age and older (pop65), as well as the percentage of households receiving public assistance income
in a local jurisdiction (pasincpop). We expect these three local level variables to be negatively
correlated with our dependent variable.
Kodrzycki (1994) used the percentage of a local government’s general non-educational
expenditures that went to human services (measured as public welfare spending plus spending on
health and hospitals) as a measure of citizens’ preferences for governmental services; her results
indicate a strong negative relationship between this variable and the use of contracting out. Thus,
we include in our model a measure of state government spending on public welfare as a percentage
of total state government general expenditures (stdgwelf) and the same measure for local
governments spending (locdgwelf); the second variable is measured using data aggregated at the
state level.
To capture the political climate in the state, we include in our model a measure of the ratio
of Democratic to Republican state legislators (legdemo). For over two decades, modern
conservative political movements in the United States, the United Kingdom and elsewhere have
advocated privatization as a means for reducing the size of the public sector, curbing the power of
the state, and making public bureaucracy more productive and competitive (Savas 2000; Sclar,
2000; see also Breaux, et al., 2002). Thus, we expect this measure to be negatively correlated with
the extent of contracting out, even though some recent findings indicate no relationship between
political partisanship and the dependent variable (Brudney, et al., 2005; Warner and Hebdon, 2001).
Finally, our model includes two local level measures of political opposition to privatization. These
measures are dichotomous ICMA survey indicators of opposition to privatization from citizens
(q5a_1) and from elected officials (q5a_2). Both should be negatively correlated with the extent of
contracting out.
Overload hypothesis
9
Convention