Abstract
Does giving citizens the power to decide budget policies improve fiscal discipline in the local public sector? This study examines the effects of local initiatives on city budgetary solvency or the ability of city governments to generate revenues to meet their service and financial obligations in a fiscal year. Budgetary imbalance in the public sector has been blamed on self-interested bureaucrats and elected officials who desire budgets that are higher than that preferred by the median voter. The initiative gives citizens the power to directly decide budget issues. Research shows that voters are more fiscally conservative than government officials, which suggests that fiscal discipline will improve if citizens exercise greater control over budgeting. Using data from audited financial reports for midsized and large cities from 2006 to 2012, the empirical analysis indicates that initiative cities have weaker budgetary solvency compared with noninitiative cities.
Introduction
From the United Kingdom’s “Brexit” vote, to the Colombian citizens’ rejection of a peace treaty with a major national insurgency group, and the State of California’s various citizen-initiated taxing and spending constraints, direct citizen participation in policymaking through institutions of direct democracy, such as initiatives and referendums, has led to the adoption of disputed policies, some of which have produced highly controversial outcomes. 1 Not surprisingly, questions have been raised about the wisdom of deciding complex public policy issues through a simple yes-or-no vote by citizens.
The question on the outcomes of direct democracy is particularly important in the United States. Many states and localities in the country have a long and vibrant tradition of direct citizen participation in governance. An important mechanism for citizen participation is the initiative, which is a “process that allows ordinary citizens to propose new laws or constitutional amendments by petition” (Lupia and Matsusaka 2004, p. 465). The state and urban politics literature has long debated the merits of direct democracy as opposed to representative democracy, that is, lawmaking by citizens in contrast to lawmaking by elected representatives (Matsusaka 1992). Some research argues that lawmaking by citizens can help curb abuses by self-interested bureaucrats and elected officials, enhance government responsiveness to the majority of voters, and improve the quality of policy choices by increasing the flow information to and between policymakers and voters (see, among others, Gerber and Phillips 2003; Lupia and Matsusaka 2004; Matsusaka 1995, 2004, 2005a, 2005b, 2005c, 2005d; M. A. Smith 2001; D. Smith and Tolbert 2004; Tolbert, McNeal, and D. Smith 2003). Other studies suggest that lawmaking by representatives helps protect the rights of minorities, is more cost-efficient than direct citizen legislation, and ensures that myopic and uninformed voters do not decide important policy questions (see the discussion in Gamble 1997; Gordon 2009; Lupia and Matsusaka 2004; Primo 2010). Nowhere is this debate more important than in the context of local budgeting. Compared with other levels of government, services provided at the local level have a more direct and palpable impact on the day-to-day lives of citizens (Jimenez 2014b, 2015). Local budget decisions, and whether voters can participate in making those decisions, shape the capacity of cities to provide services that are critical for the health, security, and well-being of citizens.
This study examines the effects of local initiatives on the budgetary solvency of midsized and large cities in the United States. Budgetary solvency is the ability of a government to generate revenues to meet its service and financial obligations in a fiscal year (Groves, Godsey, and Shulman 1981; Hendrick 2011; Jimenez 2009, 2017b, 2017c, forthcoming; Mead 2011). Others might argue that city budgetary solvency is a narrow concern, and that research on direct democracy should focus on policy debates and outcomes, as well as the distribution of government program benefits and services. This argument misses a very important point about the role of budgets in local governance: There will be no debates about policy choices or conflicts over public goods if city governments cannot pay for the costs of implementing policies and delivering services (Jimenez 2017c). Budgets are the lifeblood of cities. The inability of city governments to live within their means has been a recurring issue in urban public finance, and city fiscal difficulties have been laid bare by the Great Recession of 2007–2009 (Jimenez 2013, 2017a, 2017b, 2017c, forthcoming).
Both the theoretical and empirical literatures suggest that initiatives improve city budgetary solvency. Poor fiscal outcomes in the public sector have been blamed on self-interested bureaucrats and elected officials who desire budgets that are higher than what the median voter prefers (Brennan and Buchanan 1980; Niskanen 1972). Some research shows that voters are indeed more fiscally conservative than government officials, preferring lower spending and taxation (Besley and Case 1995; Peltzman 1992), as well as balanced budgets (Alt and Lowry 1994). If citizens exercise greater control over budget decisions through initiatives, fiscal discipline in government is likely to improve, and deficits can be minimized. At the state level, the balance of evidence indicates that initiatives improve fiscal discipline by reducing taxation, spending, or wages.
To test the hypothesis that initiatives improve city budgetary solvency, the study uses data from audited financial reports which are collected for municipalities with a population of 50,000 or more, and from fiscal years 2006–2012. Contrary to the prevailing theory suggesting that initiatives help improve budgetary balance, the results of least squares dummy variable regression and instrumental variable regression indicate that initiatives are associated with the deterioration of city budgetary solvency.
Background and Literature
Direct democracy gives people the power to decide important governance, social policy, and fiscal issues, and includes a broad array of mechanisms ranging from New England–type town meetings, to initiatives and referendums (Lupia and Matsusaka 2004). Town meetings—which are more practicable in smaller communities—require citizens to meet, discuss, and vote on local policies (Matsusaka 2005b), with the policy agenda often set by town officials. Referendums and initiatives do not require face-to-face meetings but allow citizens to participate in elections on a proposed policy, and are, thus, more frequently used in lager jurisdictions. The difference between initiatives and referendums is the origin of the policy being considered. In referendums, citizens vote on laws proposed by legislators, whereas in initiatives, citizens themselves propose new policies by petition (Blume, Müller, and Voigt 2009; Lupia and Matsusaka 2004). In either case, a majority vote by the electorate leads to the adoption of the proposed policy (Matsusaka 2005a, 2005c). This study focuses on the initiative because it is the strongest of the direct democracy mechanisms, giving power to the people not only to approve policies but also to define the policy agenda (Matsusaka 2005c).
Several studies have examined the effects of initiatives on different fiscal outcomes (for an extensive critique of the literature, see Matsusaka 2004). Extant research has largely focused on the United States, specifically on the effects of state-level initiatives on spending and revenues. 2 Focusing on studies that examine the effects of initiatives during the second half of the twentieth century, at least two conclude that initiatives have increased the size of state governments (Marschall and Ruhil 2005; Zax 1989), whereas others report insignificant results (Besley and Case 2003; Camobreco 1998). On balance, more state-level research suggests that initiatives reduce revenues, expenditures, or both (Bails and Tieslau 2000; Besley and Case 2003; Bradbury and Crain 2005; Matsusaka 1995, 2004; Matsusaka and McCarty 2001; Merryfield 2000). Other studies find that initiatives reduce the salaries of high-level elected state officials (Di Tella and Fisman 2004) and public-sector employees (Matsusaka 2009). Some research shows that state initiatives lead to the decentralization of the state and local public sector, and increase reliance on user fees at the expense of taxes (Matsusaka 2004). There are fewer studies of the budgetary effects of local initiatives, and the findings are inconclusive. Some studies find that local initiatives have a weak positive effect on city revenues (Gordon 2009) and spending (Primo 2010), but these effects disappear when different socioeconomic, political, and state-level factors are controlled for.
Expenditures and revenues are important fiscal concerns, but nonetheless, they do not tell us much about fiscal discipline in governments. For example, high spending and revenues do not necessarily signify that a government is fiscally profligate if citizens demand more services and are willing to pay for them. Similarly, low spending and revenues do not mean that a city is fiscally disciplined as these may be signs that a city is not providing an adequate level of services to its residents or does not have the political will to raise taxes to fund needed services.
Rather than analyzing revenue or spending in isolation, both must be considered simultaneously in assessing government fiscal performance. Indeed, the critical question is not whether governments spend or tax more or less but whether they have enough revenues to meet their expenditure obligations. In the public budgeting and finance literature, the interdependence between revenues and expenditures is summarized in the concept called “budgetary solvency.” Specifically, budgetary solvency is the ability of a government to generate revenues to meet its service and financial obligations in a fiscal year (Groves, Godsey, and Shulman 1981; Hendrick 2011; Jimenez 2009, 2017b, 2017c, forthcoming; Mead 2011). Budgetary solvency indicates whether a city has achieved structural balance, which means that recurring revenues match or exceed recurring spending at the end of the fiscal year and over the business cycle (Johnson, Kioko, and Hildreth 2012). For the average citizen, structural budget balance is a widely understood and concrete indicator of a government’s ability to live within its means. It is important to emphasize that budgetary solvency in the public sector is not a binary condition, that is, a city government cannot simply be classified as being solvent or insolvent. It is more appropriate to think of budgetary solvency as a continuous measure of the degree to which a city can meet its service and financial obligations.
The empirical analysis focuses on city budgetary solvency from 2006 to 2012. It covers the period of the Great Recession, which lasted from 2007 to 2009 and caused widespread budget difficulties in the local public sector, and succeeding years of fiscal recovery (Jimenez 2013, 2017a, 2017c, forthcoming). The period covered facilitates analysis of the effects of initiatives on the fiscal performance of city governments during and after a national economic recession. This is important because structural budget balance issues in city governments are not always apparent during times of economic growth when revenues automatically grow (without the need to increase tax rates) and cover any expansion of spending. Economic growth, in other words, can mask how initiatives shape city budgetary solvency. The initiative should matter more during periods of fiscal shock and subsequent fiscal recovery when governments need the support of citizens to alter expenditure or revenue policies.
Theoretical Expectations
City Government Responsiveness to Citizens: Exit and Voice Mechanisms
Do citizens influence city government policies? Research based on the Tiebout model and the median voter theory assumes that city governments are responsive to the fiscal preferences of citizens. These two models emphasize the two contrasting mechanisms of exit or citizens “voting with their feet,” and voice or citizens voting in elections. In the Tiebout model, the threat of fiscal exit ensures government responsiveness. Tiebout (1956) envisioned a quasi-public market in which numerous city governments in a region—much like private firms in a competitive market vie for market share—compete for mobile resident-taxpayers by offering different packages of services and corresponding tax costs (Jimenez 2014a, 2014b; Schneider 1989; Tiebout 1956). In regions with numerous municipalities, residents can vote with their feet and shop for jurisdictions that satisfy their tax and service preferences (Tiebout 1956). Because city governments are dependent on local taxes, the threat of fiscal exit by fiscally desirable and mobile taxpayers limits the capacity of local officials to enact budget policies that diverge from those preferred by local residents (Craw 2008; Dowding, John, and Biggs 1994; Jimenez 2014a, 2014b; Schneider 1989).
The threat of fiscal migration by local taxpayers, however, is not always effective in ensuring government responsiveness 3 because residential mobility is constrained by relocation costs, limited employment opportunities in neighboring cities, and incomplete information about the different tax and service packages offered by competing cities (see Dowding, John, and Biggs 1994). An alternative to exit is the exercise of citizens’ voice, specifically by voting in local elections. As Besley and Case (1995) argue, “In the short run, the ballot box may serve an important function and even in the long run may be a less costly alternative than migration” (p. 26).
Rather than fiscal exit, elections are the main mechanism through which voters control their governments, punishing elected officials who deviate from voters’ preferred policies by booting them out of office, and rewarding those who are more responsive to voters’ demands with reelection. Thus, electoral competition is expected to lead to the adoption of budget policies that reflect the preferences of the majority of voters. This idea is formalized in the median voter theory, which argues that the median voter is the decisive voter in majority-rule elections. The median voter is the voter whose preference lies at the middle of the electorate, so that half of the electorate supports a policy and the other half expresses less support. Because it is the position of the median voter that receives the majority’s support, vote-maximizing politicians respond to the median voter’s preferred fiscal policies, assuming single-dimensional voting and single-peaked preferences (Downs 1957).
There are no assurances, however, that elected officials will be responsive to the will of the majority. First, voters have imperfect information about the choices and actions of their elected representatives, which makes monitoring difficult. This insight is based on the principal-agent model, which sees citizens as principals who delegate budget powers to their agents or the elected representatives. The agents are expected to enact policies that promote the welfare of their principal. An agency problem occurs when agents pursue actions that promote their self-interests but are detrimental to the principal’s interests. This happens because principals face difficulty in gathering information about agents’ choices and actions. As agents develop expertise in budgeting, they gain information advantage over their principal. Beyond the expert knowledge required to understand budget issues, the presentation of budgets as an omnibus bill packed with different revenue and spending provisions adds to citizen confusion about budgetary policies (Blume, Müller, and Voigt 2009; Feld and Kirchgässner 2001; Matsusaka 1995, 2005b). Thus, for voters, gathering information about fiscal issues can be prohibitively costly.
Second, government officials control the budget agenda (Romer and Rosenthal 1979). In a representative democracy, citizens have no direct role in proposing, developing, and adopting budget options. Government officials’ monopoly of budgetary decision making, combined with weak monitoring by citizens, enhances the former’s ability to enact policies that may be contrary to the median voter’s preference (Matsusaka 1995, 2005c). 4
Third, legislative logrolling can lead to the adoption of particularistic policies, which by definition do not benefit the majority (Matsusaka 1995, 2005c). This is based on the fiscal common pool theory, which suggests that representative democracy is susceptible to pork-barrel-type decision-making (Hallerberg, Strauch, and von Hagen 2009; Jimenez 2015; Weingast, Shepsle, and Johnsen 1981). According to this theory, the general tax fund represents a common pool of resources that can be used by any representative to fund particularistic goods, or goods that primarily benefit a specific constituency (Weingast, Shepsle, and Johnsen 1981). Vote-maximizing politicians attempt to bring in more pork for their districts. Residents in a district, in essence, are being subsidized by taxpayers from other districts. The subsidy creates the illusion that the tax cost of particularistic goods is lower than what it actually is, incentivizing residents to demand higher spending. Increased spending is ultimately approved by the legislature through logrolling in which representatives trade votes for each other’s projects (Jimenez 2015). 5
If voting with their feet and voting in regular elections to choose their representatives are insufficient to ensure that citizens can control their city governments, city officials enjoy great leeway in defining budget policies independent of the will of the majority. If government officials were benevolent social planners who enact policies that maximize the welfare of citizens, their outsize role in fiscal policymaking would not be an issue. Scholars from public choice, however, argue that government officials are self-interested actors who maximize spending and taxes for private, rather than public, purposes (Brennan and Buchanan 1980; Niskanen 1972). Niskanen (1972; also see Schneider 1989) points to bureaucrats who desire higher budgets to increase their salaries and gain more nonmonetary benefits from their official positions. Others blame elected officials who overspend for particularistic goods to maximize votes (Besley and Case 1995; Weingast, Shepsle, and Johnsen 1981). Budget maximizing ultimately weakens budgetary solvency of city governments. Governments, much like households and private firms, face a budget constraint (Hallerberg, Strauch, and von Hagen 2009). Taxes cannot increase perpetually to match excessive spending, as private incomes fluctuate or stagnate, and vote-maximizing politicians realize that tax increases have electoral consequences. Because of inherent limits to a city governments’ ability to raise revenues, the higher expenditure preferred by bureaucrats and politicians leads to budgetary imbalance.
Reasserting Citizens’ Voice: Budget Control Through Initiatives
If fiscal exit and representative democracy do not work, giving power to citizens to directly decide budget policy questions through initiatives can help control budget-maximizing behavior of local officials. The literature suggests that the city initiative can push budgetary policies closer to the median voter’s preference through its direct override, indirect threat, issue unbundling, and information effects (see, among others, Feld and Kirchgässner 2001; Frey, Kucher, and Stutzer 2001; Lupia and Matsusaka 2004; Matsusaka 1995, 2004, 2005a, 2005b, 2005c, 2009, 2014; Matsusaka and McCarty 2001).
A straightforward effect of initiatives is to wrestle agenda control away from both bureaucrats and elected officials (Matsusaka 2005c, 2014). The initiative breaks the monopoly power of city officials over budget agenda setting, formulation, and adoption. This occurs either through the direct override effect or indirect threat effect. Direct override means that citizens reject the budget choices of city officials, and instead introduce new proposals and directly decide fiscal policies. Different citizen or interest groups can propose alternative budgets, and the median voter—the decisive voter—can choose a budget that is closest to her ideal point (Lupia and Matsusaka 2004; Matsusaka 2005c).
The indirect threat effect suggests that the mere availability of an initiative, rather than its actual use, is enough to change the budget decisions of city officials. In other words, an initiative need not lead to the actual approval and implementation of a measure to change extant policies (Lupia and Matsusaka 2004; Matsusaka 2005c). The threat of an initiative can change the behavior of city officials, and make them reevaluate their policy positions (Gerber 1998, 1999; Matsusaka 2005c). For example, citizen and interest groups can threaten to propose different budgets that are slightly lower than the median voter’s preferred budget. To be taken seriously by city officials, the external groups need to signal that their proposals can attract majority support, which is possible only when a proposed budget is closer to the median voter’s preferred budget. The citizen and interest groups’ preferred budgets are not acceptable to city officials because they lead to lower spending. To stave off a full election on the external groups’ budget proposals, and preempt adoption of a budget that is contrary to their preferences, city officials propose a budget that is only slightly higher than the median voter’s preferred option. Direct democracy need not result in the exact budget preferred by the median voter but can result in a budget that is close to that preference (Matsusaka 2004), assuming that officials are certain about the median voter’s preferences (Matsusaka and McCarty 2001). Matsusaka (2004) offers evidence that initiatives move budget policy closer to the median voter’s preference. He finds that state initiatives lead to lower state spending, shift in spending from the state to local levels, increase reliance on service charges, and decrease in the share of general tax revenues. Tracking opinion data, he shows that majority of state residents support these fiscal developments.
Information asymmetry and inefficiencies caused by pork-barrel politics can be corrected through the information and issue unbundling effects of initiatives. Initiatives can lower information costs through issue unbundling (Lupia and Matsusaka 2004; Matsusaka 2005c). Rather than focus on the entire budget, initiatives target specific spending or revenue provisions. By unbundling omnibus budget bills, voters can gain a better understanding of the true costs of particularistic goods. Budget unbundling also makes it harder for representatives to push projects that only benefit their district but are subsidized by taxpayers from other districts. This is because legislative logrolling cannot occur if citizens themselves decide spending programs (Feld and Kirchgässner 2001; Lupia and Matsusaka 2004; Matsusaka 2005c).
Beyond budget unbundling, initiatives lower information costs through the activation of interest groups, which shoulder the costs of generating and disseminating information to citizens. Interest groups increase information made available to voters about current city budget issues through campaign activities and greater media exposure. The education effect of initiatives has been documented in the literature (see Gerber and Phillips 2003; D. Smith and Tolbert 2004; Tolbert, McNeal, and D. Smith 2003). Initiatives not only increase information flow to citizens (Gerber and Phillips 2003) but can also improve civic engagement by mobilizing voters to participate in interest group activities, and by increasing voter turnout (M. A. Smith 2001; D. Smith and Tolbert 2004; Tolbert, McNeal, and D. Smith 2003).
By enabling citizens to limit the ability of both bureaucrats and elected officials to increase government budgets beyond that demanded by the median voter, initiatives can help cities achieve structural budget balance. The effect of initiatives on budgetary solvency, however, hinges on an important assumption—that the median voter is more fiscally conservative than city officials. Evidence from the literature suggests that this is the case, at least at the state level. Research shows that voters punish incumbents at the polls for high spending (Peltzman 1992), tax increases (Besley and Case 1995), and budget deficits (Alt and Lowry 1994). If this assumption is correct, it is expected that giving citizens the power to directly decide local budget policies through initiatives improves city budgetary solvency.
Research Method
The study tests the hypothesis on midsized and large cities in the United States, specifically those with a population of 50,000 or more. The target cities play an outsize role in the provision of local public services. As of 2014, the American Community Survey indicates that approximately 117 million people reside in these cities. This figure does not include the millions of people from smaller communities who commute to urbanized jurisdictions and central cities for work, and therefore, also benefit from city-provided public services and infrastructure. The cities covered in the study include all the major cities in the United States such as San Francisco, Los Angeles, Phoenix, Houston, Dallas, San Antonio, Atlanta, New Orleans, Chicago, Washington D.C., Philadelphia, Boston, and New York, among others.
Measuring Budgetary Solvency 6
Any analysis of budgetary solvency starts by identifying the entity for which financial information will be collected and examined. In this study, the entity being examined is the city government. Complicating the financial analysis of city governments, however, is the fund accounting system, which subdivides a city government’s accounting records into different subentities or funds. A fund is an accounting unit with its own rules for recording financial transactions. City governments create different funds because of legal requirements or restrictions about the use of different resources (Finkler 2011). A resource in governmental accounting is an “item that can be drawn on to provide services to the citizenry” (GASB 2007, p. 2). Examples of funds include the general fund, enterprise fund, and capital fund, among others. 7 The fund accounting system helps improve financial control in city governments by tracking whether resources are used for the purposes defined by law. However, creating different funds also fragments information about the financial condition of city governments (Davies, Johnson, and Lowensohn 2017; Jonhson, Kioko, and Hildreth 2012). The critical question is, “What constitutes the city government if it has numerous types of funds?” 8
The Governmental Accounting Standards Board or GASB—which is a nonprofit organization that sets accounting standards and financial reporting requirements for state and local governments—addressed the issue of the fragmentation of city financial information through Statement No. 34 or Basic Financial Statements and Management’s Discussion and Analysis for State and Local Governments, which was issued in 1999. GASB 34 makes it easier to find information about the budget condition of a city by defining what essentially is the city government. It requires cities to provide financial information for the city as a whole, or the total primary government (TPG), which includes governmental activities (GAs) and business-type activities (BTAs). GAs include services funded through general taxes and grants, such as police, fire protection, public welfare, health services, education, and parks, among others. BTAs include services funded through user fees or charges such as water and sewerage (Mead 2011). Governments can transfer resources from BTA to GA (or vice versa) to close deficits, which means that focusing on one group of activities results in an incomplete picture of city budgetary solvency. This is analogous to measuring the financial condition of an individual with multiple bank accounts by focusing only on savings in a single account. The focus in this study is on the TPG, ensuring that all the resources controlled by government that can be used to provide services to citizens are measured when assessing budgetary solvency. Fiduciary funds are not included in the TPG because these resources are not owned by government but are only held in trust by the city for their true owners such as employees in the case pensions (GASB 1999; Jimenez 2017b, forthcoming; Mead 2011)
The data on TPG budgetary solvency are from government-wide financial statements in city Comprehensive Annual Financial Reports or CAFRs. CAFRs contain audited financial information, and comply with Generally Accepted Accounting Principles and other reporting requirements issued by GASB. CAFRs are collected for 560 out of 674 cities (or 83%) with a population of 50,000 or more, 9 and for fiscal years 2006 to 2012. All states except Vermont and New Jersey are represented in the sample. Vermont has no city with a population of 50,000 or more, whereas New Jersey cities do not prepare government-wide financial statements. All relevant financial information from CAFRs is manually encoded.
Government-wide financial statements employ an economic focus, which means that they report information on all economic resources of governments including all assets and liabilities—both current and noncurrent (GASB 1999). Current assets include cash and other resources readily convertible into cash, whereas noncurrent assets include, among others, capital assets such as buildings, equipment, or land. Current liabilities include payables within a year such as amounts owed to vendors and employees, whereas noncurrent liabilities are mostly long-term debt (see Mead 2011). The government-wide statements are prepared using accrual accounting, which records an “expense” for the year when a good or service is consumed whether cash changed hands or not, and reports “revenues” in the fiscal year when tax bills are issued rather than collected (Finkler 2011). With the accrual system, it not possible for officials to use accounting gimmicks to make it appear that their cities have balanced budgets, for example, by delaying payments to vendors to the next fiscal year to reduce liabilities in the current period, or advancing collection of next fiscal years’ tax receivables to increase current year revenues (Jimenez 2017b, 2017c, forthcoming). Equally important, accrual accounting makes transparent the issue of unfunded liabilities or nonpayment of city contributions to pension and other postemployment benefits to balance the current year’s budget. Accrual accounting records such nonpayments as liabilities because current employees already “earned” or have a legal claim to such benefits, even if these will be paid out in the future. Because a city is legally obligated to pay for them, unfunded benefit contributions represent a claim on a city government’s assets (Jimenez forthcoming).
The first measure of budgetary solvency is the change in total net position, which has been used in several studies (see Jimenez 2017a, 2017c, forthcoming; Johnson, Kioko, and Hildreth 2012; Rivenbark, Roenigk, and Allison 2010; Wang, Dennis, and Tu 2007). Net position represents residual economic resources, specifically, “resources of government has left over to use for providing services after debts are settled” (Mead 2011, p. 43). A simplified accounting equation for net position is total assets minus total liabilities equal net position (GASB 2011, 2012). Change in net position is standardized by expenses, 10 which are different from expenditures. Reflecting the accrual basis of accounting, expenses include the costs of services or goods consumed by a government whether cash payment was made or not in the current fiscal year, benefits earned by employees in the current period even if such benefits will be paid out in the future, and the annual depreciated costs of capital assets, which are spread out over their operational life (Mead 2011). Change in net position is considered a measure of the flow of resources into a government. A negative change in net position indicates that a city’s fiscal position worsened in a given fiscal year (Rivenbark, Roenigk, and Allison 2010).
The second measure of budgetary solvency is unrestricted net position, which has also been used in extant public budgeting research (see Jimenez 2017a, 2017b, 2017c, forthcoming; Johnson, Kioko, and Hildreth 2012; Wang, Dennis, and Tu 2007). Unrestricted net position functions as a reserve for a city government and can be used for any purpose (Jonhson, Kioko, and Hildreth 2012). Unrestricted net position includes both liquid and less-liquid resources. The economic resources focus of the governmental-wide statements comes to light here: Unrestricted net position covers not just liquid cash but all residual and less-liquid economic resources, such as long-term securities investment and inventory, which can be sold by government to produce cash and provide services (Mead 2011). Unrestricted net position is also standardized by expenses. Unrestricted net position is considered a measure of a city government’s stock of resources. A negative ratio indicates that a city government does not have enough resources to cover its expenses in the current period (Mead 2011). It can also mean that a government is using up past savings to pay for current services, or is shifting part of current costs to future taxpayers by issuing debt or underfunding the pension and other postemployment benefits already earned by current employees but will be paid out in the future (Davies, Johnson, and Lowensohn 2017; GASB 1999, 2007; Jimenez 2017c, forthcoming; Jonhson, Kioko, and Hildreth 2012; Mead 2011).
The study uses 2006–2012 data on change in net position and unrestricted net position. In addition, the analyses smooth out the data by using three-year moving averages because single-year observations contain a lot of noise, specifically deep troughs and peaks, which can bias the results. Moving averages focus on the underlying trends in the budgetary solvency measures (Jimenez 2017b, 2017c, forthcoming).
Measuring City-Level Initiative
Data on city initiatives are from the 2001, 2006, and 2011 Form of Government Surveys of the International City/County Management Association (ICMA). The surveys cover cities with a population of 10,000 or more, and include as respondents appointed city managers and others occupying similar positions. The ICMA survey defines the initiative as a process that “allows citizens to place charter, ordinance, or homerule changes on the ballot by collecting a required number of signatures on a petition.” 11
Following the practice in the empirical literature, this study tests for the effects of the availability of the initiative by employing a dummy variable indicating whether a city allows initiatives or not. Approximately 77% of the cities in the sample have the initiative. The focus on the availability of the initiative, rather than on its actual use, is based on the argument in the literature that the total effect of an initiative on policy includes its direct and indirect effects (Gerber 1996, 1999; Matsusaka 2004, 2005a, 2005b, 2005c). As Lupia and Matsusaka (2004) argue, “the initiative has both a direct effect (measures approved by the voters) and an indirect effect (changes in legislative behavior), so its effect on policy cannot be measured by examining only the propositions that actually pass” (p. 472).
Control Variables
The models control for measures of local economic condition, sociodemographic and political factors, the intergovernmental context, city fiscal policy choices, and government professionalization. Because of space consideration, a detailed discussion of the controls is not presented here but can be found in Jimenez (2017a, 2017b, 2017c, forthcoming). Economic controls include the level of and change in housing price index, unemployment rate, and year-on-year change in the number of unemployed. Sociodemographic and political controls include population size and change, median household income, and a measure of citizens’ political ideology, which is Tausanovitch and Warshaw’s (2014) policy conservatism index. Intergovernmental controls include city own-source revenues as a percentage of intergovernmental revenue transfers, per capita operating and capital grants from state and federal governments, dummy variables indicating whether a city has access to the sales or income tax or utility revenues, and a service index, which is the count of the type of services offered by cities based on data from the Census of Governments. City fiscal policy and government professionalization controls include per capita debt and expenses, property tax as a percentage of total tax revenues, and council-manager government form. Table 1 provides the details on variable operationalization, data sources, and descriptive statistics. All nonratio fiscal and income measures are converted to year 2000 dollars to control for inflation using the implicit price deflator from the Bureau of Economic Analysis.
Variable Operationalization, Data Sources, and Descriptive Statistics.
Note. The total number of states is 48, and the total number of unique cities is 554 (although CAFRs are collected for 560 cities, six cities are not included in the analysis because they have no data on initiatives). The summary statistics presented in this table are based on actual values of the variables. CAFR = Comprehensive Annual Financial Reports 2006–2012; ICMA = International City/County Management Association Government Form Survey 2001, 2006, 2011; FHFA = U.S. Federal Housing Finance Agency (https://www.fhfa.gov/); ACS = American Community Survey 2006–2012 (https://www.census.gov/programs-surveys/acs/); CoG = Census of Governments 2007 (https://www.census.gov/govs/cog/); IRI = Institute and Referendum Institute 2005 (see http://www.iandrinstitute.org/).
Estimation Approach
The models are estimated using least squares dummy variable or LSDV regression to address the issue of unobserved heterogeneity. All models include state dummies to control for unobserved, time-invariant, state-specific factors that affect city budgetary solvency (such as tax and expenditure limits [TELs], balanced budget requirements, homerule law, and debt limit, among others), and year dummies to account for contemporaneous error correlation.
Marschall and Ruhil (2005) argue that the initiative is endogenous. Specifically, residents in initiative cities might prefer higher spending and taxation, predisposing their cities to fiscal difficulties, in which case any observed relationship between the initiative and budgetary solvency is spurious. Possible endogeneity is directly addressed using instrumental variable or IV regression, which involves two stages. In stage 1, the initiative is modeled as a function of all exogenous variables in the solvency models and a set of excluded instruments. In stage 2, the predicted values of the initiative are used instead of the actual values in the solvency models.
Valid instruments satisfy two conditions, specifically, that they are correlated with the endogenous regressor (the availability of city-level initiative) but are uncorrelated with the outcome variable of interest (measures of city budgetary solvency). The analysis uses two excluded instruments measured at the state level. States generally define the powers and responsibilities of their local governments (Krane, Rigos, and Hill 2001). The first instrument is a modified version of Matsusaka’s (2009) instrument, 12 which is an index of state legal provisions on local initiatives that measures whether state law does not allow all cities within a state to have the initiative, allows only some cities, and allows all cities. Because these are state laws on local initiatives, the index should strongly predict the presence or absence of city initiative. The index is also unlikely to be directly correlated with city budgetary solvency because the state laws do not contain specific provisions on city policies on expenses, revenue, or debt. The second instrument is the status of state-level initiative prior to 1950. It is assumed that citizens who prefer direct participation in state-level decision making via the state-wide initiative are also likely to demand participation rights in city-level governance via the local initiative. Thus, the presence of state initiatives should correlate strongly with the availability of city initiatives. State-level initiatives are unlikely to be directly correlated with city budgetary solvency because they do not directly affect individual cities’ fiscal policy choices. 13 Although the theoretical justifications for the instruments are sound, their validity needs to be established empirically, which is done in the succeeding section.
Empirical Results
Primary Models
Table 2 presents the results of the primary models. All standard errors are heteroskedasticity-robust. Note that variables with skewed distribution, including the measures of budgetary solvency, are log transformed. Because logged zero or negative values are undefined, a constant value is added prior to log-transformation to ensure that no observations are dropped. The outcome variable in panels 1 and 2 is the unrestricted net position ratio (hereafter called the level ratio), and in panels 3 and 4 is the change in total net position ratio (hereafter called the change ratio). Panels 1 and 3 are estimated using LSDV, and panels 2 and 4 using IV regression.
Results of Primary Models LSDV Regression and IV Regression.
Note. Standard errors are heteroskedasticity robust. Although budgetary solvency data are from 2006 to 2012, inclusion of single-year-lagged independent variables means that 2006 data are dropped. The base year is, therefore, 2007. The base state is Alabama, whereas the base region is the West. Results for year, state, and region dummies are not shown because of space consideration. LSDV = least squares dummy variable; IV = instrumental variable; DV = dependent variable.
Significance at 1%, **significance at 5%, *significance at 10%, based on two-tailed tests.
Panel 1 shows that the local initiative is negatively correlated with the level ratio, and this relationship is moderately statistically significant (p < .05). Because the level ratio is logged, the coefficient for the city initiative has no straightforward interpretation. To assess the magnitude of the effect of the city initiative, predicted effects are calculated for a city with an initiative and those without, holding other control variables constant at their means. The predicted effects are then exponentiated. Figure 1 shows that a city with an initiative has a level ratio of 0.47, which means that its TPG or government-wide reserves can cover 47% of expenses in case of a significant drop in revenues because of a recession or other unexpected external event. A city without an initiative can cover 53% of expenses, or 6 percentage points higher than in an initiative city. The economic significance of 6% can be examined by putting the figure in dollar terms, and comparing the amount with city budgets for basic law enforcement, maintenance, and social services. For the average city, 6% of TPG expenses is equivalent to $21.8 million (in year 2000 dollars). For all cities in the sample, $21.8 million is approximately 154% of the average real operational expenditures for public welfare ($14.2 million), 104% of fire protection ($20.9 million), 96% of health and hospital services ($22.7 million), 71% of sewer and waste management ($30.9 million), and 54% of police services ($40.4 million). 14
Marginal effect of initiative on unrestricted net position ratio.
For control variables, cities with high level ratios include those that have high median household income, high housing price index, more politically conservative residents, council-manager governments; earn revenues from public utilities; and provide more services. In contrast, cities with low level ratios include those with a big population, and those that levy an income or sales tax, receive higher operating grants, spend more and issue higher debt in previous fiscal years, and are dependent on intergovernmental revenues and property taxes.
Panel 2 reestimates panel 1 using IV regression. It is important to note that because the excluded instruments are measured at the state level and are time-invariant, they are perfectly collinear with the state dummies. For this reason, state dummies need to be replaced with region dummies. Stage 1 of the IV regression shows that the state law index and state-level initiative are significantly and positively correlated with the presence of city initiatives. The F-statistic (F = 94.13, Prob.> F = 0.00) is nine times the prescribed floor of 10 (see Staiger and Stock 1997), indicating that both variables are strong excluded instruments. The underidentification statistic in stage 2 is highly statistically significant (χ2 = 134.74, Prob. > χ2 = 0.00), which means that the instruments, jointly, are strong predictors of the city initiative. In stage 2, the Hansen J statistic is insignificant (χ2 = 0.74, Prob. >χ2 = 0.39), which means that there is no direct correlation between the instruments and the level ratio. The instruments meet the two validity conditions. Addressing possible endogeneity, city initiative continues to have a negative and moderately significant (p < .05) relationship with the level ratio. Note that the coefficients of city initiative in the LSDV and IV models cannot be directly compared because LSDV includes state dummies, which are replaced with region dummies in the IV regressions.
Panel 3 shows the LSDV results for the change ratio. The coefficient for city initiative is negative and moderately statistically significant (p < .05). Figure 2 shows the predicted effects of the initiative. For a city with an initiative, the change ratio is approximately 0.12, which means that the average annual increase in its total net position represents 12% of government-wide expenses. In comparison, the increase in the net position of a noninitiative city is equivalent to 14% of its expenses, or a difference of two percentage points. For the average city, 2% of government-wide expenses is equivalent to $7.1 million (in year 2000 dollars). This amount is equivalent to 50% of the average real operational expenditures for public welfare services, 34% of fire protection, 31% of health and hospital services, 23% of sewers and waste management, and 18% of police services.
Marginal effect of initiative on change in total net position ratio.
For control variables, cities with positive change in total net position include those with high housing price index, lower unemployment, small population, faster population growth rate, and those that offer more services, are least dependent on intergovernmental revenue relative to own-source revenue, receive more operating and capital grants, do not have access to sales and income taxes, have utility revenues, and have lower expenses in previous fiscal years.
Panel 4 reestimates panel 3 using IV regression. The same instruments used in panel 2 are also used in panel 4. Relevant tests indicate that the instruments are valid (F test: F = 94.13, Prob. > F = 0.00; Underidentification statistic: χ2 = 134.74, Prob. > χ2 = 0.00; Hansen J statistic: χ2 = 2.49, Prob. > χ2 = 0.11). Controlling for possible endogeneity, the coefficient of city initiative remains negative and moderately statistically significant (p < .05).
Robustness Tests
How robust are the main findings to the type of controls included in the models, alternative operationalization of variables, and change in the estimation approach? Table 3 presents the results of additional tests. First, some might disagree with the inclusion of certain control variables in the primary models. Dropping all other controls except for state and year dummies, city initiative continues to have a negative and statistically significant relationship with the level and change ratios. Second, because budgetary solvency in the current year is highly dependent on solvency levels in previous years, serial correlation is a possibility. Using Prais–Winsten regression to control for first-order autocorrelation does not change the study’s main conclusions. 15 Third, the findings remain robust when using annual values of the change and level ratios, instead of three-year moving averages. Finally, the city initiative is interacted with year dummies to assess whether the effect of initiative varies across years, specifically, whether such effect is more pronounced during the recession years. The estimates for the interaction terms are statistically insignificant.
Results of Robustness Tests.
Note. Standard errors are heteroskedasticity robust. For all panels, the base state is Alabama. Panels 7 to 12 include single-year-lagged independent variables, which means that 2006 data are dropped. The base year is, therefore, 2007. Panels 5 to 6 have no lagged independent variables and covers 2006 to 2012. Consequently, the base year for panels 5 to 6 is 2006. Results for year and state dummies are not shown because of space consideration. DV = dependent variable; LSDV = least squares dummy variable.
Significance at 1%, **significance at 5%, *significance at 10%, based on two-tailed tests.
Discussion and Conclusion
The empirical analysis shows that initiative cities have weaker budgetary solvency compared with noninitiative cities, which goes against expectations derived from current theories on the effects of initiatives on fiscal discipline in the public sector. The question that begs an answer is why? There are two possible explanations. First, initiatives fail to make city governments more responsive to voters’ wishes. City officials continue to ignore voters, and maximize spending and revenues for personal gain. A problem with this explanation is that it goes against the preponderance of evidence in the literature that initiatives have brought government policies closer to the median voter’s preference (in the case of fiscal policies, see Matsusaka 2004, 2005a, 2014; for social policies and political reforms, see Gerber 1996, 1999; Matsusaka 2005c, 2014).
A second more plausible explanation is that initiatives are in fact successful in forcing city governments to respond to voters’ demands, but what citizens prefer at the local level does not necessarily improve city budgetary solvency. Specifically, it is possible that initiatives increase spending at the local level but reduce revenues. Primo (2010) argues that local initiatives empower interest groups, which demand particularistic goods from city governments. However, it may not be interest groups only that demand higher spending, but citizens in general. Oates (1999), for example, recognizes the possibility that because local governments are closer to citizens compared with higher-level governments, such as state and federal governments, local residents’ access to city officials makes it easier to transmit demand for more local public services. Matsusaka (2004) also points out that citizens prefer local spending to avoid redistribution. Local spending ensures that city revenues are used to finance public services that benefit only city residents and not redistributed to other jurisdictions as what happens when spending is concentrated at the state and federal levels.
Achieving budgetary solvency, however, requires that an increase in local spending is matched by an increase in local revenues. Research shows that most voters oppose property taxes (Jimenez 2014b, 2017c). Numerous states and localities have adopted TELs that primarily target the size and growth of property taxes (Jimenez, 2017a, forthcoming), and most of these limits were adopted through initiatives (Brooks and Philips 2008). Several studies show that TELs have reduced local property taxes but do not constrain local spending (see the studies cited in Jimenez 2017a, forthcoming). This is unfortunate for cities because the property tax is a major revenue source, comprising more than half of total tax revenues of the cities in the sample (see Table 1). Because of the difficulty of raising property taxes, extant studies find that TELs lead to deficit financing in midsized and large cities, especially during an economic recession (Jimenez forthcoming; Ross, Yan, and Johnson 2015). 16
Very few disagree with the democratic ideal that citizens should exercise their right to directly shape policies that affect their lives. However, the result of this study is a sobering reminder that the fulfillment of a normative ideal—giving power to the people—does not always lead to desirable fiscal outcomes. This study shows that initiative cities have weaker budgetary solvency in comparison with noninitiative cities. Of the two measures of budgetary solvency, the result for unrestricted net position should be a subject of deep concern for local policymakers and citizens alike. Unrestricted net position is considered a measure of interperiod equity, which means that a “government is raising sufficient revenues each period” and not “deferring costs to the future or using up accumulated resources to provide current-period services” (GASB 1999, p. 84). A consistently declining unrestricted net position indicates that a government is not raising adequate revenues in the current fiscal year to meet expenses in the same period, and is instead raiding past savings, or passing the burden of paying for the costs of current services to future taxpayers by issuing debt or underfunding pension and other postemployment benefits obligations (GASB 1999, 2007; Jimenez 2017c, forthcoming; Jonhson, Kioko, and Hildreth 2012; Mead 2011). 17 Because the city initiative is associated with declining unrestricted net position, the unavoidable conclusion is that the initiative process has inadvertently allowed current users of city services to shift part of the responsibility for paying for the benefits they now enjoy to the past and future taxpayers.
It is apt to end this study by pointing out questions that were not fully answered here. First, because of limited data on city initiatives, it was not possible to assess how signature requirements moderate the relationship between initiatives and budgetary solvency. Second, it is important to find out what exactly happened in the cities covered in this study, that is, the actual measures initiated, passed, or rejected by voters, to provide context to the findings. Unfortunately, there is no single organization that tracks ballot measures across the thousands of cities in the United States. Finally, how initiatives interact with local sociopolitical characteristics (such as population size, income, or citizen political ideology) and other local political institutions designed to increase responsiveness to citizens (such as term limits and at-large versus district elections, among others) needs to be assessed. These crucial questions and data-gathering challenges will hopefully encourage more research on how local initiatives shape fiscal outcomes in cities.
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.