Abstract
As more local governments consider consolidation of government functions, officials are concerned about the expected impact on expenditures. Using a treatment group of consolidated city-counties and a control group of city-counties that considered but rejected consolidation through the referendum process, the authors examine differences in per capita local government expenditures. The statistical analysis shows that per capita expenditures in consolidated communities are not statistically different from those that considered and rejected consolidation. These results suggest that consolidation is not likely to decrease expenditures in the typical consolidated local government.
The most recent recession and slow recovery have taken a toll on local government revenues. Facing tight budgets for the foreseeable future, many local governments are considering various strategies to decrease spending. Numerous types of structural and functional consolidation are among the policies that local governments are reviewing. 1 Among the questions often asked by local government officials contemplating consolidation is the expected impact on expenditures. The goal of this article is to address this issue by studying the most extreme form of local government consolidation: city-county consolidation where most local government functions of a city and county are combined. 2 In most consolidated governments, smaller municipalities and school districts are not included, and some city and county departments remain independent, although they may be governed by the elected body of the consolidated government. We examine the impact of consolidation on total expenditures in sixty-two cities and counties that have held consolidation referenda since 1970.
Most public goods are characterized by large initial fixed costs and small variable costs. Economies of scale and scope drive the effort to reduce cost and provide cost-efficient public goods. Proponents of consolidation posit that by increasing the service area, the merged government units take advantage of economies of scale and economies of scope, and the average cost of service provision decreases. In addition, combining local governments will reduce spending by reducing duplication and redundancy in the provision of public goods and services, so expenditures should decrease as should the number of local government employees.
There are several theoretical models that explain why these potential savings never emerge. Oates (1985, 1989) hypothesizes that even though a fragmented system of government technically means more competition, the small government units may be too small to take advantage of economies of scale. Other arguments for a decentralized system of government are based on the heterogeneity of preferences for public goods and services and a Tiebout-like mechanism of competition among local government units for residents that will lead to an efficient allocation of resources (Tiebout 1956).
Another influence on the debate is the Leviathan hypothesis advanced by Brennan and Buchanan (1977). In the case where government’s power to tax is not constrained through some mechanism, total tax revenue and total government expenditure will rise, similar to the principle of revenue maximization by businesses. The decrease in the number of independent government units through consolidation reduces the competitive pressure on local government units. As a consequence, the reduced competitive pressure allows government units to levy excessive tax rates to support unnecessarily large and inefficient government bureaucracies. The number of papers (Oates 1985, 1989; Nelson 1987; Eberts and Gronberg 1990; Marlow 1988; Grossman 1989; Schneider 1989; Zax 1989; Joulfaian and Marlow 1990; Hoyt 1990; Taylor 2000) in response to the Leviathan hypothesis bears witness to the importance of this issue. Hirsch (1959) predates the Leviathan hypothesis but introduces important concepts for the subsequent debate.
In response to the theoretical arguments for consolidation and the drive to reduce public expenditures, many local governments across the United States have proposed consolidation, especially consolidation of cities with the county that surrounds a landlocked city. However, the literature is missing a rigorous quantitative investigation of whether consolidation reduces public expenditures as proponents of consolidation hypothesize. This serves as the motivation of this study.
Literature Review
In his review of the literature examining whether consolidation leads to greater efficiency, Feiock (2004) concluded that consolidation has not produced the political, economic, or fiscal benefits that reformers tout but instead has led to increased taxes and expenditures. He points out that the consolidation charter is central to understanding how consolidation will affect the ability of consolidation to reduce costs/increase efficiency and that special interests have weakened charters so that the promises of consolidation have not been realized. In their review of the literature on the impact of city-county consolidation, Martin and Schiff (2011) find little evidence that local government consolidation improves efficiency through improved delivery of services or lowers costs for the same level of service.
A few case studies have included an examination of the impact of consolidation on expenditures and have found that consolidation did not decrease expenditures in large part due to differences in personnel costs, a large component of total expenditures, among consolidating jurisdictions. In his examination of personnel issues associated with the merger of Athens-Clarke County, Georgia, based on over 800 employee interviews over a two-year period, Condrey (1994) found that increased efficiencies were the result of spending cuts rather than from merging the two governments. Since the consolidated government’s charter mandated that no employees should lose their jobs due to consolidation and that city and county workers with similar responsibilities should receive uniform compensation by the end of the fourth year after unification, projected personnel costs actually increased 7 percent. Vojnovic (2000) examined the transition and short-term effects of consolidation in five Canadian municipalities. Of the five, one city, Abbotsford, limited salary increases to an average of 1.5 percent while the other cities had considerably larger increases. The ability of Abbotsford to contain salary increases was due to explicit agreements regarding salary changes before the consolidation. Overall costs increased (or were expected to increase) in three municipalities and decreased in two.
Other studies have found increases in pay and per capita expenditures after consolidation. In her analysis of consolidation in Ottawa, Canada, Reese (2004) noted that pay and benefit levels tended to increase as differing pay systems and union contracts were renegotiated. Since wages and salaries are usually the largest component of local government budgets, increases in pay and benefit levels are likely to be associated with increases in overall expenditures. In their analysis of the Athens-Clarke consolidation, Selden and Campbell (2000) found cost savings in some departments and in real operating expenditures but overall an increase in real and per capita expenditures. The findings of these studies suggest that the act of consolidating city and county government is unlikely to lead to lower expenditure levels. The decisions of local government officials and personnel are the driving force behind spending decisions. So the structure of local government, consolidated or not, may have little impact on overall expenditures.
Leland and Thurmaier (2010) examined nine comparative case studies of consolidated and similar nonconsolidated cities and counties to test the hypothesis that consolidated governments operate more efficiently than nonconsolidated governments. They found limited evidence that consolidation leads to increased efficiency (measured as expenditure growth per capita) and concluded that efficiency gains are not a predictable outcome of consolidation.
Empirical Strategy and Hypotheses
The approach in this study is to test for differences in per capita expenditures among cities and counties that approved consolidation via referendum and those that rejected consolidation. We use a regression approach incorporating a dummy variable to distinguish between those local governments that approve the consolidation referendum and those that do not while controlling for other factors that influence local government expenditures. This is fundamentally an issue of technical efficiency: the savings from economies of scale, reducing duplicate services and professional management practices that result when two jurisdictions are combined. We are interested in determining whether evidence of these efficiency gains exists. Proponents argue that consolidation should increase efficiency and decrease the average costs of government. There is a theoretical basis for such claims, but to date quantitative evidence to support or refute these claims is lacking.
We test the following hypothesis: counties with consolidated city-county government have lower average expenditures than counties without consolidated local governments. We test this hypothesis for total local government expenditures.
Data
To test this hypothesis, this study employs data on sixty-two communities that voted on city-county consolidation referenda between 1970 and 2002, some more than once. The data on communities voting on consolidation referenda were initially compiled from Leland and Thurmaier (2006). We limit our analysis to the lower forty-eight states and to those states where communities voted on consolidation referenda. 3 The consolidation referenda passed in twelve of these communities. 4 The referenda data were augmented with county area data from the 2002 Census of Governments and the 2000 Census of Population and Housing. 5 County area data in the Census of Governments are an aggregation of expenditures for all local governments in a county. 6 The city governments considering consolidation are the largest in the county.
After adjusting for inflation, twelve of the forty-six (26 percent) counties with nonconsolidated governments had lower per capita expenditures after the referenda than before. Of the counties with consolidated governments, two of eleven (18 percent) had lower per capita expenditures after consolidation. In the typical (average) county after the referendum, per capita expenditures were 30 cents higher in nonconsolidated counties and 25 cents higher in consolidated counties. The standard deviation is larger than the mean for each type of government structure indicating a high level of variability in the data. Of the nonconsolidated counties, twenty-seven of fifty (54 percent) had per capita local government expenditures lower than all local governments in the state in 2002. Among the twelve consolidated governments, seven (58 percent) had lower per capita expenditures than all local governments in the state. On average, per capita expenditures were 23 cents higher in nonconsolidated counties and 14 cents lower in consolidated counties relative to the state where each is located, again with high variability in the data. 7
Statistical Method
Ordinary least squares (OLS) regression models with robust standard errors are used to test the hypothesis that, after controlling for other factors, consolidated city-county governments have lower average expenditures than nonconsolidated governments (among governments where consolidation has been considered via referendum). Our empirical strategy includes a series of regressions where consolidation variables are initially isolated, and then demographic and local government variables are added to examine whether the results are robust to alternative specifications. 8
The basic model takes the following form
Government size can be measured through tax revenue or government spending. While tax revenue may provide a good indicator of government size, government spending is more closely related to actual economic and political activity. For the purpose of this study, we use local government expenditures per capita (Eˉi ) as the dependent variable.
The consolidation variables (Ci ) include a dummy variable equal to one if the consolidation referendum passed and zero if the consolidation referendum did not pass and a variable measuring the number of referendum attempts between 1970 and 2002. The number of consolidation referendum attempts range from 1 to 5. Additional referendum attempts may put pressure on local governments to lower the cost of providing services or increase efficiencies. The number of consolidation referendum attempts measures this effect. The consolidation dummy is included to determine if there is a difference in average expenditures between consolidated local governments and local governments that considered but rejected consolidation.
Local government characteristics (Gi ) in the model are the share of revenue raised from intergovernmental transfers in 2002 and two measures of local government concentration/fragmentation: the number of local governments per square mile and the number of local governments per capita in a county. If the flypaper effect holds (Hines and Thaler 1995), we expect expenditures per capita to increase with the share of local government revenue from intergovernmental transfers. The measure of intergovernmental transfers is an aggregate measure spanning all functional services performed by local governments.
The number of local governments per square mile measures the spatial concentration of local government units in a county. The number of local governments per capita measures the number of local government units per resident. The signs for both government units per square mile and per capita are indeterminate. For counties with more government units per capita or per square mile, consolidation would allow government units to take advantage of economies of scales and scope, thus lowering local government expenditures per resident. However, if larger governments are less efficient, consolidation may be positively related to expenditures per capita or per square mile.
Demographic characteristics (Di ) in the model include median household income and population. If goods and services provided by local government are normal goods, we expect expenditures per capita to be positively related to income. Population serves as a measure of the size of the local jurisdiction. The relationship between population and expenditures per capita is indeterminate. If economies of scale exist for local government services, expenditures per capita should decrease as the population increases (holding the quality of services constant). However, a larger and more affluent population will demand more specialized public goods and services, increasing expenditure per capita.
In the final specification, we also include dummy variables for states to control for unspecified differences in state characteristics that influence per capita local government expenditures.
The variables included in these models are commonly used to explain local government expenditures or revenue (Zax 1989). We initially included additional demographic variables that were highly insignificant and were removed from subsequent regressions. 10
Results
Consolidation Variables
The regression results (Table 1) suggest that there is no difference in total expenditures per capita between consolidated local governments and those that considered but rejected consolidation. The consolidation dummy is negative and significant in the first two specifications indicating that per capita expenditures are lower in consolidated governments, but as additional variables are added to the model, this variable is no longer significant. The number of consolidation referendum attempts is positive and significant in the first specification but not in other specifications. These results are consistent with case studies that provide little evidence that consolidation decreases spending.
Regression Estimates Measuring Factors Affecting Per Capita Local Government Expenditures
Note: The residuals were estimated using White (1980).
***0.01 level of significance. **0.05 level of significance. *0.10 level of significance.
Local Government Characteristics
Intergovernmental transfers are a significant determinant of total expenditures. This result indicates that higher intergovernmental funding resulted in lower expenditures (by $5 to $8 per capita, $1.2 to $1.9 million in the average county) which is the opposite of what is predicted by the flypaper effect. This variable is significant in each of the specifications in which it is included indicating that it is a robust determinant of local government spending.
Government units per thousand population, a commonly used measure for the level of government fragmentation/consolidation, is a positive and significant determinant of local spending in two specifications. Local government units per square mile are not significant in any specification examined. We include general purpose governments (county, municipalities, and townships), school districts and special districts in our measure of local governments. Once state dummy variables are added to the model, these variables are no longer significant. This later result indicates that the number of local governments in a county has little influence on spending and refutes the argument by proponents of consolidation that a more fragmented system of government units will cause duplication of efforts, hence increasing costs.
Demographic Characteristics
Population and income are the demographic characteristics considered in the model. Population is positive and significant in five of the six specifications in which it is included, but the magnitude is small indicating that the size of the county has a small impact on per capita expenditures. When the state dummies are added to the model, population is no longer significant. Median household income is not a significant determinant of spending in any specification. 11
State Dummy Variables
The coefficients on the state dummies suggest that state characteristics have differential effects on local government spending. Georgia is the omitted category since this state had the most consolidation referendum attempts over this time period. With the exception of intergovernmental transfers, other variables included in the model become insignificant when state dummies are added. This result suggests that there are state specific factors, perhaps related to the legal and institutional environment, that influence local government expenditures.
The results of this analysis illustrate that controlling for state effects matters. The majority of local governments considering consolidation are located in the south. Of the sixty-two counties considered in the analysis, forty-three (69.3 percent) are located in the Deep South and thirty-five (56.4 percent) are from three states, Florida, Georgia, and Tennessee, suggesting that regional effects are also important. Voters in the South may be more aware of local government consolidation as a potential solution for local government problems since more referenda attempts have taken place in this region than any other part of the United States. As a result, the institutional knowledge and frameworks for mounting consolidation referenda would be more prevalent in this region.
Discussion and Conclusions
The statistical analysis presented in this article suggests that consolidation is not likely to decrease expenditures in the typical consolidated local government. It is possible that isolated consolidated governments do reduce spending, but in general this result is not likely. Lowering government expenditures is contingent on strategic planning and cooperation, conscious and deliberate. A few case studies indicate that savings from consolidation were achieved but were the result of improved local government processes not general effects resulting from economies of scale.
The presence of interlocal agreements and the debate during the referendum process may explain why there are few differences in spending between consolidated and nonconsolidated governments. Nonconsolidated governments may participate in various cooperative or regionalism schemes such as interlocal agreements that achieve the same results as consolidation. Similarly, if major city and county services were functionally merged before consolidation occurs so that few services are left to be merged, there would be little savings from consolidation. Leland and Thurmaier (2010) noted that this was the case for several of the consolidated governments considered in their book. Perhaps, the debate around consolidation itself during the referendum process leads to lower local government expenditures. Both sets of communities could have higher expenditures if there had been no consolidation debate.
Even if consolidation does not decrease government spending, there are other potential benefits. There may be improvements in the quality of government services. The available data do not measure changes in quality. Limited survey evidence shows that citizens of consolidated governments see improvements in quality of service provision after consolidation. DeHoog, Lowery, and Lyons (1990) found that citizens of consolidated governments have higher levels of satisfaction with local government in general and service provision in particular. Higher satisfaction may result from clearer lines of authority and accountability in consolidated governments. Consolidated governments may be better able to implement changes to service quality; changes that could have been implemented through interlocal agreements but lacked support in the two separate governments. Faulk and Hicks (2011) find that differences in per capita expenditures among local governments is not the driving force behind consolidation referenda and conclude that differences in service quality is a key determinant of successful referenda. Further investigation of the impact of consolidation on service quality is needed.
Footnotes
Declaration of Conflicting Interests
The authors declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.
Funding
The authors received no financial support for the research, authorship, and/or publication of this article.