An exploration of the impact of macro-level governing structure on local public sector employment in the United States

Abstract

This research investigates how characteristics of the macro-level governing structure affect local public sector employment. The empirical model is tested using fixed effects with panel data that consist of 3031 county-type areas in the United States for five-year intervals from 1992 to 2012. We find that an increased level of fiscal decentralization is associated with larger labor input in the production of public services. Our statistical results also show that interjurisdictional competition leads to the growth of local public sector employment. Nevertheless, we do not find that spatial fragmentation and jurisdictional overlap have any significant relationship with the level of public sector employment.

Keywords:

• Public sector employment
• fiscal decentralization
• local fragmentation
• jurisdictional overlap

Notes

1. PSE is used to represent public sector employment throughout the paper.

2. MLGS is used to represent Macro Level Governing Structure throughout the paper.

3. As of the 2012 Census of Governments, there are 3031 county-type areas. Prior to 2012, there were 3033, 3034, 3043, and 3043 county-type areas in years 2007, 2002, 1997, and 1992, respectively. The change in the number of county-type areas occurred in the following states, including 1 in Kansas (2007–2012), 1 in Kentucky (2002–2007), 1 in Tennessee (1997–2002), 3 in Georgia (2002–2012), 7 in Massachusetts (1997–2002), and 2 in Alaska (2007–2012). The majority of the changes are due to the consolidation of the city and the county. After the consolidation, the Census of Government counted the combined city and county government as a municipal government, rather than a county government, in census reporting. To illustrate, the Unified Government of Greeley County in the state of Kansas is the result of the consolidation of Greeley County and the city of Tribune. It is also counted as a municipal government, rather than a county government, in Census Bureau reporting (2012). The composite units of the city and borough of Anchorage in the state of Alaska are counted as municipal governments in Census Statistics on governments. Because the city of Anchorage is counted as a municipal government, so the combined city and borough of Anchorage is still counted a municipal government in the Census data. The only exception is seven abolished counties in the state of Massachusetts. From 1997 to 2000, the state of Massachusetts abolished seven counties. These counties still remain, but the offices of the abolished counties have been administratively placed under state Secretariats. However, the Census of Government no longer counted them as county governments, in census reporting.

4. The US Census of Government classifies ‘parish’ governments in Louisiana and ‘Borough’ governments in Alaska as county governments for Census Bureau statistics on governments. Thus, it is more appropriate to use the term county-type areas.

5. The way to measure ethnic fractionalization is presented as follows:

where race or ethnicity i denotes the share of population identified as belonging to race i including White, Black, Hispanic, American Indian, as well as Asian and Pacific Islander.

6. We tried to use the fiscal decentralization level at 1972 as one more instrumental variable for fiscal decentralization when we ran the regression. However, after adding the state fixed effects, the variable-fiscal decentralization level at the 1972 was automatically dropped by Stata due to collinearity issue.

7. In Stata, we first ran 2SLS estimation with a robust VCE, and then performed Wooldridge’s (1995) robust score test and a robust regression-based test using ‘estat endogenous.’ In all cases, if the test statistic is statistically significant (usually with p-values less than 0.05), we can reject the null hypotheses and then conclude that the variables being tested must be treated as endogenous. Because there is only one instrumental variable, we did not perform tests of overidentifying restrictions.

8. We only found that the correlation between county population and population over 65 years is about 0.4720, and the correlation between county population and interjurisdictional competition is about 0.4027, but these should be acceptable. Also, adding county population might help to control for cross-sectional variation in county size for the regression model.