Government form and performance: fiscal illusion and administrative ability in U.S. counties.

1. Introduction

Counties in the United States typically take one of several forms of government: council-elected executive, council--administrator, or council--commission. The primary differences lie in the separation of powers between executive and legislative functions, and whether the administrative duties are performed by an elected or appointed executive. The question is: Do these differences in government structure matter? And, if so, in what way?

In the council-elected executive (CE) form, the executive powers are in the hands of the county executive, who is elected directly by voters. The legislative powers rest in the hands of the elected council. In the council--administrator (CA) form, the executive and legislative powers are unified in the elected council. Administrative functions are delegated to the county administrator or manager, who serves as the chief executive officer appointed by and answerable to the elected council. In the council--commission (CC) form of government, the council of commissioners is responsible for budgeting and legislating while individual members serve as agency heads, performing the executive functions of one or more agencies. The commission typically shares administrative duties with the other county officials who are elected to run specific functions such as the county clerk, coroner, sheriff, tax assessor, and treasurer. The CE form exhibits the greatest degree of separation of powers while the CA and CC forms exhibit unified powers. At the same time, these forms also differ with respect to management structure. The CE and CC forms rely on elected administrative officers while the CA form relies on a professional administrator hired by the council. These characteristics are summarized in Table 1.

This study sheds new light on the relationship between organization form and government performance. We begin with a simple model of local government spending to identify the channels through which government form likely affects spending outcomes. The model assumes an output-maximizing bureaucracy constrained by the elected officials' desire to meet the voters' demand for the publicly provided services. The framework highlights the roles of managerial ability and asymmetric information in the form of voter fiscal illusion in the local public sector, factors leading to observable spending effects.

While the economics literature has not paid much attention to the separation of powers embodied in local government structure, arguments justifying separation of powers for the federal and state governments in the United States can also be used to explain its role in local government. On the one hand, it is often assumed that the public sector bureaucracy and possibly even elected officials pursue expansionary behavior when they can (Brennan and Buchanan 1980). If the separation of powers successfully introduces checks and balances that increase the responsiveness of both the administration and elected council to voters' demands, then it is more likely to curb expansionary tendencies of the public bureaucracy and lead to lower spending. On the other hand, the separation of powers in local governments typically gives the executive specific agenda control or veto powers in the budgetary process. Agenda control and veto power change the decision-making dynamics in ways that are difficult to predict. In sum, stronger separation of powers can arguably raise or lower spending levels. It remains an empirical question whether spending levels are higher or lower under separation of powers than under a unified executive--legislative structure. Rigorous empirical evidence is scant. Campbell and Turnbull (2003) find that separation of powers leads to greater spending only for Southern counties; they find no spending differences between separation of powers and unified forms of government for other regions of the United States.

In terms of professional versus elected administration, much of the economics literature concerned with local government form focuses on cities. There is, however, little agreement about how differences in this one dimension of government form affect spending. Booms (1966) argues that unelected professional management is not only more capable but is also freer of politics and the influence of interest groups by virtue of not having to answer directly to the voters, thereby leading to more efficient production and lower spending. (1) Hayes and Chang (1990) and Turnbull and Chang (1998) instead argue that elected administrators must answer directly to voters while professional administrators are hired by and answer to elected members of government. This means that there is an additional layer of a principal--agent relationship in the professional administrator form of government that is missing in the elected executive form. This additional principal--agent relationship might give the administrator wider latitude to pursue personal goals, like increasing the emollients (e.g., staff and perquisites) associated with greater spending or output. Further, the existing empirical evidence on this point is mixed as well. Booms (1966) and Turnbull and Chang (1998) conclude that the difference between professional and elected management matters for cities while Deno and Mehay (1987) find it does not. Focusing on costs for a select set of city services, Grosskopf and Hayes (1993) find no robust differences among different types of city government management, although Hayes and Chang (1990) find some differences for the larger cities in their sample.

The approach taken here identifies a simple formal link between local spending, managerial ability, and variations in the principal--agent relationships across government forms that may result in different degrees of voter fiscal illusion. The fundamental notion underlying the principal--agent relationship is asymmetric information; the principals (the voters) cannot perfectly monitor their agents (the administration and county career bureaucrats). We allow for the possibility that the additional layer of bureaucracy between the voter-taxpayers and the ultimate provision of the publicly provided service creates greater information asymmetry regarding the tax price--public service nexus. This type of information asymmetry is often viewed as the source of voters' fiscal illusion, which is the notion that a government tends to overstate the marginal benefits of spending and understate the marginal cost to taxpayers to increase its command over resources in the economy. Buchanan (1960, p. 60) traces this fiscal illusion notion to the Italian economist Puviani at the turn of the twentieth century. Modern applications to local governments view voter fiscal illusion as a systematic underestimation of marginal tax prices of publicly provided services, which by itself tends to increase public spending (Wagner 1976; Oates 1979; Turnbull 1998). Thus, the degree to which the additional principal--agent relationship inherent in the CA form of government (in which the administrator is the agent of the council, who are in turn agents of the voters) leads to higher or lower spending, reflects how it affects voter fiscal illusion and production cost.

There are several advantages from thinking about the effects of government structure within the context of fiscal illusion and production cost effects. The model offers a clear link between the separate branches of the literature dealing with spending effects and cost effects of government form and shows how they can be reinforcing or offsetting. The model is also empirically refutable, in that there are combinations of empirical estimates that lead to rejecting the theoretical framework. Finally, the theoretical framework provides the key relationships among estimable parameters needed to interpret the effects of organization structure on both fiscal illusion and cost efficiency.

The rest of the paper proceeds as follows. Section 2 presents the theoretical model to demonstrate the predicted effects of managerial ability and voter fiscal illusion on county spending in the United States. Section 3 reports the results of the empirical study, estimates that show how fiscal illusion and spending vary across government characteristics by separation of powers and type of administration. The results clarify previous conclusions about how government form affects local government spending. Decomposing the effects of organization form on spending into its components, we find that separation of powers leads to less taxpayer fiscal illusion and to services offered at greater cost relative to counties with unified decisionmaking structure. We find similar comparisons for counties relying on professional administrative officers relative to those with elected administrators. Section 4 concludes.

2. Fiscal Illusion, Administrative Ability, and Spending

Our analysis focuses on fiscal illusion and professional managerial proficiency as two channels through which government form affects local government spending. This section offers a simple model that combines the effects of managerial ability and fiscal illusion to provide a framework for interpreting the empirical results.

The local government spending model presented here assumes that the public sector bureaucracy maximizes public output subject to meeting voters' demands. The output maximization assumption for the public bureaucracy is widely used and is supported by a growing pool of empirical evidence (Blais and Dion 1991; Mitias and Turnbull 2001; Chang and Turnbull 2002). Under the fiscal illusion assumption, voters are not fully informed about the details of the public budget process when ratifying local government decisions or when deciding the politician for whom they will vote. An implicit assumption is that each individual voter finds it too costly to become fully informed even if he or she was to discover that he or she is not fully informed about local fiscal information. The fiscal illusion model assumes that voters know the total amount they are paying their local government in taxes and what they are enjoying in the way of publicly provided goods and services, but they do not know precisely how marginal changes in their individual tax bills will translate into increases or decreases in actual service levels or precisely how much additional services will increase their tax bills. Nor do they know the extent to which the local government obtains revenue from higher government grants or perhaps other key fiscal variables as well. (2) Simply put, this imperfect information environment allows output maximizing local governments to provide public services at lower perceived tax prices, so that voters end up supporting greater local spending than they would under perfect information. Of course, rational voters can continue to be mistaken about the true marginal tax prices of services in equilibrium only if their perceived total tax bills equal their actual tax bills--the consistency condition imposed below. Nonetheless, there are several consequences of this view of fiscal illusion that directly relate to the question of how government form affects spending, especially when combined with possible differences in managerial ability. This section considers the consequences of fiscal illusion for local government spending behavior.

Equilibrium County Spending

Define the following notation for the pivotal voter: x = consumption of the local governmentally supplied service, y = private goods consumption, p = price of private goods, m = money budget before local tax, and s = voter's share of the local tax base. Both x and y are composite commodities. The voter's utility is given by the well-behaved utility function u(x,y).

The total of the lump-sum intergovernmental grants received by the local government from higher level governments is G. The unit cost of providing the service is C(a), where a is an index of managerial ability. We assume that greater managerial proficiency--whether by the professional administrator or the elected administrator forms of government--results in lower costs, so that [C.sub.a] < 0. (3)

The voter's perceived marginal tax price for the locally provided service is t, which can differ from the true marginal tax price sC(a). Under our asymmetric information assumption, the voter does not observe the entire amount of intergovernmental grants received by the community. Denote by [phi] the proportion of the grants that escapes the voter's notice, 0 [less than or equal to] [phi] [less than or equal to] 1. Thus [phi] = 0 indicates full information (no fiscal illusion) and [phi] = 1 indicates complete fiscal illusion. (4) The voter's perceived tax bill for the service x is given by the voter's perceived share of total spending, tx, less the share of perceived intergovernmental aid, or

[T.sup.p] = tx - s(1 - [phi])G. (1)

The demand constraint on the output maximizing public bureaucracy is characterized by the pivotal voter's behavior under the perception constraints in t and [phi] reflecting the degree of fiscal illusion. The voter's problem is to maximize u(x,y) subject to the perceived budget constraint m = py + [T.sup.p], or

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

which yields utility u. It turns out to be convenient to characterize the demand side of the model using duality. For the equilibrium utility u, the perceived tax price multiplied by the quantity of the publicly provided good plus private spending is

e(t, p, u) = py + tx.

e(t,p,u) has all of the usual expenditure function properties with respect to its parameters and satisfies the following identity:

e(t, p, u) [equivalent to] m + s(1 - [phi])G. (2)

On the supply side of the model, the service providing government bureaucracy directly observes the entire amount of intergovernmental grants G. Following convention, the local government must satisfy a balanced budget constraint, which requires that total local taxes equal C(a)x - G. The individual pivotal voter's actual share of local taxes is s, which, following the usual approach in public spending models, is assumed to be exogenous. (Equivalently, the pivotal voter's property is assumed to capitalize fiscal differentials at the same rate as other taxable property in the jurisdiction.) The voter's actual tax bill is

[T.sup.a] = s[C(a)x - G]. (3)

Fiscal illusion does not mean that taxpayers behave irrationally, but rather that they systematically underestimate the marginal tax price of additional public services. But this type of misperception in equilibrium can be consistent with rationality only when individual taxpayers each end up paying the same total amount of taxes that they expect to pay according to their misperceived tax price. Therefore, for the voter's systematic misperception of the marginal tax price to persist in equilibrium, the voter's total perceived tax bill, [T.sup.p], must equal the actual total tax bill, [T.sup.a], in equilibrium. Equating (1) and (3) and solving for t, the perceived tax price in equilibrium must, therefore, satisfy what we label the consistency condition

t = s[C(a) - [phi]G/x]. (4)

In addition, the public service level satisfies the voters' demand as a constraint (2), hence Shephard's lemma,

[e.sub.t](t, p, u) [equivalent to] x, (5)

where [e.sub.t](t,p,u) is the compensated or Hicksian demand of the pivotal voter.

The system of Equations 2, 4, and 5 represent a concise statement of the complete model determining the equilibrium level of public service provision, perceived tax price, and taxpayer utility, {[x.sup.*], [t.sup.*], [u.sup.*]}. The equilibrium is depicted in Figure 1. When there is no fiscal illusion, the voter's perceived tax price is the actual marginal tax price, sC(a), the absolute slope of the actual budget line for the voter, cf. The vertical intercept of this line is m + sG and the voter maximizes utility at [x.sub.1], where the indifference curve [u.sup.1] is tangent to cf. Under fiscal illusion, however, [phi] > 0 and the voter's perceived marginal tax price is t, the absolute slope of the perceived budget constraint for the voter, ed. The vertical intercept of this line is m + s(1 - [phi])G. The consistency condition for equilibrium (that is, the condition that actual and perceived total taxes are the same for the voter) is fulfilled for the perceived tax price t that ensures the voter's indifference curve [u.sup.2] is tangent to the perceived budget constraint, where the perceived constraint intersects the actual budget line, at [x.sub.2]. In terms of the diagram, condition (2) ensures the tangency of [u.sup.2] and ed, (4) ensures that the tangency occurs where cf and ed intersect, and (5) defines the equilibrium service level [x.sub.2] satisfying the other two conditions.

Fiscal Illusion Effects on Spending

Totally differentiating the system of equations and solving for the comparative static results, we find the effect of greater fiscal illusion on public spending:

([partial derivative][x.sup.*]/[partial derivative][phi]) = - sG[e.sub.u][e.sub.tt]/D > 0, (6)

where

D = [e.sub.u] + ([e.sub.t][e.sub.tu] - [e.sub.u][e.sub.tt])s[phi]G/[x.sup.2] > 0 (7)

and the sign follows from the standard expenditure function properties for interior equilibria ([e.sub.u] > 0, [e.sub.t] > 0, and [e.sub.tt]< 0) and the assumed normality of x (i.e., [e.sub.tu] > 0). Given (7) the sign of (6) follows as well. This is the overspending effect of fiscal illusion: fiscal illusion increases equilibrium public spending (Turnbull 1998). Figure 1 provides a straightforward illustration of this result. The equilibrium under perfect information is [x.sub.1] and the equilibrium under fiscal illusion is [x.sub.2]; clearly, [x.sub.2] > [X.sub.1].

Managerial Ability and Cost Effects on Spending

Now consider how administrative ability or other aspects of cost management can affect the equilibrium. To do so, first note that a change in the pivotal voters' tax share, s, has two channels of influence on the equilibrium: It changes the underlying marginal tax price and it changes the voter's share of intergovernmental aid flowing to the locale. It turns out to be useful to isolate these two effects. Differentiate (2), (4), and (5) to get

[([partial derivative][x.sup.*]/[partial derivative]s).sub.d(sG) = 0] = ([e.sub.u][e.sub.tt] - [e.sub.t][e.sub.tu])C/D < 0 (8)

The sign follows (7) and the price effect on the voter's ordinary or Marshallian demand, which is ([e.sub.u][e.sub.tt] - [e.sub.t][e.sub.tu]) < 0.

We can now evaluate the effect of managerial ability. Differentiate (2), (4), and (5), and use (8) to simplify the result. Doing so, the effect of managerial ability can be expressed as

([partial derivative][x.sup.*]/[partial derivative]a) = [C.sub.a]s/C [([partial derivative][x.sup.*]/[partial derivative]s).sup.d(sG) = 0] (9)

Of course, the publicly provided service level, administrative ability, and even unit cost, are not observable in the empirical study. But we can still derive some useful conclusions for spending, a variable which is observable. Exploiting (9), the effect of administrative ability on observed total spending is

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (10)

where

E(Cx, s) = s/C[x.sup.*][([partial derivative]C[x.sup.*]/[partial derivative]s).sup.d(sG) = <0]

is the elasticity of public spending with respect to the pivotal voter's tax share (see Eqn. 11 below). This parameter depends, in part, on the underlying tax price elasticity of demand for the publicly provided goods. It is a reduced form parameter and is readily estimated using standard empirical methods. Once we have an estimate of this elasticity, the effects of managerial ability on spending immediately follow from using (10); managerial ability increases or decreases equilibrium spending because expenditure is elastic or inelastic with respect to the median voter's tax share, a relationship overlooked thus far in the literature concerned with the government form-spending nexus. It turns out that this relationship (10) plays a key role in the interpretion of empirical results, as explained in the next section.

Empirical Implications of the Model

It is useful to consider the full implications of the theoretical model before turning to the empirical results. In principle, it is straightforward to measure the effect of management form on total municipal spending. Ascertaining the individual influences of fiscal illusion and managerial ability, however, is a more subtle exercise. The previous comparative static results (6) and (10) provide some help in this direction.

Consider first the question of whether the one type of government form increases or decreases spending because it engenders greater or less fiscal illusion than another type of government form. As explained in the next section, it turns out that the expenditure data can be used to estimate the illusion parameter [phi] directly. We can use the difference in expenditures implied by differences in fiscal illusion for the two forms of government to examine the likely impact of administrative production cost between the alternative forms of government. Table 2 summarizes the possible cost relationships implied by the theoretical model. There are three key empirical estimates used in the comparisons: the degree of fiscal illusion (in column 1), the observed net effect of government form on total spending (in column 2), and the tax share elasticity of spending (in column 3). Using these estimates, we see that the model implies the cost relationships given in column 4 of the table, relationships arising from the otherwise unobservable differences in managerial ability.

For example, consider the first case listed in column 1. Suppose that we are interested in evaluating the performance of local governments with separation of powers relative to local governments with a unified executive-legislative decision-making structure. Suppose that we find that the separation of powers type of government leads to greater fiscal illusion than the unified form. By itself, (6) shows that the greater fiscal illusion under the separation-of-powers government tends to increase spending relative to the unified form. Therefore, if we also find that spending of separation-of-powers governments is less than or equal to that of unified governments (the first subcase listed in column 2) then the model implies that the difference between production costs for the two forms must be reducing separation-of-powers spending relative to unified governments. If the separation-of-powers form has superior administrative ability relative to the unified form, then (10) reveals that by itself the separation of powers' higher managerial ability (larger a) tends to decrease spending when [absolute value of [[beta].sub.1] < 1 and increase spending when [absolute value of [[beta].sub.1] > 1. If, in contrast, the separation of powers form exhibits worse managerial skills than the unified form then the city manager's lower a tends to increase spending when [absolute value of [[beta].sub.1] < 1 and decrease spending when [absolute value of [[beta].sub.1] > 1. Thus, for the case where spending by the separation-of-powers government is less than or equal to spending by the unified form government ([E.sub.sop] [less than or equal to] [E.sub.unified]), an estimated tax share elasticity [absolute value of [[beta].sub.1] < 1 is consistent with the separation-of-powers form of government exhibiting greater managerial proficiency than the unified form, indicated in the last column of the table. On the other hand, finding [absolute value of [[beta].sub.1] > 1 is consistent with the separation-of-powers form of government exhibiting less managerial proficiency than the unified form. Finding [absolute value of [[beta].sub.1] = 1 is inconsistent with the model and would lead us to reject the framework.

When spending is greater under separation-of-powers governments than under unified governments (the second subcase in column 2), we can infer nothing about the relative efficiency of the management types in the context of our model. The greater fiscal illusion of separation-of-powers governments leads to higher spending; as long as this effect overshadows the managerial ability effect (and we have no means of determining that it does or does not in this particular case), differences in ability alone can lead to higher or lower spending.

Continuing down the first column possibilities in Table 2, suppose instead that fiscal illusion is less under the separation-of-powers than the unified form of government, the third case in column 1. Because this by itself tends to decrease spending for the separation-of-powers relative to the unified form of government, [absolute value of [[beta].sub.1] > 1 and an observed spending effect [E.sub.sop] [greater than or equal to] [E.sub.unified] together imply greater production efficiency under separation of powers while [absolute value of [[beta].sub.1] < 1 implies less production efficiency under the separation-of-powers government relative to the unified. Of course, as illustrated by the last subcase listed in column 2, in this situation lower spending under the separation of powers form ([E.sub.sop] < [E.sub.unified]) allows us to infer nothing about relative production efficiency for the two types of government.

The relationships in Table 2 can be similarly applied to evaluate the relative performance of professional versus elected administration forms of government.

Broadly speaking, empirical studies of city spending tend to find inelastic demand for general spending. There are fewer county spending studies, but they also tend to find [absolute value of [[beta].sub.1] < 1 (Mitias and Turnbull 2001; Campbell and Turnbull 2003).

3. Empirical Evidence

The data set comprises 2243 counties in 38 of the contiguous U.S. states. Town governments in the New England states of Connecticut, Massachusetts, Maine, New Hampshire, Rhode Island, and Vermont perform the basic functions of counties in other states. These New England states are, therefore, not included in the sample. The sample also excludes three states (Arizona, Delaware, and Nevada) because each lacks a sufficient number of counties with complete data for analysis at the state level. Virginia cities are independent of counties; this institutional structure is unique so Virginia counties are omitted from the sample as well. There are several counties in other states that contain independent cities, and these individual counties are omitted from the sample as well. In addition, the sample excludes all unified city--county governments because their institutional forms fundamentally differ from the other counties. The 38 states in the data set are Alabama, Arkansas, California, Colorado, Florida, Georgia, Iowa, Idaho, Illinois, Indiana, Kansas, Kentucky, Louisiana, Maryland, Michigan, Minnesota, Missouri, Mississippi, Montana, North Carolina, North Dakota, Nebraska, New Jersey, New Mexico, New York, Ohio, Oklahoma, Oregon, Pennsylvania, South Carolina, South Dakota, Tennessee, Texas, Utah, Washington, Wisconsin, West Virginia, and Wyoming. Note that some counties in these states also lack complete data and so are also excluded from the sample. Table 3 summarizes data means and standard deviations for the partitioned samples used in the estimation.

The reduced form spending function follows the popular log-linear form typically used in the expenditure analysis literature, with spending a logarithmic function of the median voter's perceived tax price, t; median income, m; the median voter's perceived share of intergovernmental grants to the locale, s(1- [phi])G; population, N; population density, D; expenditure; and the Herfindahl index of expenditure concentration, H, and state dummy variables:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII.] (11)

Previous studies show that, although not necessarily appropriate for individual spending categories, for general spending the median property tax share and income variables most often associated with the single-tax and single-service median voter model provide a better depiction of the important demand side variables than do popular alternatives based on jurisdiction averages or per capita tax share and income measures (Pommerehne and Frey 1976; Pommerehne 1978; Turnbull and Djoundourian 1994; Turnbull and Chang 1998). Following this literature, we use median family income and the median value house divided by the jurisdiction property tax base as our income (m) and tax share (s) variables in (11), respectively.

The equations include population (N) and population density (D) as additional explanatory variables. These variables do not play a direct role in our study but are included as controls that have been identified with a variety of roles in the empirical literature. (5)

// The empirical model includes the Herfindahl index of expenditure concentration, H, to control for any spending effects arising from the budgetary structure of local governments aside from those that give rise to the systematic misperception of marginal tax prices envisioned in the fiscal illusion model just described. This variable is calculated using the Herfindahl concentration index of county spending on public safety, roads and highways, health, sanitation, and other expenditures. This variable provides an empirical measure of the overall level of budgetary complexity (Wagner 1976; Turnbull and Djoundourian 1994; Turnbull 1998). It is calculated as

H = [(public safety spending/total spending).sup.2] + [(roads and highways spending/total spending)].sup.2] + [(health services spending/total spending]).sup.2] + [(sanitation spending/total spending).sup.2] + [(other spending/total spending).sup.2].

The Herfindahl index ranges from a maximum of 1 (if spending is concentrated in only one category) to a minimum of 0.20 (if spending is allocated equally among all five categories). Counties spending most of their budgets on few categories of services have greater H indices and come closest to a single purpose government while governments spending on a wide range of services have lower H values reflecting more complicated fiscal systems. Turnbull (1998) argues that greater budgetary complexity indicates greater voter uncertainty about how incremental taxes translate into services and incremental spending translates into taxes. That paper also shows why greater budgetary uncertainty by itself decreases the demand for public spending by risk averse taxpayers, either reinforcing or offsetting the effect of fiscal illusion in the form of systematic misperception of the tax price identified in the previous section of this paper.

Total tax base (used in the construction of the tax share), county government general expenditure, expenditures across broad categories (used in the Herfindahl index of spending concentration), and state and federal aid receipts for 1992 are from the 1992 Census of Governments. The data for median house value (used in the construction of the tax share), median household income, population, and population density are 1990 population census data drawn from the 1990 Census of Population and the 1994 County and City Data Book. (6)

To derive the estimating equation, substitute the equilibrium perceived tax price (4) into (11) and rearrange terms:

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII],

where [[??].sub.0] = [[beta].sub.0] + [[beta].sub.3] ln(1 - [phi]). Since the dependent variable, E, appears in a highly nonlinear form on the left side of the previous equation, we use an iterative search procedure to estimate the parameters as follows. Conditional on [phi] and [[beta].sub.1], construct the variable

[z.sub.i]([phi], [[beta].sub.1]) = ln [E.sub.i] - [[beta].sub.1] ln (1 - [phi] [G.sub.i]/[E.sub.i]) (12)

and estimate the equation

[MATHEMATICAL EXPRESSION NOT REPRODUCIBLE IN ASCII] (13)

using least squares. The procedure is repeated, searching over values of [phi] and [[beta].sub.1] in (12) to maximize the likelihood function for (13). This method does not directly yield standard error estimates for [phi] and [[beta].sub.1], so the standard errors of these parameters are estimated using the bootstrap method.

Table 4 reports the parameter estimates for county governments broken down by elected (first and second columns) versus professional (third and fourth columns) administrators and separation of powers (fifth and sixth columns) versus unified (seventh and eighth columns) governments. Two versions of each model are presented, one with and one without the state dummy variables capturing state fixed effects. As is evident, the inclusion or exclusion of the state-fixed-effects dummy variables do not appreciably alter the parameter estimates. The price, income, and grants elasticities are all in line with the broader empirical literature and resemble to a surprising degree the type of results typically found for city governments. Of these, the price elasticity is of the greatest interest to this study; in all cases, the demand is inelastic.

Table 4 indicates significantly greater fiscal illusion in counties with elected administrators than in counties with professional administrators. This is the opposite of what the principal-agent argument would lead us to expect. It appears that the additional principal--agent relationship layered between the county CEO and voters in the CA type of government does not inexorably lead to greater voter fiscal illusion or its spending consequences. Table 5 reports the predicted spending levels under the alternative government characteristics evaluated at the pooled sample means. Whether using the fixed effects model or not, total spending is greater under professional administration than under elected administration. According to the estimates reported in Table 4, the demand for public spending is inelastic ([[absolute value of [[beta].sub.1] < 1). Therefore, referring back to Table 2 yields the conclusion that whatever the virtues of professional administration in terms of fiscal illusion, costs appear to be greater under professional than under elected administration. This is contrary to the rationale offered by advocates of professional city and county managers; professional training and certification, it is argued, makes professional administrators better able to organize inputs in the production process, deal with public sector unions, and avoid the entanglements with interest groups that are inherent in the election process. All of these attributes are supposed to lead to lower cost. But, while the parameter estimates reveal that these skills do lead to less fiscal illusion, that effect is outweighed by the effect of greater cost on total spending.

Now turning to separation of powers versus unified organization forms, the fiscal illusion parameter estimates in Table 4 indicate that separation-of-powers governments enjoy significantly less fiscal illusion than their unified counterparts. But at the same time, the spending predictions in Table 5 show that separation of powers leads to greater overall spending than in the unified decision-making structure. Again referring back to the relationships in Table 2, inelastic demand, therefore, implies that cost is greater under separation of powers than under unified governments. Apparently, the checks and balances and executive agenda control inherent in the separation of powers form are successful in reducing fiscal illusion while at the same time increasing the cost of services. Perhaps the organization advantages that bring clarity to the overall budgeting process and thereby reduce the effects of taxpayer fiscal illusion also hamper the government's ability to manage the provision of services in the most cost-effective manner.

4. Conclusion

The main differences between the different forms of local governments in the United States lie in the separation of powers and elected versus professional administration. With respect to separation of powers, we note that in the elected executive form of county government, the executive and legislative powers of the local government are split between the elected chief executive and the elected council, respectively. In contrast, the CA and CC forms exhibit unified executive and legislative powers in the elected council. With respect to elected versus professional administration, we note that both the elected executive and the CC forms of county government rely upon elected administrators while the CA form relies upon professional management. This paper examined how these differences in local government structure affect government performance in terms of fiscal illusion, production efficiency, and the resultant effects on spending. We presented a simple theoretical model linking spending, managerial ability, and variations in the principal-agent relationships across government forms. The principal--agent relationship between voters and the administration and government bureaucrats implies asymmetric information, the source of fiscal illusion in the locale. Thus, we argued that whether the additional layer of a principal--agent relationship inherent in the council--administrator form of government leads to higher or lower spending, reflects whether it results in greater or lesser voter fiscal illusion. Coupled with the presumably greater administrative ability of professionally trained public sector managers relative to their elected counterparts, the fiscal illusion and managerial efficiency effects may be reinforcing or offsetting. The net effect of government structure remains an empirical issue. Nonetheless, the model provides key relationships needed to interpret empirical results in terms of the relative performance of the different types of government

The empirical study yields new results. The estimates show that separation of powers leads to less taxpayer fiscal illusion than under unified executive--legislative government structure. But at the same time, the empirical model predicts that spending is greater under separation of powers than under the unified structure. Since the public goods demand is price inelastic, this implies that separation-of-powers governments are providing services at greater cost than their unified counterparts. The empirical estimates for comparing professional versus elected administration yield similar conclusions. While fiscal illusion is lower under professional administrators, the cost of the bundle of services is greater than in governments with elected administrative officers.

The author gratefully acknowledges the helpful comments and suggestions provided by two anonymous referees. The usual disclaimer applies.

Received September 2004; accepted January 2006.

References

Blais, Andre, and Stephane Dion. 1991. The budget-maximizing bureaucrat: Appraisals and evidence. Pittsburgh, PA: University of Pittsburgh Press.

Booms, Bernard H. 1966. City government form and public expenditures. National Tax Journal 19:187-99.

Brennan, Geoffrey, and James A. Buchanan. 1980. The power to tax: Analytical foundations of a fiscal constitution. New York: Cambridge University Press.

Buchanan, James A. 1960. Fiscal theory and political economy: Selected essays. Chapel Hill, NC: University of North Carolina Press.

Cambell, Rebecca J. 2004. Leviathan and fiscal illusion in local government overlapping jurisdictions. Public Choice 120:301-29.

Campbell, Rebecca J., and Geoffrey K. Turnbull. 2003. On government structure and spending: The effects of management form and separation of powers. Urban Studies' 40:23-34.

Chang, Chinkun, and Geoffrey K. Turnbull. 2002. Bureaucratic behavior in the local public sector: A revealed preference approach. Public Choice 113:191-210.

Deno, Kevin T., and Stephan L. Mehay. 1987. Municipal management structure and fiscal performance: Do city managers make a difference? Southern Economic Journal 53:627-42.

Grosskopf, Shawna, and Kathy Hayes. 1993. Local public sector bureaucrats and their input choices. Journal of Urban Economics 33:151-66.

Hayes, Kathy, and Semoon Chang. 1990. The relative efficiency of city manager and mayor-council forms of government. Southern Economic Journal 51:167-77.

Mitias, Peter M., and Geoffrey K. Turnbull. 2001. Grant illusion, tax illusion, and local government spending. Public Finance Review 29:347-68.

Oates, Wallace E. 1979. Lump-sum intergovernmental aids have price effects. In Fiscal federalism and grants-in-aid, edited by Peter Mieszkowski and William H. Oakland. Washington, DC: Urban Institute, pp. 23-30.

Pommerehne, Werner W. 1978. Institutional approaches of public expenditures: Empirical evidence from Swiss municipalities. Journal of Public Economics 9:255-80.

Pommerehne, Werner W., and Bruno S. Frey. 1976. Two approaches to estimating public expenditures. Public Finance Quarterly 4:395-407.

Turnbull, Geoffrey K. 1998. The overspending and flypaper effects of fiscal illusion: Theory and empirical evidence. Journal of Urban Economics 44:1-26.

Turnbull, Geoffrey K., and Chinkun Chang. 1998. The median voter according to GARP. Southern Economic Journal 64:1001-10.

Turnbull, Geoffrey K., and Salpi S. Djoundourian. 1994. The median voter hypothesis: Evidence from general purpose local governments. Public Choice 81:223-40.

Wagner, Richard E. 1976. Revenue structure, fiscal illusion, and budgetary choice. Public Choice 25:45-61.

(1) This point of view ignores the role of public demand in spending determination. Supplying the public good to voters at lower unit cost will increase or decrease total spending because the demand is elastic or inelastic.

(2) This is the essence of the theories of fiscal illusion based on systematic misperception. See, for example, Oates (1979), Turnbull (1998), and Mitias and Turnbull (2001). Alternatively, according to the uncertainty theory of fiscal illusion, voters know the planned or expected taxes or service levels, but they are uncertain about what the actual outcomes will be over the period covered by the agreed upon local government budget, largely because the actual outcomes are in part determined by stochastic or uncertain economic factors in the broader economy (Turnbull 1998). The misperception view of fiscal illusion is the more popular perspective in the local public finance literature and so is the view articulated in this paper.

(3) Jurisdiction population and population density can affect the public service provision-consumption nexus or service delivery costs. Therefore, we include population and density in our empirical models. To simplify notation, however, we exclude these variables from the theoretical model without loss of generality.

(4) The condition [phi] > 1 means that taxpayers perceive that their local government receives more intergovernmental grants than it does in actuality (Mitias and Turnbull 2001). It can be shown that this type of fiscal illusion would lead to lower levels of spending than under perfect information. Therefore, the condition [phi] > 1 cannot hold in equilibrium when the local government is maximizing spending subject to the voter-taxpayers' demand constraint because it could then both satisfy the voters' demand constraint and increase spending by offering voters the perfect information spending level, that is, by behaving as if [phi] = 1. Additionally, [phi] > 1 implies a reversed flypaper effect, so that income is more stimulative than intergovernmental grants on local spending, a result that runs contrary to the empirical literature.

(5) The population and population density variables are brought into the empirical model through their roles in consumption congestion. While it seems reasonable to include additional ad hoc socioeconomic variables in the empirical model, their inclusion or exclusion do not affect the conclusions of this study. This is not surprising in light of the existing literature for cities that tends to find no empirical role for these variables as determinants of local fiscal behavior once the structural variables implied by theory are included in the estimating equations. The relevant literature for counties is sparse, but see, for example, Campbell and Turnbull (2003) and Campbell (2004).

(6) The Census of Governments discontinued collecting the assessed valuation of county property tax bases in 1997.

Geoffrey K. Turnbull, Department of Economics, Georgia State University, P.O. Box 3992, Atlanta, GA 30302-3992, USA; E-mail [e-mail].

Table 1. Characteristics of the Prevalent County Government Types

Executive-
Legislative
Powers            Government Forms

Separate          Council-executive
Unified           Council-commission, council--administrator

Management Form   Government Forms

Elected           Council-executive, council--commission
administration

Professional      Council-administrator
administration

Table 2. Fiscal Illusion, Observed Spending Differences and Implied
Cost Differences: Comparing Government Type i with Type j

Fiscal Illusion    Expenditures          Price         Implied Cost
                                      Elasticity        Comparison

[[phi].sub.i] >   [E.sub.i] [less      [absolute        [C.sub.i] >
  [[phi].sub.j]    than or equal       value of         [C.sub.j]
                   to] [E.sub.j]    [[beta].sub.1]
                                          > 1

                                       [absolute       Inconsistent
                                       value of         with model
                                    [[beta].sub.1]
                                          = 1

                                       [absolute        [C.sub.i] <
                                       value of         [C.sub.j]
                                    [[beta].sub.1]
                                          < 1

                    [E.sub.i] >      [MATHEMATICAL    No implication
                     [E.sub.j]      EXPRESSION NOT     for relative
                                    REPRODUCIBLE IN        cost
                                        ASCII]

[[phi].sub.i] =     [E.sub.i] >      [MATHEMATICAL     [MATHEMATICAL
  [[phi].sub.j]      [E.sub.j]      EXPRESSION NOT    EXPRESSION NOT
                                    REPRODUCIBLE IN   REPRODUCIBLE IN
                                        ASCII]            ASCII]

                    [E.sub.i] =      [MATHEMATICAL      [C.sub.i] =
                     [E.sub.j]      EXPRESSION NOT      [C.sub.j]
                                    REPRODUCIBLE IN
                                        ASCII]

                    [E.sub.i] <      [MATHEMATICAL     [MATHEMATICAL
                     [E.sub.j]      EXPRESSION NOT    EXPRESSION NOT
                                    REPRODUCIBLE IN   REPRODUCIBLE IN
                                        ASCII]            ASCII]

[[phi].sub.i] >      [E.sub.i]         [absolute        [C.sub.i] <
  [[phi].sub.j]    [greater than       value of         [C.sub.j]
                   or equal to]     [[beta].sub.1]
                     [E.sub.j]            > 1

                                       [absolute       Inconsistent
                                       value of         with model
                                    [[beta].sub.1]
                                          = 1

                                       [absolute        [C.sub.i] >
                                       value of         [C.sub.j]
                                    [[beta].sub.1]
                                          < 1

                    [E.sub.i] <      [MATHEMATICAL    No implication
                     [E.sub.j]      EXPRESSION NOT     for relative
                                    REPRODUCIBLE IN        cost
                                        ASCII]

Table 3. Sample Summary Statistics

Variable                        Observation   Mean

Elected executive sample
  Expenditure                      1969            3.96E+07
  Tax share                        1969            0.0005705
  Income                           1969       23,385.94
  Grant                            1969         1628.094
  Population                       1969       68,466.28
  Density                          1969          125.297
  Spending concentration           1969            0.5132225

Professional executive sample
  Expenditure                       274             1.92E+08
  Tax share                         274             0.0002787
  Income                            274        27,793.19
  Grant                             274          3119.094
  Population                        274       224,395.1
  Density                           274           298.1686
  Spending concentration            274             0.5679558

Separation of powers sample
  Expenditure                       364             1.19E+08
  Tax share                         364             0.0002803
  Income                            364        26,629.33
  Grant                             364          1687.017
  Population                        364       167,195.9
  Density                           364           337.9349
  Spending concentration            364             0.572236

Unified powers sample
  Expenditure                      1879             4.62E+07
  Tax share                        1879             0.0005841
  Income                           1879        23,400.31
  Grant                            1879          1834.1
  Population                       1879           109.3133
  Density                          1879           109.3133
  Spending concentration           1879             0.5097718

Variable                        Standard Deviation

Elected executive sample
  Expenditure                         1.42E+08
  Tax share                           0.0014191
  Income                           6005.271
  Grant                            5952.961
  Population                    203,126
  Density                           438.4272
  Spending concentration              0.162541

Professional executive sample
  Expenditure                         6.89E+08
  Tax share                           0.0007822
  Income                           7032.851
  Grant                            8177.458
  Population                    629,742.5
  Density                           558.4959
  Spending concentration              0.1842306

Separation of powers sample
  Expenditure                         2.55E+08
  Tax share                           0.0006214
  Income                           8135.804
  Grant                            3401.788
  Population                    368,737.8
  Density                           881.1246
  Spending concentration              0.1789645

Unified powers sample
  Expenditure                         2.82E+08
  Tax share                           0.0014562
  Income                           5743.013
  Grant                            6700.843
  Population                        303.3917
  Density                           303.3917
  Spending concentration              0.1617992

Table 4. Estimates of Fiscal Illusion Model for Different Government
Forms

                         Elected Administration

                       Without State    With State
Elasticity Estimate       Dummies         Dummies

Tax price                -0.4900 **      -0.5100 **
                        -35.06#         -42.68#
Income                    0.8811 **       0.7963 **
                         16.02#          14.80#
Grant                     0.3599 **       0.3522 **
                         55.43#          45.61#
Population                0.4656 **       0.4376 **
                         24.81#          22.92#
Density                  -0.1109 **      -0.0865 **
                         -7.18#          -5.20#
Spending                  0.8872 **       0.9648 **
  concentration          23.97#          24.52
Fiscal illusion           0.3200 **       0.2700 **
  ([phi] = 0 perfect     62.82#          85.68#
    information)
  ([phi] = 1 perfect    -29.56#         -31.69#
    illusion)
SSE                     394.5645        320.6143
Sample size            1969            1969

                       Professional Administration

                       Without State    With State
Elasticity Estimate       Dummies         Dummies

Tax price               -0.5100 **      -0.5400 **
                       -17.56#         -15.70#
Income                   0.2395 *        0.2409
                         2.02#           1.88#
Grant                    0.3584 **       0.2921 **
                        21.09#           9.68#
Population               0.4474 **       0.4425 **
                        15.79#          12.80#
Density                  0.0262          0.0124
                         0.88#           0.35#
Spending                 0.4474 **       0.4338 **
  concentration          6.55#           4.50#
Fiscal illusion          0.2300 **       0.2000 **
  ([phi] = 0 perfect    61.37#          51.52#
    information)
  ([phi] = 1 perfect   -18.33#         -19.05#
    illusion)
SSE                     28.0092         23.9428
Sample size            274             274

                          Separation of Powers

                       Without State    With State
Elasticity Estimate       Dummies         Dummies

Tax price               -0.5000 **      -0.5300 **
                       -18.27#         -17.99#
Income                   0.7387 **       0.6651 **
                         6.62#           6.08#
Grant                    0.3992 **       0.3796 **
                        21.49#          17.70#
Population               0.4715 **       0.4591 **
                        12.66#          11.85#
Density                 -0.0744 *       -0.0736 *
                        -2.24#          -2.13#
Spending                 0.8226 **       0.9417 **
  concentration         10.71#          10.66#
Fiscal illusion          0.2800 **       0.2300 **
  ([phi] = 0 perfect    42.66#          48.71#
    information)
  ([phi] = 1 perfect   -16.59#         -14.55#
    illusion)
SSE                     58.1166         44.0498
Sample size            364             364

                            United Powers

                       Without State    With State
Elasticity Estimate       Dummies         Dummies

Tax price                -0.5000 **       -0.5100 **
                        -35.28#          -36.15#
Income                    0.8543 **        0.7811 **
                         14.86#           13.79#
Grant                     0.3588 **        0.3486 **
                         54.67#           43.73#
Population                0.4651 **        0.4599 **
                         25.16#           24.70#
Density                  -0.1059 **       -0.0900 **
                         -6.83#           -5.40#
Spending                  0.8162 **        0.8782 **
  concentration          21.60#           21.86#
Fiscal illusion           0.3200 **        0.2800 **
  ([phi] = 0 perfect     90.45#           71.73#
    information)
  ([phi] = 1 perfect    -42.56#          -27.89#
    illusion)
SSE                     375.3276         313.9411
Sample size            1879             1879

All models include constant term. t-statistics in bold.

* Significant at 5% level.

** Significant at 1% level.

Note: t-statistics in bold indicated with #.

Table 5. Predicted Expenditures at Full Sample Means

Organization Characteristic   Without State Dummies   With State Dummies

Elected administration             14,902,718             14,012,817
Professional administration        19,981,161             15,506,956
Separation of powers               17,011,418             16,429,783
Unified powers                     14,880,950             14,989,481