscal variables were defined in terms of the various national accounts measures of government receipts, expenditures and transfers. Using two econometric models of the Sims type (1978; 1980), we examined the reactions of both aggregate and disaggregate fiscal variables to movements in five economic indicators: a measure of the gap between actual and potential income, inflation, potential output growth, a real interest rate measure, and real income per capita. The result is a stylized description of government policy that can be used as a basis for structural econometric modelling.

Our principal finding was that the fiscal variables of all levels of government are endogenous; they responded to movements in the economic indicators considered in the study, particularly inflation and cycles in real income. Of general relevance for econometric modelling, moreover, is the finding that neither highly aggregated nor highly disaggregated models of fiscal policy reflect the true extent of this policy endogeneity. Too much aggregation results in an underestimated measure of policy endogeneity because policies with dissimiliar characteristics are combined. Too much disaggregation also results in an underestimate of policy endogeneity because many policy interdependencies are overlooked. According to our results, the level of aggregation at which policy feedback is most evident occurs when government balances are disaggregated only into their revenue, expenditure and transfer components, and these components are not disaggregated further.

In addition, some of our more specific results have a bearing on a number of current issues in public finance. According to our measures, fiscal variables move more consistently in a contracyclical direction than some studies of fiscal policy would have us believe. We also found that, while inflation does systematically influence the revenues and expenditures of government, corporate tax revenues have been relatively insulated from the effects of inflation.