observation) as a variable. However, this does not address the problem of multicollinearity and heteroscedastic disturbances in the crosssectional analysis. Consequently the regression analysis will often use a two-stage weighted least squares procedure to compensate for the systematic extraneous variance within countries. A two-stage estimation technique becomes necessary when the hypothesized causal linkages among the causal variables imply that some of them may be partly a function of other causal variables in that model and cannot be considered entirely exogenous. Relative political capacity and level of economic development are such variables. The rate of inflation, the share of central government expenditures devoted to subsidies, and the size of the government deficit as a percentage of GDP are often heavily influenced by the magnitude of relative political extraction. In such cases, the estimates for the models are two-stage weighted least squares estimates.
Some readers will inevitably conclude that this work shows excessive concern for methodological precision to the point where I appear to have labored mightily to bring forth a mouse. My response is that inattention to the methodological side of this enterprise would have resulted in bringing forth not a mouse, but a sewer rat.
An Empirical Test of the Components
of Aggregate Political Capacity
Table 1.2 presents the results of three tests to unravel the nonlinear effects of one of the dimensions of aggregate political capacity (APC), namely relative political extraction (RPE). The dependent variables used are the percentage change in gross domestic product (GDP) expressed in current dollars, lending risk as measured by the Institutional Investor country credit rating index, and public consumption as a percentage of GDP. All three components of aggregate political capacity are included as causal variables in order to track the direction of impact of any of them on the outcome variables when the observations are restricted to cases where relative political extraction is equal to or less than its grand mean (1.06). The differential effects of these variables on the outcomes across levels of economic development are included since these data do not permit pooling across levels of development (see Chapter 2 below). Since the initial position of countries on the dependent variable varies