Sorting Through the Determinants of Local Government Competition

Abstract

Competition among local governments for business investment and residents is a key feature of metropolitan governance scholarship. Despite the excellent work exploring interjurisdictional competition, the conceptualization and operationalization of competition still lack the necessary complexity to fully capture the determinants of competition. In reality, the degree of competition between local governments is a multidimensional concept. How do the different dimensions of competition impact a city’s own-source revenue yield? Using a Spatial Durbin Model (SDM) to analyze a sample of 2,299 U.S. cities, this study finds that household income differentiation and manufacturing differentiation are important in a city’s revenue yield, and both types of differentiation limit head-to-head competition among local governments. In addition, the results indicate that entry barriers and collaboration affect a city’s revenue yields, while the number of cities in a metropolitan statistical area (MSA) does not influence those collections.

Keywords

interjurisdictional competition polycentric government product differentiation local market for public goods

Introduction

Local governments do not operate in a vacuum. Instead, they are part of a complex polycentric system of governments where politically autonomous and self-ruled cities collaborate, outsource, and compete with one another (Ostrom, Tiebout, & Warren, 1961). Scholars have debated about the efficacy of polycentricism—especially as it pertains limiting government size. On one hand, proponents of polycentricism argue that multiple local governments competing over taxable wealth (i.e., interjurisdictional competition) increase local government sensitivity to citizen preferences for public goods (Ostrom et al., 1961; Tiebout, 1956) and prevent revenue maximizing bureaucrats from overtaxing (Brennan & Buchanan, 1980; Buchanan, 1971; Craw, 2008; Schneider, 1989).

On the other hand, critics claim that the theoretical assumptions made by polycentrists are unrealistic (John, Dowding, & Biggs, 1995; Lyons & Lowery, 1989; Orbell & Uno, 1972), that total tax liability actually increases with an increase in the number of local governments (Berry, 2008; Hendrick, Jimenez, & Lal, 2011), and that large non-overlapping governments are cheaper, more democratic, and less wasteful than fragmented, polycentric forms of government (Carr & Feiock, 2004).

Central to this debate is whether interjurisdictional competition effectively limits the monopoly power of city governments. Proponents of polycentricism argue that a local market for public goods exists where citizens move to a city that best matches their preferences for public goods (Schneider, 1989; Tiebout, 1956). Interjurisdictional competition for taxable wealth minimizes the ability of local governments to exercise monopolistic tendencies, particularly their ability to unilaterally increase expenditures and revenues (Brennan & Buchanan, 1980; Craw, 2008).

While this theory is highly intuitive and appealing from a policy standpoint, the empirical literature both supports (Carr & Feiock, 2004; Craw, 2008; Forbes & Zampelli, 1989; Oates, 1985; Schneider, 1989; Stephens & Wikstrom, 2000; Zax, 1989) and refutes (Berry, 2008; Epple & Zelenitz, 1981; Hendrick et al., 2011; Lowery & Lyons, 1989) this perspective, calling into question the efficacy of interjurisdictional competition and perhaps the existence of a local market for public goods (Lowery & Lyons, 1989). Excellent work has been done examining the impact of interjurisdictional competition on government size (see Schneider, 1989 and Craw, 2008), but the conceptualization and operationalization of competition lack the necessary complexity to fully capture the determinants of competition, which partially explains the divergent findings.

The degree of competition in a market—the market’s structure—is a function of more than just the number of competitors. In reality, it depends on the degree of product differentiation, barriers to entry, potential for collusion, and each firm’s market share. How do the different dimensions of competition affect a city’s own-source revenue yield—the amount of revenue raised directly by the city? This study (a) explicates the dimensions of interjurisdictional competition that limit the size of government, (b) provides a framework for evaluating the competitiveness of a local market for public goods, and (c) explores the spatial context of interjurisdictional competition. Particular attention is given to the degree to which a city’s consumer base¹ is differentiated from that of neighboring cities.

By exploring the spatial context and multidimensionality of competition in a local market for public goods, an important gap in our understanding of interjurisdictional competition is addressed. This study provides a more complete answer to how interjurisdictional competition and tax base differentiation affect a city’s revenue yield. By thoroughly clarifying the different dimensions shaping the structure of a local market for public goods, new insights into the potential benefits and limits of using competition as a policy tool to promote local government efficiency are acquired. Using census data from 2007, this study’s analysis reveals that many previously unaccounted determinants of competition and some types of tax base differentiation play an important role in limiting a city’s revenue yield.

Interjurisdictional Competition

Interjurisdictional competition prevents city governments from gaining a monopoly and maximizing revenue (Brennan & Buchanan, 1980). Local governments that fear the migration of residents to neighboring jurisdictions if property taxes are too high are less inclined to raise property tax rates. Yet, imperfect competition allows local governments to set tax rates at a higher level than would be tolerated in a perfectly competitive marketplace (Hindriks & Myles, 2006). For example, cities with unique natural amenities—such as beaches, lakes, or mountains—can raise tax rates with less trepidation than other local governments because the number of proximate cities with the same natural amenities is few, limiting competition. In a private market, there are many confounding factors that perpetuate imperfect competition—few producers, few consumers, high entry costs, low consumer information about alternatives, and transaction costs to name a few.

While the determinants of imperfect competition in a private market are well known, few attempts have been made to synthesize these into a comprehensive theoretical framework explaining the structure of local markets for public goods. This study fills that gap by establishing a theoretical framework to evaluate the market structure for locally produced public goods and the effects of the market’s structure on revenue yield.

A city’s revenue yield—the amount of revenue raised directly by a city (i.e., own-source revenue) within a year—is modeled as a function of consumer base differentiation, contestability, concentration, collusion, consumer base diversification, control variables, the spatial lag of the amount of revenue collected, and the spatial lag of all the independent variables.

Y_{i} = α + π_{i} + Ω_{i} + ω_{i} + ψ_{i} + Φ_{i} + ξ_{i} + δ_{i} + W_{i} + ρ W Y_{i} + W π_{i} + W Ω_{i} + W ω_{i} + W ψ_{i} + W Φ_{i} + W ξ_{i} + W δ_{i} + ε .

Consumer base differentiation ( $π_{i}$ ) refers to the degree of similarity between a city’s consumer base and the consumer base of the metropolitan statistical area (MSA) in which the city is located. Contestability ( $Ω_{i}$ ), concentration ( $ψ_{i}$ ) and collusion ( $Φ_{i}$ ) are dimensions of a market that determine the degree of price competition. Unlike consumer base differentiation, consumer base diversification $ξ_{i}$ refers to the internal diversity of a city’s economy in terms of both its resident and business bases. Control variables $δ_{i}$ include the various sociodemographic variables that affect a city’s revenue yield and a citizen’s willingness to pay for services. The spatial lag of the dependent $ρ W Y_{i}$ and independent variables $W π_{i}, W Ω_{i}, W ω_{i}, W ψ_{i}, W Φ_{i}, W ξ_{i}, and W δ_{i}$ captures the metropolitan context of these concepts. The following section elaborates on each of these concepts.

Product Differentiation

Product differentiation theory explains how private firms compete for a market share by making their product a variant of their competitors’ products. When this theory is applied to cities, it can explain how local governments capture a share of the regional tax base. As a production strategy, product differentiation is employed by private firms to maximize profits. Firms produce products that have varying degrees of similarity to those produced by their competitors. The degree to which differentiated products act as perfect substitutes determines the degree of differentiation (Beath & Katsoulacos, 1991). The closer the products are to becoming perfect substitutes, the more undifferentiated the products.

However, as a firm produces a product that differs increasingly from that of its competitors, it is producing a differentiated product. Consumers in a differentiated product market choose among competing producers based on their individual preferences rather than solely on price differences. Differentiated products are not perfect substitutes for each other although they do satisfy similar overarching needs.

Producing differentiated or undifferentiated products is driven by profit maximizing motives, and affects a firm’s price competition and market share. Price competition refers to the elasticity of demand for a firm’s product when a competing firm lowers or raises the price of its product. Undifferentiated product markets result in higher levels of price competition. Similarities in characteristics and quality allow an informed consumer to easily substitute one undifferentiated good for another. Thus, in an undifferentiated product market, customers make their purchase based on the lowest price.

In a differentiated product market, firms have limited monopoly power over a subset of consumers, resulting in imperfect competition. Firms are no longer competing with other firms over a uniform preference among consumers. Instead, these firms are trying to capture a niche market. The more differentiated a product, the less sensitive a firm is to price changes by competing firms, and conversely the less impact one’s firm has on competing firms. By differentiating their product, a firm lowers the amount of price competition it faces in the market and increases the amount of control it has over the price of their product.

Applying these principles to a public goods market provides a theoretical framework for understanding interjurisdictional competition. A public goods market is composed of private firms (a business base) and households (a resident base). Cities compete to attract private firms and households to locate within their city (Schneider, 1989). Both these bases form the foundation for a city’s revenue yield. Businesses provide employment and sell goods that can be taxed while residents pay a variety of taxes, including property taxes, utility fees, and sales taxes on their purchases. By successfully attracting residential and business bases to locate within their jurisdiction, local governments can then tax this base and subsequently increase revenues.

According to product differentiation theory, a city will behave increasingly like a monopoly when its consumer base is highly differentiated compared with the geographic region in which it is located. A differentiated business and resident base indicates that a city is providing a bundle of public goods that attracts a niche consumer base but may not necessarily appeal to the average consumer in a larger geographic region. This niche, or differentiated resident and business base, makes the city’s consumer base less sensitive to competition and allows the city to raise taxes higher than if the city was in a more undifferentiated local market for public goods.

Conversely, when a city has an undifferentiated business and resident base, it is more sensitive to competition. Cities with similar consumer bases must compete to attract and retain their resident and business base to a greater degree than those with niche business and resident bases. When multiple cities have equally appealing bundles of public goods, then tax rates will be weighed more heavily by consumers making location decisions. In short, if a city has differentiated resident and business bases, the result will be lower price competition and higher own-source revenues. The following hypotheses are generated from this theoretical proposition:

Hypothesis 1: As the resident base differentiation increases, own-source revenue will increase.

Hypothesis 2: As the business base differentiation increases, own-source revenue will increase.

Determinants of Competition

In addition to differentiation, a variety of factors affect the structure of a market and the subsequent degree of competition in a local market for public goods. In a private market, competition is a function of three dimensions: contestability, concentration, and collusion (Hindriks & Myles, 2006). The logic behind these three dimensions can be extended to a local market for public goods.

First, contestability refers to the ability of rival producers to freely enter a market. Entry barriers play a key role in the degree of competition that exists in a local market for public goods. Entry barriers reduce interjurisdictional competition, enabling a city to increase revenue yields above what otherwise would be possible.

In a local public goods market, entry barriers can be either economic or political. Economic barriers are those that directly affect the cost of competition over a tax base. Competition costs are associated with financing incentives used to entice businesses and citizens to locate within a city’s border. As competition becomes fiercer, the costs associated with entering a market increase and naturally create a barrier for local government competition. Entry barriers allow cities to exercise monopoly power over their business and residential bases.

Political barriers prevent a local government from effectively entering a market because of the electoral costs to politicians. Political barriers are strong deterrents for local government politicians who want to be reelected (Mayhew, 1974). For example, sex businesses or polluters require that local government officials sacrifice votes in the next election to allow the business to locate within their jurisdiction (Fischel, 2001; Sharp, 2004). Sex businesses and manufacturing plants that give off pollution are profitable and increase local revenue, but they are not always pursued because of the negative electoral consequences caused by externalities from these businesses. Contestability in a local market for public goods leads to the following hypotheses:

Hypothesis 3: As economic barriers to entry increase, own-source revenue will increase.

Hypothesis 4: As political barriers to entry increase, own-source revenue will increase.

Second, concentration refers to the number of competitors in a market. Multiple producers in a market increase price competition, while few producers result in less competition. In a public goods market, the effects of concentration should not differ from those of a private market. As the number of cities in an MSA increases, the public goods market has more producers giving residents and businesses more choices, thereby increasing price competition among cities.

Besides the raw number of competitors in a market, an important element of concentration is the relative market share of each competitor or the dispersion of market share. If the market share of each competitor is nearly the same, then market concentration is considered low and competition is considered high. However, if one or a small minority of firms has captured most of the market, then the concentration of the market is low even if there are a lot of firms in the market. When one or a few firms have captured most of the market, they are less responsive to price competition from competing firms because they control most of the market.

In the case of a public goods market, dispersion of market share does not directly equate to that of a private market. Instead, relative revenue shares in a MSA can be used to capture a city’s equivalent of market share because a city that controls most of the revenue can set tax rates in the region. The following hypothesis follows:

Hypothesis 5: As concentration increases, own-source revenue will increase.

Third, collusion measures the degree to which firms set prices independently. In a public goods market, collusion is the degree to which cities set tax rates independently. Collusion can either be explicit or tacit. Explicit collusion occurs when firms purposefully agree to work together to set prices higher than would be in a perfectly competitive market. This type of collusion between cities is unlikely to occur.

By contrast, tacit collusion occurs when firms avoid price undercutting strategies without explicitly agreeing to do so. This strategy minimizes short-term risk and only requires knowledge of a neighboring city’s tax rate. In a public goods market, exit threats can be minimized and revenue maximized in a MSA when cities adopt tax rates similar to their neighbor’s tax rates. According to the theory of yardstick competition, voters use the decisions and policy outcomes of neighboring local governments to evaluate the performance of their local government (Besley & Case, 1992; Brueckner, 1998, 2003; Buettner, 2001; Heyndels & Vuchelen, 1998; Ladd, 1992; Shleifer, 1985). By pursuing a taxing level similar to the regional norm, the risk of outmigration and voter dissatisfaction are decreased. Tax rate information about neighboring cities is widely available and something that city officials watch closely. Collusion in a local market for public goods leads to the following testable hypotheses:

Hypothesis 6: As tacit collusion increases, own-source revenue will increase.

Spatial Context and the Proposition of a Rippling Monopoly

In metropolitan areas, interjurisdictional competition cannot be separated from its spatial context (Minkoff, 2012). In a private market, the prices a firm charges for its products are constrained by aggressive competitors seeking to increase their market share by undercutting their competitor’s prices. In a local market for public goods, however, tax rate increases are not as easily prevented by competition. When a city increases its tax revenue by capitalizing on imperfect competition, the threat of outmigration should prevent neighboring cities from subsequently raising their taxes because these neighboring cities do not have the same monopolistic power (Brennan & Buchanan, 1980; Tiebout, 1956).

However, research has shown that an increase in spending or tax rates in one city results in an increase in spending or tax rates of a city’s neighbors (Besley & Case, 1992; Brueckner, 1998, 2003; Buettner, 2001; Heyndels & Vuchelen, 1998; Ladd, 1992; Minkoff, 2012; Shleifer, 1985). Neighboring cities raise their prices with the monopolistic city rather than undercutting the city in question. In other words, as neighboring cities become less sensitive to price competition, the focal² city will become less sensitive to price competition and will increase its own-source revenue.

Two competing spatial processes are theorized. The first argues that the spatial process at work is a competitive process, where the monopolistic behavior of a city’s neighbors leads to increased price competition in the focal city. Price undercutting is an attempt to capture some of the revenue share of the monopolistic neighboring city, and thus serves an economic purpose. In addition, price undercutting when your neighbors have monopolistic characteristics can be used to gain electoral benefits when politicians use their neighbor’s tax exploitation as a yardstick for their lower revenue yields.

The second theory argues that there is a rippling monopoly. A rippling monopoly occurs when one city’s monopolistic behavior ripples out to other surrounding local governments who increase their revenue yield despite the absence of having monopolistic characteristics. A city next to a monopolistic city can shift the political risk of increased revenue yield onto its neighbor through yardstick comparisons. These two opposing viewpoints provide the following competing hypotheses:

Hypothesis 7a: As consumer base differentiation, contestability, or collusion increase in neighboring cities, own-source revenue in the focal city will decrease.

Hypothesis 7b: As consumer base differentiation, contestability, or collusion increase in neighboring cities, own-source revenue in the focal city will increase.

Research Design

Data

To test the hypotheses, U.S. cities in a sample of MSAs³ are utilized as the unit of analysis. Only cities within MSAs are included in the sampled population because cities in an MSA are considered legally independent but economically interdependent, which means that cities in a shared MSA are also sharing a local market for public goods. In addition, cities have the power to set zoning laws shaping the overall development of a location (Fischel, 2001). The data are collected from three sources: the 2007 Census of Government Finance, the 2007 Economic Census, and the 2005-2009 American Community Survey.

All MSAs in 10 purposefully selected states were the subject of analysis.⁴ California, Delaware, Florida, Louisiana, Montana, North Carolina, Oklahoma, South Dakota, Tennessee, and Texas were chosen because they provide a wide range of population sizes, industrial bases, demographics, environmental endowments, and sales tax administration and collection procedures. The resulting sample contains 2,299 cities from 115 MSAs, and can be reviewed in Table 1.

Table 1.

Sample Breakdown.

State	TX	CA	DE	MT	SD	LA	FL	NC	OK	TN	Total
MSA cities	699	432	35	13	50	121	324	273	192	160	2,299
0% of sample		19	2	1	2	5	14	12	8	7

Note. MSA = metropolitan statistical area.

The primary reason for this purposive sample was to control for sales tax administration and collection procedures that limit or enhance interjurisdictional competition over retail businesses—an important element of a city’s business base. States vary in their laws and procedures governing sales tax administration and collection, and these “institutional rules” affect the behavior of actor’s within those institutions (North, 1990). Depending on those state-level policies, a state’s sales tax may encourage interjurisdictional competition. The selected states create a sample of cities split evenly between cities in states with optimum competition for a sales tax base (i.e., California and Texas,) and states whose sales tax administration procedures limit competition (i.e., Delaware, Florida, Louisiana, Montana, North Carolina, Oklahoma, South Dakota, and Tennessee).⁵

Three institutional rules were used to determine optimum competition: (a) All cities in these states collect a local sales tax, (b) state law sets a maximum rate for the local-level sales tax that effectively creates a uniform rate of taxation statewide, and (c) city sales tax revenue is generated from point of sale as opposed to point of residence. These three rules encourage cities to compete over business location, by limiting cross-border issues (i.e., Rules 1 and 2) and internalizing gains in revenue (i.e., Rule 3).

Dependent Variable

This study examines the effects of consumer base differentiation on price competition among cities in metropolitan areas. Price competition is the dependent variable and is operationalized using two measures: total own-source revenue and total own-source revenue per capita. Cities experiencing more intense price competition should collect lower total revenue and lower per capita revenue all else being equal. Own-source revenue includes all revenue generated by a city through taxes, fees, and enterprise activities but excludes revenue from intergovernmental sources. Total own-source revenue is logged to normalize its distribution.

By using total own-source revenues and revenues per capita, a more complete picture of price competition is provided. Total own-source revenues represent and acknowledge revenue yield from the perspective of the city government and is based on what is needed to balance revenues with expenditures. Revenue per capita captures revenue yield from the perspective of individual citizens and is an indication of the average revenue burden incurred by residents of the city. These two measures also provide insight on the nature of interjurisdictional competition—do cities compete against the total revenue yields of other cities or do they compete over individual tax burdens. In short, the effects of interjurisdictional competition on revenue yield are explored on an institutional level (i.e., total revenue) and an individual level (i.e., revenue per capita).

Independent Variables

Consumer base differentiation

The key independent variables are residential base and business base differentiation. Residential base differentiation is operationalized both as racial differentiation and as income differentiation. Racial differentiation is calculated as the distance standard deviation (DSD) between a city’s % non-White population and the MSA % non-White population. This variable gives an approximation of how differentiated the city’s racial composition is from that of the reference region.

The following formula is used to calculate racial differentiation, where i is the city and j is the city’s MSA:

Racial differentiation = \sqrt{{(% non - {White population}_{i} - % non - {White population}_{j})}^{2}} .

Income differentiation is calculated as the DSD of a city’s median household income and the MSA household median income. The following formula is used to calculate racial differentiation, where i is the city and j is the city’s MSA:

Income differentiation = \sqrt{{({Median Household Income}_{i} - {Median Household Income}_{j})}^{2} .}

By accounting for differences in both racial and income composition of a city’s population, two elements that characterize a residential population and determine that population’s preferences are captured (Hendrick & Shi, 2014). Population groups differ in their preferences for public goods and will relocate to cities that match their desired preferences. Consequently, as a city’s racial composition differs from that of its MSA, that city is in limited price competition with neighboring cities in the MSA.

Business base differentiation is operationalized as the DSD of the location quotient (LQ) of retail employment or manufacturing employment in each city. By using the DSD, the LQ of both retail and manufacturing employment accounts for business base differentiation for political reasons (manufacturing agglomeration) and bureaucratic reasons (retail agglomeration).

Retail and manufacturing agglomeration are selected to test the hypotheses because each industry provides distinctly different benefits to city officials. Retail businesses increase in sales tax revenue and are attractive to bureaucrats because bureaucrats are revenue maximizers (Niskanen, 1975). Politicians are motivated by electoral benefits (Mayhew, 1974), and manufacturing businesses increase employment—an electorally beneficial outcome.

An LQ is an index that measures the agglomeration of an industry in a city relative to the regional concentration of that industry (Leigh & Blakely, 2013). Specifically, an LQ uses industry employment data to calculate the ratio between the percent of an industry’s employment in a city and the percent of an industry’s employment in an MSA.

The LQ of industry i in geographic area j is calculated using the following formula:

L Q_{i j} = \frac{(X_{i j} / X_{* j})}{(X_{i *} / X_{* *})},

where $X_{i j}$ is the total number of full-time employees in industry i in geographic area j; $X_{* j}$ is the total number of full-time employees in all industries * in geographic region j; $X_{i *}$ is the total number of full-time employees in industry i in the smallest geographic area that contains all geographic areas in the sample *; $X_{* *}$ is the total number of full-time employees in all industries * in the smallest geographic area that contains all geographic areas in the sample *. When a city’s LQ is greater than 1, then industry i is geographically clustered in area j.

The DSD LQ of industry i in geographic area j is calculated using the following formula:

DSD LQ = \sqrt{(L Q_{i j} - 1^{2})} .

As 1 is the MSA agglomeration mean for industry i, this measure gives a measure of the actual differentiation of a city’s retail and manufacturing employment agglomeration compared with the MSA average. The DSD LQ accounts for the impact of actual differentiation away from the regional agglomeration mean rather than simply examining agglomeration’s impact on revenue yield.

Contestability

A city’s contestability is operationalized as its reliance on sales tax and market share as measures of economic barriers to entry. A city’s status as a principal city is used to measure political barriers to entry. Competition costs represent a real entry barrier to many local governments. As the costs associated with pursuing a certain consumer base increase, the relative benefits decrease.

Competition costs that represent entry barriers are operationalized as the percentage of a city’s own-source revenue acquired from sales tax and the spatially lagged percentage of a city’s own-source revenue acquired from sales tax. Cities that rely more heavily on these revenue sources are more likely to compete over businesses that allow local governments to exploit exported tax revenue (Feiock, Steinacker, & Park, 2009; Hendrick & Shi, 2014). Conversely, cities that do not rely on sales tax revenue are less likely to pay the entry costs associated with pursuing a retail sales tax base. Spatially lagging a city’s reliance on sales tax provides a measure of the average sales tax reliance of its neighbors, which gives another approximation of competition costs. Thus, as a city’s neighbors increasingly depend on sales tax revenue, the entry costs associated with pursuing retail business increase.

A city’s market share also creates barriers to entry. This variable is operationalized as a city’s own-source revenue divided by the sum of all own-source revenues from cities in that same MSA. The larger a city’s revenue yield relative to the metropolitan tax base, the more monopolistic power possessed by the city because a change in the way that city operates affects a greater portion of the market. Cities with a larger MSA market share are likely to have more control over tax rates. Thus, these cities will be less sensitive to tax competition. In addition, being next to a city with large market shares allows neighboring cities to also raise their tax rates without fear of being undercut. In short, cities with a larger market share are less sensitive to tax competition, which in turn allows neighboring cities to raise their tax rates.

In a local market for public goods, entry into a market is sometimes prevented by political barriers. Principal cities capture one aspect of a political entry barrier. For purposes of this study, principal cities are defined by the U.S. Census Bureau as those at the core of an MSA with the largest population, with a population between 50,000 and 250,000 and with more workers commuting into the city than residents, or a city with a population between 10,000 and 50,000 where at least one third of the population of the largest city and the number of commuting workers exceed the population.

Principal cities face unique challenges to economic development not shared by other cities in the MSA. Often principal cities are where the majority of residents are employed but not where they live. Principal cities have a high-service demand, a weakened tax base because of the large number of commuters, and regional prestige (Pagano & Bowman, 1995). Being next to a principal city presents unique challenges for neighboring cities. Specifically, a small city with a relatively small business base will have difficulty attracting businesses away from a prestigious principal city. The prestige of the principal city is an entry barrier for competing local governments. It shields the principal city and prevents local governments from competing.

Concentration

Concentration is operationalized as both the number of cities in an MSA—otherwise referred to as MSA size—and the Herfindahl–Hirschman index (HHI) of the metropolitan revenue share. Both these variables are MSA fixed effects, meaning they vary across MSAs but not across cities within an MSA.

The size of an MSA is the number of cities in an MSA and captures the number of potential interjurisdictional competitors. As the number of cities increases, the number of competitors increases. HHI of a metropolitan revenue share measures the dispersion of competitors in a market. Unlike the number of cities, the HHI is an index that weights how competitive a region is based on the market share of cities in an MSA. The HHI is calculated using the following formula:

HHI = \sum_{i}^{N} S_{i}^{2},

where i is the share of a city’s revenue in an MSA and N is the number of cities in an MSA. HHI ranges in value from 0 to 1. As the HHI approaches 1, the regional revenue market share is increasingly located in fewer cities. Conversely, as HHI approaches 0, the regional revenue market share is increasingly located in more cities.

Collusion

Tacit collusion is operationalized using various measures of collaboration and the spatial lag of the dependent variable. Cities that collaborate are interrelated and more likely to be aware of the other’s taxing behavior because collaboration provides an opportunity to share information and expertise (Agranoff, 2006).

Two variables are used to account for collaboration: collaboration-service provider and collaboration-service receiver. Local governments collaborate with their neighbors in multiple ways, and this exchange of resources indicates that collaboration exists. If a local government received money from another local government, then collaboration-service provider is coded 1; otherwise, it is coded 0. If a local government paid money to another local government, then collaboration-service receiver is coded 1; otherwise, it is coded 0. These variables give a measure of the service interdependence and increased potential for information sharing among local governments.

Tacit collusion is operationalized as the spatially lagged dependent variable (i.e., total own-source revenue or own-source revenue per capita). Tacit collusion refers to cities setting their tax rates without explicitly discussing the matter. Due to the openness of government revenue yield and tax rates, tacit collusion is easy for local governments to undertake. Simply by comparing neighboring cities’ revenue yields, a city can avoid reducing its own revenue yield as long as the difference falls within an acceptable range. A positive spatial relationship between the dependent variable and the spatially lagged dependent variable would indicate such a process is at work because an increase in revenue in one city results in a revenue increase in neighboring cities.

Consumer base diversification

In addition to base differentiation, contestability, concentration, and collusion, base diversification affects a city’s overall revenue yield. Differentiation measures business base similarities among cities, whereas diversification measures the similarity of a business base within a city.

Where differentiation gives a city a competitive advantage, a city’s residential or business base diversification measures how dependent the city is on a particular base. A city that is diversified does not suffer the same risks as a city reliant on one business or residential base. Specifically, diversification decreases the chance of one base gaining a monopoly power over the city, and it stabilizes revenues by allowing for alternative means of revenue if one base fails.

Residential base diversification is operationalized as the percentage of residents in a city who are non-White. Business base diversification is operationalized using the traditional HHI, but with the two-digit North American Industry Classification System codes of employment by types of businesses in a city. This measure gives a general sense of the industrial composition of a city. The HHI is calculated using the same formula as previously stated. However, i is the share of industry’s employment in a MSA and N is the number of industries. As the HHI approaches 1, a city is less diversified and more reliant on one industry, yet as the HHI approaches 0 the city’s industrial base becomes more evenly distributed between many industries.

Control variables

The control variables for this particular study include the percentage of the population below 18 and above 65 called the vulnerable population, unemployment rate, median year structures were built in a city, median household income, population change between 2000 and 2007, population (logged), and city size in square acres (logged). Variable descriptions are presented in Table 2, and hypothesis and data sources are presented in Table 3.

Table 2.

Variable Description.

Variable	Variable description
Total revenue (in 1,000s)	Total own-source revenue collected by a city logged
Per capita revenue (in 1,000s)	Total own-source revenue collected by a city divided by the city’s population logged
Racial differentiation	The standard deviation of a city’s % non-White population from the MSA % non-White population
Income differentiation	The standard deviation of a city’s median household income from the MSA household income
DSD LQ (manufacturing)	Standard deviation of a city’s manufacturing LQ away from the MSA average
DSD LQ (retail)	Standard deviation of a city’s retail LQ away from the MSA average
Sales tax reliance (%)	The total sales taxes revenue divided by the total own-source revenue collected by a city
Market share	A city’s total own-source revenue divided by the aggregate of all city own-source revenue in an MSA
Principal city	1 if a city is a principal city in an MSA, 0 otherwise
MSA size	Number of cities in the same MSA
MSA revenue HHI	HHI comprised of city revenue market shares
Collaboration receiver	1 if a city showed a local government intergovernmental expenditure, 0 otherwise
Collaboration provider	1 if a city showed a local government intergovernmental revenue, 0 otherwise
Non-White residents (%)	A city’s non-White population divided by the city’s total population
HHI	HHI comprised of North American Industry Classification System two-digit industry employment figures
Vulnerable population (%)	A city’s population 18 years old and younger plus a city’s population 65 years and older divided by a city’s total population
Unemployment rate (%)	The percentage of population who are unemployed in a city between the ages of 20 and 64
Median structure age	The median year that buildings were constructed in a city
Population change (%)	The difference between the 2000 population and 2007 population
City area	City size in square miles logged
Population	Total population in a city logged
Median income	Median household income in a city logged

Note. MSA = metropolitan statistical area; DSD = distance standard deviation; LQ = location quotient; HHI = Herfindahl–Hirschman index.

Table 3.

Data Source and Hypotheses.

Variable	Hypotheses			Data
Variable	H1-H7	H8a	H8b	Source
Total revenue (in 1,000s)				COG
Per capita revenue (in 1,000s)				COG
Racial differentiation	+	−	+	ACS
Income differentiation	+	−	+	ACS
DSD LQ (manufacturing)	+	−	+	EC
DSD LQ (retail)	+	−	+	EC
Sales tax reliance (%)	−	+	−	COG
Market share	+	−	+	COG
Principal city	+	−	+	ACS
MSA size	−			TSF
MSA revenue HHI	−			COG
Collaboration receiver	+	−	+	COG
Collaboration provider	+	−	+	COG
Non-White residents (%)	−			ACS
HHI	−			EC
Vulnerable population (%)	−			ACS
Unemployment rate (%)	−			ACS
Median structure age	+			ACS
Population change (%)	+			CEN and COG
City area	+			ACS
Population	+			COG
Median income	+			ACS

Note. COG = 2007 Census of Governments-Finance; ACS = 2005-2009 American Community Survey; DSD = distance standard deviation; LQ = location quotient; EC = 2007 Economic Census; MSA = metropolitan statistical area; TSF = 2007 Census Tiger Shape File; HHI = Herfindahl–Hirschman index; CEN = 2000 Census.

Missing Data

Anytime a spatial pattern is present in data, complete data are needed, or else the threat of bias increases with every missing observation. Imputation methods allow for the inclusion of spatially lagged data in otherwise incomplete observations (LeSage & Pace, 2004). While the imputed observations are of limited inferential value, imputation does allow for non-imputed values in an observation with missing data to be spatially lagged. Missing data are imputed using an iterative process that utilizes random forests called MissForest (Stekhoven & Buhlmann, 2012). MissForest makes no parametric assumptions, can account for unspecified interactions and non-linear relationships, and outperforms other methods regardless of the number of missing values or the number of variables in a dataset (Richards et al., 2012; Stekhoven & Buhlmann, 2012).

From the dataset, there are a total of 86 missing observations across all of the independent variables, which constitutes 0.12% of the observed independent variables. Because some of the variables come from the economic census, a considerable amount of the observations is missing because the economic census only provides data for cities with a population upto 2,500. There are 1,712 missing values for the manufacturing agglomeration, 1,208 for the retail agglomeration.⁶ The out-of-bag (OOB) error estimates provide a good measure of imputation accuracy (Stekhoven & Buhlmann, 2012), where values close to 0 indicate good performance and values close to 1 indicate bad performance (Stekhoven, 2011). The OOB for the independent variable imputation was .00201356, and the OOB for the dependent variable (estimated at the same time) was 7.08e-08 indicating excellent performance.

Method

Local markets for public goods are, by nature, spatial arrangements. A city’s taxing decisions are made, given a variety of neighboring city characteristics. In spatial analysis, the attributes of a city are analyzed in conjunction with the attributes of its neighbors to account for spatial autocorrelation. Spatial autocorrelation occurs when random variables take on similar or less similar values at observations that are geographically proximate (Legendre, 1993). Because of the uniqueness of spatial analysis and spatial autocorrelation, the specialized terms “focal city” and “spatial lag” are used throughout this article. A focal city refers to all non-spatially lagged outputs from a spatial regression. Focal city is used to identify the internal, non-neighboring, attributes of a city. A spatially lagged variable is the average value of a variable for all neighboring cities.

A spatial weights matrix (SWM) is employed that spatially relates cities to other cities in the sample. The SWM employed is a distance-based inverse weights matrix (ID-SWM). A city’s neighbors are classified as all cities within 100 km of the focal city. Although all cities within 100 km of the focal city are considered neighbors, not all neighbors are weighted equally. An ID-SWM weights neighbors 1 / (distance), which means that as distance from the focal city increases, a neighboring city is weighted less.

A Global Moran’s I test was conducted on the dependent variables, and indicates that there is spatial autocorrelation present in both total revenue and per capita revenue variables at a 99% confidence level.⁷ Moran’s I test statistics are not directly interpretable; instead, they just indicate if there is a spatial relationship.

In addition, the significance of Moran’s I test statistic indicates that spatial autocorrelation is present, and thus assumptions of the ordinary-least-squares (OLS) model are violated. Spatial Durbin Models (SDMs) are employed to account for various specifications of spatial autocorrelation and are chosen for three key reasons. First, the SDM model estimates spatially lagged dependent variables and independent variables by weighting these variables by the previously specified ID-SWM, which allows for efficient and unbiased estimation in the presence of spatial dependence. Spatial error models (SEMs) only include a spatially lagged error term and no spatially lagged dependent or independent variables. Given the statistical significance of the Moran’s I statistic, any model that did not include a spatially lagged dependent variable would be biased. Second, selection of the spatial autoregressive (SAR) model over the SDM model runs the risk of introducing bias into the analysis. SAR models only include spatially lagged dependent variables but do not include spatially lagged independent variables. If the spatial lags of the independent variables are statistically insignificant but, included in the model (a potential outcome for an SDM model), no bias would be introduced into the regression results, but if the spatial lags are statistically significant and not included (a potential outcome for a SAR model), then bias is introduced into the results (LeSage & Pace, 2009). Therefore, using a SDM to estimate the regression models introduces the least bias into the regression model. Third, the SDM models in this article are estimated using maximum-likelihood estimation procedures to avoid inflation of standard errors from the correlation between the dependent variable and its spatial lags (Lesage & Pace, 2009).

Results

Descriptive statistics are presented in Table 4. Models 1 (Total Revenue) and 2 (Per Capita Revenue) in Table 5 present the results of the SDM analyses, and Table 6 presents the goodness-of-fit statistics for Models 1 and 2. Overall, both models have excellent goodness-of-fit statistics. The likelihood ratio (LR) statistic compares each model with a model using the intercept only. Both models have a statistically significant LR statistic at a 99% confidence level, indicating that the independent variables of each model are jointly significant and add predictive power compared with a model with no independent variables.

Table 4.

Descriptive Statistics.

	M	SD	Median	Minimum	Maximum	Range
Total revenue (in 1,000s)	61,990	440,851	4,906	—	16,526,171	16,526,171
Per capita revenue (in 1,000s)	3	55	1	—	2,583	2,583
Racial differentiation	18	14	16	—	73	73
Income differentiation	15,839	19,387	10,483	42	194,541	194,499
DSD LQ (manufacturing)	1.94	1.07	2.20	—	7.24	7.24
DSD LQ (retail)	0.66	0.44	0.68	—	3.48	3.48
Sales tax reliance (%)	13	15	10	—	100	100
Market share	0	0	—	—	1	1
Principal city	0	0	—	—	1	1
MSA size	58	58	34	1	203	202
MSA revenue HHI	0.43	0.25	0.37	0.04	1.00	0.96
Collaboration receiver	0.21	0.41	—	—	1.00	1.00
Collaboration provider	0.41	0.49	—	—	1.00	1.00
Non-White residents (%)	34	26	27	—	100	100
HHI	0.31	0.12	0.32	0.11	0.78	0.68
Vulnerable population (%)	40	7	39	—	100	100
Unemployment rate (%)	7	4	6	—	57	57
Median structure age	1,976	12	1,976	1,940	2,004	64
Population change (%)	23	66	7	(95)	1,179	1,275
City area	39,360,771	112,230,031	11,790,597	42,051	2,265,279,368	2,265,237,000
Population	29,024	119,850	4,238	8	3,849,378	3,849,370
Median income	53,992	27,586	46,936	6,250	250,000	243,750

Note. DSD = distance standard deviation; LQ = location quotient; MSA = metropolitan statistical area; HHI = Herfindahl–Hirschman index.

Table 5.

Model 1 and Model 2.

	Model 1		Model 2
	Total revenue (log)		Per capita revenue (log)
		Spatial lag		Spatial lag
Intercept	22.081		20.767
Racial differentiation	−0.006***	0.026	−0.005***	0.021
Income differentiation (1,000s)	0.006***	−0.021	0.004***	−0.014
DSD LQ (manufacturing)	0.122***	−0.616	0.070*	−0.458
DSD LQ (retail)	0.012	0.406	−0.032	0.346
Sales tax reliance (%)	−0.070	1.500	−0.341**	0.635
Market share	0.652***	9.098***	0.642***	7.332***
Principal city	−0.094	−1.778	−0.047	−1.550
MSA size	0.000		0.001
MSA revenue HHI	−0.175		−0.147
Collaboration receiver	0.140***	0.451	0.099**	0.457
Collaboration provider	0.214***	0.671	0.181***	0.477
Non-White residents (%)	−0.001	0.012	−0.001	0.010
HHI	−4.757***	9.067*	−3.771***	7.842*
Vulnerable population (%)	0.017***	0.052	0.013***	0.034
Unemployment rate (%)	−0.002	−0.115	0.002	−0.088
Median structure age	−0.013***	−0.008	−0.010***	−0.008
Population change (%)	0.140***	0.451	0.109**	0.322
City area (log)	0.061	−0.222	0.108**	−0.256
Population (log)	0.998***	−0.900**	−0.068	−0.045
Median income (log)	−0.281**	1.939**	−0.179*	1.420*
Texas	0.218		0.239
Louisiana	0.547		0.500
Tennessee	0.266		0.270
California	0.778*		0.666*
North Carolina	0.591		0.516
Oklahoma	−0.034		0.080
South Dakota	−0.008		0.079
Florida	0.647		0.515
Montana	0.217		0.101
Rho (spatially lagged DV)	0.737***		0.818***

Note. DSD = distance standard deviation; LQ = location quotient; MSA = metropolitan statistical area; HHI = Herfindahl–Hirschman index; DV = dependent variable.

p <90%. **p <95%. ***p <99%.

Table 6.

Goodness-of-Fit Statistics for Model 1 and Model 2.

	Model 1	Model 2
	Total revenue (log)	Per capita revenue (log)
AIC	6,694	5,596
AIC LM	6,715.9	5,625.9
LM: Residual autocorrelation	6.5447**	6.7918***
LR: Model significance	3,842.3***	673.68***
N	2,299	2,299.0

Note. AIC = Akaike information criterion; AIC for LM only = AIC statistic for a linear version of the model; LM = Residual Autocorrelation-Lagrange Multiplier test for Residual Autocorrelation; LR = Model Significance-Likelihood ratio test comparing each model with an intercept-only model.

p <95%. ***p <99%.

In addition, the Akaike information criterion (AIC) statistics indicate that Models 1 and 2 are good models. AIC statistics are used to compare nested models by measuring how good the model fits the data while penalizing models for overparameterization. For both Models 1 and 2, two AIC statistics are included: an AIC statistic and an AIC LM statistic. The AIC statistic was generated using the SDM model specification, and AIC LM is the linear model without the spatially lagged dependent variable. Both Models 1 and 2 have smaller AIC statistics than their corresponding AIC LM statistics, indicating that the spatially lagged dependent variable is a prudent and appropriate additional measure.

Limited statistical evidence supports Hypothesis 1 that an increase in a city’s residential base differentiation results in an increase in own-source revenue. In both Models 1 and 2, racial differentiation is statistically significant at a 99% confidence level. But it is negatively associated with total revenue and per capita revenue, which is opposite to the predicted direction. As a city’s racial composition becomes increasingly similar to the MSA’s racial composition, both total revenue and per capita revenue increase.

One possible explanation comes from the resurgent ethnicity framework for racial sorting in metropolitan areas. According to the resurgent ethnicity framework, higher socioeconomic households locate adjacent to culturally dissimilar clusters within a region because cultural clustering creates a psychological barrier for non-members of that culture (Chung & Brown, 2007). As a city’s racial composition becomes increasingly differentiated from the region, it is less likely to attract higher socioeconomic individuals.

While the results for racial differentiation do not support Hypothesis 1, the estimates for income differentiation do support the hypothesis. Income differentiation is statistically significant at a 99% confidence level and is positively associated with both total revenue and per capita revenue. These findings provide strong support for Hypothesis 1 and are consistent with previous evidence that cities use zoning policies and development plans to target higher-income residents (Evans, 2008).

In addition, Wagner’s Law holds that public good preferences are income elastic, meaning that as household income increases their demand for public goods increases at an even greater rate (Wagner & Weber, 1977). Thus, the statistical significance of income differentiation provides strong support for the impact of residential base differentiation. While the unanticipated direction for racial differentiation does not necessarily work against the general theory or residential base differentiation, it implies that cultural clustering counteracts the monopoly power provided through racial differentiation.

Hypothesis 2 states that an increase in a city’s business base differentiation results in an increase in total revenue and per capita revenue. Again, both Models 1 and 2 produce strikingly similar results. Manufacturing differentiation is statistically significant (at a 99% confidence level in Model 1 and a 90% confidence level in Model 2), while retail differentiation is statistically insignificant at a 90% confidence level. Furthermore, manufacturing differentiation is positively associated with total revenue and per capita revenue, which corresponds to the predicted direction of the relationship. These findings indicate that manufacturing, but not retail, business base differentiation affects revenue yield in a city.

In general, partial support is provided for the impact of differentiation on revenue yield in an MSA. Specifically, income differentiation and manufacturing differentiation function in a fashion predicted by product differentiation theory. However, racial and retail differentiation operate differently than expected. These findings paint a clear picture that residential and business base differentiation do limit direct revenue competition among local governments but only in specific ways.

The results of Models 1 and 2 corroborate Hypothesis 3 that economic entry barriers result in greater total revenues and per capita revenues. Market share is statistically significant at a 99% confidence interval in both models and is positively associated with total revenue and per capita revenue. A city with a larger market share controls prices and can effectively price other cities out of a certain market. Market share is robust across Models 1 and 2.

The analysis provides mixed evidence regarding the effects of competition costs (i.e., sales tax reliance) on revenue yield. Specifically, sales tax reliance is statistically insignificant at a 90% confidence level in Model 1 but is statistically significant (at a 95% confidence level) in the predicted direction in Model 2. Given the lack of similarity in statistical significance between both Models 1 and 2, it can be inferred that sales tax reliance does little to affect a city’s overall revenue yield but does create an entry barrier from a per capita perspective. In other words, competition costs affect revenue yield in terms of the relative burden placed on its resident base. A city will be more inclined to pay competition costs when the relative burden on citizens is low, but it becomes a greater barrier as the relative burden (i.e., per capita revenue) increases. In short, competition costs are a barrier insofar as competition costs increase the relative burden on citizens.

Both Hypothesis 4—own-source revenue will increase as political barriers increase—and Hypothesis 5—own-source revenue will increase as concentration measures increase—are not supported by the analysis. Hypothesis 4 was tested using a city’s status as a principal city and the variable principal city was statistically insignificant at a 90% confidence level in both models. Although principal cities are associated with greater regional prestige, the results of this study indicate that this prestige does not serve as a political entry barrier that prevents interjurisdictional competition.

Hypothesis 5 was tested using two measures of concentration: MSA size and MSA revenue share. The concentration variables were statistically insignificant at a 90% confidence level in both models. This finding indicates that the number of competitors and the regional distribution of revenue do not affect a city’s revenue yield. Competition in an MSA is often thought of in terms of the number of competitors; however, this study shows that, once other dimensions of competition are included, the number of competitors does not affect revenue yield in a city.

There is strong support corroborating the important role that collusion plays in a public goods market place (i.e., Hypothesis 6) because all measures of tacit collusion are statistically significant in the predicted direction. Both the collaboration measures—collaboration receiver, collaboration provider—are statistically significant across both models and positively associated with total revenue and per capita revenue. As a city pays another local government or receives funds from another local government, total revenue in that city increases. These findings support Hypothesis 6 that collusion results in increases in revenue.

In both Models 1 and 2, the spatially lagged dependent variables are statistically significant at a 99% confidence level and are in the predicted direction. This finding corroborates the idea that local governments use their neighbor’s as a yardstick measure for revenue yield. Overall, there is strong and robust support for Hypothesis 6, which provides strong evidence that collusion plays a key role in a city’s total revenue yield and per capita revenue yield.

In general, the spatially lagged independent variables are statistically insignificant at a 90% confidence level, providing no support for either Hypothesis 7a or Hypothesis 7b. The lone exception is the spatial lag of market share, which is statistically significant in both models at a 99% confidence level and is positively associated with total revenue and per capita revenue. These results support Hypothesis 7b that there is a rippling monopoly effect when a city’s neighbors have a relatively large market share. When a city’s neighbors have a large market share, then the focal city can increase tax rates with little impact on revenue yield. Subsequently, a city will find that price undercutting strategies will not affect their business or resident location substantially, and will opt to raise their revenue yield efforts knowing that outmigration will not lead to lower tax rates for businesses and residents.

This proposition is also supported by the statistical insignificance of the concentration variables MSA size and MSA revenue share. The threat of price undercutting from competitors appears to be low in a local market for public goods. In short, limited evidence supports the theoretical existence of a rippling monopoly in specialized circumstances, but no evidence supports price undercutting strategies articulated in Hypothesis 7a.

Turning to the control variables, the results suggest that business base diversification operates in the predicted fashion. First, in both models, HHI is statistically significant at a 99% confidence level and is negatively associated with total revenue and per capita revenue. As cities become more dependent on one industry, they collect less revenue because they must increasingly bolster that industry and have limited alternative business base options.

Second, the spatial lag of HHI is statistically significant at a 90% confidence level and is positively associated with total revenue and per capita revenue. As a city’s neighbors become more dependent on one industry, the focal city collects more revenue because that city’s neighbors are focused on securing a limited business base, which leaves more industries with fewer suitors. Thus, spatially lagged industry diversification will lead to increased competition because businesses will have many suitors when no one city specializes and all cities welcome many different industry bases.

Unlike business base diversification, residential base diversification is not supported by the analysis. Both percent non-White residents and the spatial lag of the variable percent non-White residents are statistically insignificant across all models.

Overall, the control variables operate in the predicted fashion. The variables median structure age, population change, and population are statistically significant at a 99% confidence level in the predicted direction. In addition, city area is statistically insignificant in Model 1 but statistically significant at a 95% confidence level in Model 2, and is positively associated with per capita revenue. Population is statistically insignificant in Model 1 at a 99% confidence level, but it is not statistically significant in Model 2. Both city area and population are associated with per capita revenue and total revenue in the directions predicted from the economic distress perspective.

Some of the variables act in unexpected ways. Vulnerable population and median household income are statistically significant in the opposite direction as predicted, while the unemployment rate is statistically insignificant at a 90% confidence level in both models. The study finds that as a city’s population is increasingly made up of a vulnerable population, total revenue and per capita revenue increase. This finding is not necessarily unintuitive.

The assumption was that a population comprised of vulnerable individuals would have a weaker tax base. However, a large vulnerable population is more likely to demand more services, which requires more revenue. So rather than the variable vulnerable population capturing a supply side variable, the direction of the variable indicates that it is a demand side variable. Both models also indicate that median household income is negatively associated with revenue yield. As a city’s median household income goes up, total revenue and per capita revenue decrease. Like the vulnerable population variable, it is a distinct possibility that median household income is not a supply side variable but a demand side variable. In other words, as median household income increases, the need for public services decreases.

Furthermore, the spatially lagged control variables are, in general, statistically insignificant at a 90% confidence level. However, the spatial lag of median household income is statistically significant in Model 1 (95% confidence level) and Model 2 (90% confidence level), and the spatial lag of population is statistically significant at a 95% confidence level in Model 1 but not in Model 2.

Conclusion

The purpose of this study was to explore the various dimensions of competition in a local market for public goods, and how these dimensions affect a city’s own-source revenue yield. The analyses paint a vivid picture of the dynamics of the local market for public goods. First, resident base differentiation is supported from an income differentiation perspective but not a racial perspective. This finding suggests that citizen preferences for public goods are differentiated along income. However, racial sorting actually creates cultural entry barriers for households of different cultures, which limits a city’s revenue yield.

Second, business base differentiation occurs as predicted for manufacturing differentiation but not for retail differentiation. While unexpected, this occurrence is not necessarily incongruent with product differentiation theory. Retail location is largely short-term compared with long-term manufacturing placement. Retail firms locate to maximize foot traffic, while manufacturing firms choose locations to minimize supply costs. Comparatively, retail firm placement is short-term, while manufacturing plants are generally long-term. Thus, manufacturing differentiation has long-term regional implications where retail differentiation is less likely to have the same magnitude of regional impact as manufacturing differentiation.

Third, evidence suggests that, of the three competition dimensions, only economic barriers and collusion play an important role in determining the degree of interjurisdictional competition in a region. Of particular importance is the notion that cities with large market shares are able to increase revenue yield and also create rippling monopolies—increases in revenue yield in neighboring cities. Surprisingly, no concentration variables resulted in statistically significant changes in revenue yield. This finding is particularly important because studies that do include measures of interjurisdictional competition often use concentration measures to operationalize competition when, in fact, the analysis indicates that this might be the least effective way to capture the competitiveness of a local market for public goods.

Overall, the local market for public goods is a complex, multidimensional structure that operates differently than private markets. This study serves as a start to a needed line of inquiry into the nature of interjurisdictional competition. From this study, the author feels safe concluding that residential and business base differentiation limit interjurisdictional competition, but only certain types of residential and business base differentiation actually capture differentiated preferences in a consumer base. In addition, collusion appears to play a significant role in how competitive a region is while the number of competitors does not make a statistically significant difference. Furthermore, economic entry barriers and competition costs are important considerations when conceptualizing public good markets.

Future study is needed to determine how residents and businesses differentiate in a substantial way. Furthermore, this study believes that there is an opportunity to study the public goods marketplace from the perspective of various political and economic entry barriers. While this study provides important insight into the local public goods marketplace, it is a starting point.

Footnotes

Declaration of Conflicting Interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) received no financial support for the research, authorship, and/or publication of this article.

Notes

Author Biography

Michael Overton is a research associate with the Center for Public Management at the University of North Texas. He received his doctorate in public administration from the University of North Texas. His research focuses on the intersection of fiscal policy and local governance.

References

Agranoff

(2006). Inside collaborative networks: Ten lessons for public managers. Public Administration Review, 66, 56-65.

Beath

Katsoulacos

(1991). The economic theory of product differentiation. Cambridge, UK: Cambridge University Press.

Berry

(2008). Piling on: Multilevel government and the fiscal common pool. American Journal of Political Science, 52, 802-820.

Besley

Case

(1992). Incumbent behavior: Vote seeking, tax setting and yardstick competition. American Economic Review, 85, 25-45.

Blair

(1995). Local economic development: Analysis and practice. Thousand Oaks, CA: Sage.

Brennan

Buchanan

(1980). The power to tax: Analytic foundations of a fiscal constitution. Cambridge, UK: Cambridge University Press.

Brueckner

(1998). Testing for strategic interaction among local governments: The case of growth controls. Journal of Urban Economics, 44, 438-467.

Brueckner

(2003). Strategic interaction among governments: An overview of empirical studies. International Regional Science Review, 26, 175-188.

Buchanan

(1971). Principles of urban fiscal strategy. Public Choice, 11, 1-16.

10.

Buettner

(2001). Local business taxation and competition for capital: The choice of the tax rate. Regional Science & Urban Economics, 31, 215-245.

11.

Carr

Feiock

(2004). City-county consolidation and its alternatives: Reshaping the local government landscape. New York, NY: M.E. Sharpe.

12.

Chung

Brown

(2007). Racial/ethnic residential sorting in spatial context: Testing the explanatory frameworks. Urban Geography, 28, 312-339.

13.

Craw

(2008). Taming the local leviathan: Institutional and economic constraints on municipal budgets. Urban Affairs Review, 43, 663-690.

14.

Epple

Zelenitz

(1981). The implications of competition among jurisdictions: Does Tiebout need politics? Journal of Political Economy, 89, 1197-1217.

15.

Evans

(2008). Economics and land use planning. Malden, MA: Blackwell.

16.

Feiock

Steinacker

Park

(2009). Institutional collective action and economic development joint ventures. Public Administration Review, 69, 256-270.

17.

Fischel

(2001). The homevoter hypothesis: How home values influence local government taxation, school finance, and land-use policies. Cambridge, MA: Harvard University Press.

18.

Forbes

Zampelli

(1989). Is leviathan a mythical beast? American Economic Review, 79, 568-577.

19.

Hendrick

Jimenez

Lal

(2011). Does local government fragmentation reduce local spending? Urban Affairs Review, 47, 467-510.

20.

Hendrick

Shi

(2014). Macro-level determinants of local government interaction how metropolitan regions in the United States compare. Urban Affairs Review, 51, 414-438.

21.

Heyndels

Vuchelen

(1998). Tax mimicking among Belgian municipalities. National Tax Journal, 51, 89-101.

22.

Hindriks

Myles

(2006). Intermediate public economics. Boston, MA: MIT Press.

23.

John

Dowding

Biggs

(1995). Residential mobility in London: A micro-level test of the behavioural assumptions of the Tiebout model. British Journal of Political Science, 25, 379-397.

24.

Ladd

(1992). Mimicking of local tax burdens among neighboring counties. Public Finance Review, 20, 450-467.

25.

Legendre

(1993). Spatial autocorrelation: Trouble or new paradigm? Ecology, 74, 1659-1673.

26.

Leigh

Blakely

(2013). Planning local economic development: Theory and practice. Thousand Oaks, CA: Sage.

27.

LeSage

Pace

(2004). Models for spatially dependent missing data. Journal of Real Estate Finance and Economics, 29, 233-254.

28.

LeSage

Pace

(2009). Introduction to spatial econometrics. Boca Raton, FL: Chapman & Hall/CRC.

29.

Lowery

Lyons

(1989). The impact of jurisdictional boundaries: An individual-level test of the Tiebout model. Journal of Politics, 51, 73-97.

30.

Lyons

Lowery

(1989). Governmental fragmentation versus consolidation: Five public-choice myths about how to create informed, involved, and happy citizens. Public Administration Review, 49, 533-543.

31.

Mayhew

(1974). Congress: The electoral connection (2nd ed.). New Haven, CT: Yale University Press.

32.

Minkoff

(2012). The proximate polity: Spatial context and political risk in local developmental goods provision. Urban Affairs Review, 48, 354-889.

33.

Niskanen

(1975). Bureaucrats and politicians. Journal of Law & Economics, 18, 617-643.

34.

North

(1990). Institutions, institutional change and economic performance. Cambridge, UK: Cambridge University Press.

35.

Oates

(1985). Searching for leviathan: An empirical study. American Economic Review, 75, 748-757.

36.

Orbell

Uno

(1972). A theory of neighborhood problem solving: Political action vs. residential mobility. American Political Science Review, 66, 471-489.

37.

Ostrom

Tiebout

Warren

(1961). The organization of government in metropolitan areas: A theoretical inquiry. American Political Science Review, 55, 831-842.

38.

Pagano

Bowman

(1995). Cityscapes and capital: The politics of urban development. Baltimore, MD: John Hopkins University Press.

39.

Richards

Starr

Miller

Bloom

Butler

Brink

. . .Crellin-Quick

(2012). Construction of a calibrated probabilistic classification catalog: Application to 50k variable sources in the all-sky automated survey. The Astropysical Journal Supplement Series, 203: 32.

40.

Schneider

(1989). Intermunicipal competition, budget-maximizing bureaucrats, and the level of suburban competition. American Journal of Political Science, 33, 612-628.

41.

Sharp

(2004). Metropolitan structure and the sex business. In Feiock

(Ed.), Metropolitan governance: Conflict, competition, and cooperation (pp. 159-182). Washington, DC: Georgetown University Press.

42.

Shleifer

(1985). A theory of yardstick competition. Rand Journal of Economics, 16, 319-327.

43.

Stekhoven

(2011). Using the MissForest package. Swiss Federal Institute of Technology Zurich: Seminar for Statistics. Retrieved from http://stat.ethz.ch/education/semesters/ss2013/ams/paper/missForest_1.2.pdf

44.

Stekhoven

Buhlmann

(2012). MissForest–non-parametric missing value imputation for mixed-type data. Bioinformatics, 28, 112-118.

45.

Stephens

Wikstrom

(2000). Metropolitan government and governance: Theoretical perspectives, empirical analysis, and the future. New York, NY: Oxford University Press.

46.

Tiebout

(1956). A pure theory of local expenditures. Journal of Political Economy, 64, 416-424.

47.

Wagner

Weber

(1977). Wagner’s law, fiscal institutions, and the growth of government. National Tax Journal, 30, 59-68.

48.

Zax

(1989). Is there a leviathan in your neighborhood? American Economic Review, 79, 560-567.