The Dynamic Impacts of State Revenue Capacity on Highway Investment

Abstract

This study investigates the short-term effects of state-highway EARs (EARs) and federal-grant expenditure on present-year state revenue sources and highway investment, when another state-highway source, general revenue, is controlled for. In the dynamic framework, this article argues that general revenue interacts with highway EAR and federal-grant reimbursements since these revenue sources are substitutable. The results from the panel vector auto regression (PVAR) indicate that (a) EAR and federal grants are inversely related to general revenues used in highway programs, (b) when controlled for simultaneity, autocorrelation, and omitted-variable bias, EAR and federal grants are not significant for state-investment decisions, and (c) general revenue inversely relates to state investment because there is competition for public resources between highway and general-revenue programs.

Keywords

infrastructure capital budgeting capital financing transportation highways

Introduction

In the private sector, a firm makes capital-investment decisions based on its expected profits, market demands, and internal-financing capacity. Unlike the private sector, governments rarely compare profits with costs (Stanley & Block, 1984). Since state and local governments are subject to institutional rules, such as tax policies, debt limits, as well as balanced budget requirements, their capital-investment decision may depend on internal-resource availabilities. Previous studies have found significant relationships among fiscal capacity, that is state’s major tax rates (Eberts & Fox, 1992; Holtz-Eakin, 1991), earmarked revenue (EAR; Jung, 2002), federal grants (Becker, 1996; Gamkhar & Oates, 1996), monetary policies and long-term debt rates (Phelps, 2001), macroeconomy (Ho, 2008), and state or local public investment. If subnational governments make investment decisions based on these fiscal availabilities, the next question is whether these revenue sources respond to the availability of other sources, that is, crowding in or crowding out general revenues prior to determining subnational government investment? In terms of highway literature, understanding the effects of each fiscal source on other sources and on public investment is necessary because (a) highway-investment costs are relatively large compared to subnational government revenue capacities, (b) these governments must pay highway costs up front, and (c) investment expenditure is subject to appropriation and balanced-budget requirements every fiscal year.

In the state-highway literature, revenue availability, including per capita state-highway EAR (Dye & McGuire, 1992; Nesbit & Kreft, 2009) and per capita highway federal grants (Ghamkar, 2000, 2003; Knight, 2000; Nesbit & Kreft, 2009), determine per capita state-highway investment. Since these studies investigate the effect of fiscal capacity on state-highway investment in a static framework, the dynamic roles of these revenue sources are not well understood, for example, whether the major state-highway revenue sources, earmarked taxes, federal grants, general revenue, and long-term bond proceeds interact against one another to alter highway-revenue compositions and investment in the following year. Furthermore, given that these previous studies assume the insignificant roles of other highway-revenue sources, that is, general revenue, the results may be subject to omitted-variable bias. If these omitted variables do not determine state-highway investment, and at the same time, do not determine the two included revenue sources, that is, earmarked highway revenue and federal grants, the results will be unbiased (Stock & Watson, 2007). However, it is possible that the previously excluded variable, general revenue used to finance highway spending, determines state investment at the marginal level and at the same time alters the role and amount of other highway-revenue sources.

Since positive studies (Borg & Mason, 1988; Deran, 1965; Dye & McGuire, 1992; Miller & Pierce, 1997; Spindler, 1995) found that EAR is used as the fiscal substitution for general revenue, the availability of the EAR may be fungible for general revenue that would otherwise be appropriated. When the EAR is used to substitute general revenue, the EAR does not necessarily increase earmarking programs on a one-to-one dollar basis. As Ghamkar (2003) describes, “states generally start a project using their own money; that is, they provide front-end financing for the project and receive cash for the federal share of the project’s costs (federal expenditure) as work is completed” (p. 5). As a result, state governments may need to look for resource availability elsewhere, including general revenue for financing highway investment costs up front. In this case, there will be resource competition between general revenue programs and highway programs, thus increasing federal aid may result in decreased reliance on general revenue but may not necessarily result in increased highway spending.

This study is different from previous studies in two ways. First, it investigates the roles of highway-revenue sources in a dynamic framework where lagged revenues alter current revenues prior to determining current spending. This understanding is important for future highway policy options and financing reforms, given that federal highway grant institutions are unique and state fuel-tax rates are not necessarily a good indicator of highway demands (Wachs, 2005). Second, this study controls for the role of general revenue as another revenue source for highway spending, especially when previous results suggest that EAR and general revenue are substitutable.

This article is organized as follows. The next section describes the theoretical arguments regarding EAR and federal grants in relation to state highway spending. The third section provides the methodology and data. The following section presents the empirical results, and the last section presents the conclusion.

Theoretical Background

State-Highway Financing Practices

State governments rely on four major sources to finance their state-administered highway systems: (a) earmarked state highway revenue, which includes fuel taxes, highway-user charges, and vehicle fees; (b) federal grants; (c) long-term bond proceeds; and (d) general-revenue funds (Puentes & Prince, 2005). Data from U.S. Department of Transportation, Federal Highway Administration (2001) indicate that in 2001, 43% of total state highway revenues came from earmarked state-highway taxes, 30% from federal grants, 13% from long-term bond proceeds, and 4% from state general-revenue funds. In 2006, 42% of total state-highway revenue came from earmarked state-highway taxes, 30% from federal grants, 15% from long-term bond proceeds, and 3% from state general-revenue funds. These statistics suggest that state governments, in an aggregated term, rely heavily on earmarked state-highway revenues and federal grants and relatively less on state general revenues or bond proceeds. Hence, the main question is how the EAR interacts with general-revenue funds in state-government finance, that is, whether the EAR is fungible and whether state highway investment programs are competing with general-revenue programs.

At individual level, state governments vary in terms of highway-revenue sources, revenue composition, and reliance on state general revenue. This variation is not only across states in the same year, but also across times within the same state. Figure 1 presents the percent of earmarked state-highway revenue used to finance government spending in state nonhighway programs as reported by U.S. Department of Transportation, Federal Highway Administration (2010) from 2001 to 2006. In this study, EAR refers to state taxes and fees imposed on the owners and operators of motor vehicles for their uses of public highways (Federal Highway Statistics, 2010). The EAR also includes per gallon motor-fuel taxes, motor-vehicle registration fees and motor-vehicle sales taxes. The total amount of EAR used in this study was drawn from Federal Highway Statistics Table MF3: disposition of State Motor-Fuel Tax Receipts and Table MV3: disposition of state motor-vehicle and motor-carrier tax receipts in which only the portion available for distribution to state-administered highway (roads) is reported (U.S. Department of Transportation, Federal Highway Administration, 2010). Thus, the EAR for interstate-highway spending is excluded in the data presented in Figures 1 and 2.

Figure 1.

Percentage of highway EAR used for state nonhighway programs.

Figure 2.

Percentage of general-revenue funds used for state-administered highways.

Due to limited space, only 10 states are presented in the figure to show the variations across states and times. In the Figure 1 data table, looking across the 2006 column in the table, Massachusetts ranked first in exporting earmarked highway revenue to other state nonhighway programs; however looking across the Massachusetts row, the level of EAR exported to other programs is volatile, for example, 48%, 21%, 22%, 27%, 11%, and 75% in 2001, 2002, 2003, 2004, 2005, and 2006, respectively. Delaware is an example of a state for which 100% of its earmarked highway revenue was used to finance state-administered highways. However, from 2001 to 2006, Delaware was the only state that did not export earmarked state-highway revenue for other spending purposes. During the same period, West Virginia used 100% of its earmarked highway revenue only in years 2001 and 2002, and exported EAR for about 2% to 6% for the rest of the 4 years. During the same period, 46 continental states (excluding Delaware and Virginia) transferred some amount of their earmarked highway revenue to other state nonhighway programs; such practices cannot be regarded as a fixed effect, since they change from year to year within the same state, except for Delaware.

Figure 2 presents the percent of general revenues used in state-administered highway program financing. As shown in the figure, looking across the columns, within the same year, states vary in terms of using their general revenue. For example, Utah, Massachusetts, and Kentucky ranked the first in general-revenue reliance at 24%, 47%, and 20%, in 2001, 2002, and 2005 respectively. Looking across the rows, each state is volatile in terms of general revenue reliance. For example, Utah relied on general-revenue funds for 24%, 15%, 9%, 5%, 14%, and 18% in the years 2001, 2002, 2003, 2004, 2005, and 2006, respectively. Indiana relied on general-revenue funds for 0%, 23%, 0%, 0%, 5%, and 4%, in the years 2001 through 2006, respectively.

The data and pattern of highway financing presented in Figures 1 and 2 indicate that at the individual-state level, the composition of the four major highway revenue sources vary across states and times. When the revenue composition in a state is not only different from others, but also from itself across years, the difference is not due to a state’s fixed effect. This signals the need to control for other revenue sources than earmarks and grants when examining the effects of fiscal resources on highway spending.

EAR: Fiscal Substitution or Fiscal Constraint?

An empirical finding indicates that at the state and local level, a jurisdiction’s reliance on user-charge revenue is inversely related to personal income (PI) and the capacity of state and local government to export business taxes from a jurisdiction (Netzer, 1992). This implies that, everything else equal, taxpayers prefer to keep their tax rates low and create new revenue sources to finance new public services rather than raising their major tax rates, for example, income taxes and sales taxes, which usually go to general-revenue funds. Another study indicates that as federal income-tax policies allow a greater deductibility rate for major state and local taxes, state and local governments invest more in their highway infrastructure (Eberts & Fox, 1992). In another study, per capita income and federal grants do not significantly determine state-highway spending at the conventional level, while interest rates for government borrowing significantly explain the level of state-highway investment (Phelps, 2001). These results suggest that at the subnational level, governments tend to maintain their major tax rates and look for new revenue sources to finance their capital investment.

The above empirical results imply that EAR may be used to substitute for general revenue. This is known as fiscal substitution. Such extra revenues may not marginally increase spending in the favorable programs because, (a) the spending in that program would otherwise be financed by general revenue had extra resources not been available, or (b) since governments at the state level are relatively large compared with the local level, they may have more latitude to divert resources at the marginal level to balance revenue and expenditure (Netzer, 1992; Phelps, 2001). As many studies suggested (e.g., Borg, Mason & Shapiro, 1993; Dye & McGuire, 1992; Evans & Zhang, 2007), when funds are earmarked for program A, for instance, the expenditure of program A does not increase by EAR amount. The EAR might have been used for program A but instead revenue from other sources, for example, general funds might have been reduced or the revenue might have been used for other programs.

In the public-finance literature, the role of EAR in earmarked spending is debatable. On the one hand, EARs are considered as a public demand indicator on the earmarked program; thus, using EAR to finance earmarked programs yields optimal government size (Browning, 1975; Buchanan, 1963). In other words, EAR is viewed as a tool to separate spending decisions in general-revenue programs and earmarked programs so that there will be no resource competition between the two programs. It should be noted, however, that this assertion is based on the condition that government general-revenue funds are exogenous; and thus, EAR and general revenue are not crowded in or out (Browning, 1975; Buchanan, 1963). If this condition holds, everything else equal, the relationship between EAR and earmarked spending should exhibit a ratio of one-to-one: a US$1 increase in EAR results in a US$1 increase in earmarked spending. Furthermore, general revenue used to finance earmarked spending should have zero effect on earmarked spending, given that this is definitely separated from earmarked-revenue sources.¹ Under such an assumption, there is little need to examine the relationship between EAR and earmarked spending.

On the other hand, EAR is considered a political tool used to avoid increasing major tax rates. This occurs in the situation where citizens’ trust in government has declined (Rubin, 2008), tax burdens cannot be exported to outsiders (Netzer, 1992), state legislature wants to reduce competition for fiscal resources in general-revenue funds by locking down the priority of spending on earmarked programs (Rubin, 2008), and where the demand for public service increases but the willingness to tax themselves is stable (Goetz, 1968). If EAR is meant to be a tool to avoid major tax hikes, EAR will not increase earmarked spending because it will simply substitute for other fiscal resources that otherwise would be appropriated (Michael, 2008). Previous findings support this argument since they indicate that there is no significant relationship between EAR and government spending (Deran, 1965). However, there is a significant but inverse relationship between EAR and general revenue appropriated for earmarked programs (Borg & Mason, 1988).

Since the data presented in Figures 1 and 2 indicate that state-government highway revenue and general revenue are transferrable, the two revenue sources are substitutable in financing state-administered highways. At the marginal level, because there is competition between state-highway and state-nonhighway programs, the larger the general revenue, the smaller the spending on state-highway programs. At the marginal level, because the earmarked highway revenue is substitutable, an increase in earmarked highway revenue does not necessarily result in an increase in state-highway spending.

Federal Grants: Fiscal Substitution or Flypaper Effect?

Income effect means that the purchasing power of consumer voters increases by additional revenues, either from grants or other sources. The flypaper effect means that when grants hit the government the monies stick in the public sector. Majorities of early empirical studies indicates that income effects of federal grants on subnational highway spending is about 5 to 10 cents per US$1 (e.g., Stotsky, 1991), whereas the flypaper effects on state-highway spending is about 25 to 50 cents per $1 (e.g., Meyers, 1987; Ryu, 2006).

In the highway literature, there are mixed empirical results for the roles of federal grants in subnational highway spending when some variables are included or excluded in empirical models. For example, Ghamkar (2003) found that a US$1 increase in per capita federal obligations results in a 78 cent increase when earmarked highway revenue is excluded in the author’s testing models. Nesbit and Kreft (2009) found that a US$1 increase in per capita federal-grant expenditure, which is predicted by lagged federal-highway obligations, results in a 92-cent increase on highway spending, controlling only for the effect of earmarked highway revenue. Eberts and Fox (1992) found an insignificant effect of per capita federal aid on state-highway spending when long-term debt proceeds, which is another major source of state-highway revenue, is controlled for. Dye and McGuire (1992) also found an insignificant effect of federal grants when EAR, which is another source of federal-highway spending, is controlled for. Since it is difficult to predict the effect of federal grants based on these previous findings, given that these studies are varied in terms of model specifications, institutional-highway-budget knowledge was consulted to predict the effect of grants as follows.

Ghamkar (2003) describes that federal government provides highway aid to state governments in the form of contract authorities that will be passed through highway-authorization bills every 4 to 6 years. The federal-highway administration, then, distributes resources based on a formula which usually includes such elements as income, population, highway-lane miles, and rural and urban usage (U.S. Department of Transportation, Federal Highway Administration, 2007). The appropriated aid is available in terms of granted credits to create obligations during the specified period, which usually ranges from 4 to 6 years. Unlike other federal aid, for highways, state governments do not receive cash from the federal government prior to project construction; instead, they receive cash reimbursement for grants later once the project is completed (Ghamkar, 2003). State governments, thus, in practice, decide and initiate highway projects based on their own capacities and receive reimbursement for the partial costs of the previously completed projects as specified in federal-grant contracts.

Excluding the points that, (a) state governments receive federal-aid reimbursement as an award for previous projects, (b) that state governments need to use own-source revenue to pay for the costs of highway projects up front and such costs must be appropriated or reappropriated each fiscal year, and (c) that state governments are subject to balanced-budget requirements, the effect of federal grants on state-highway spending can be examined under a static framework. In this static setting, federal aid is the only external resource which is exogenously determined by federal-highway obligations. The empirical studies in this setting usually find a relatively large effect of federal grants on highway spending (around 28 to 78 cents) and the authors conclude that the flypaper effect hypothesis holds (Nesbit & Kreft, 2009).

However, when the three points mentioned above are considered, federal resources may not necessarily be exogenously determined, and thus, a static framework may yield different results. For the first point, the external resource, that is, federal-grant expenditure, transferred to subnational governments in each fiscal year, is in fact an internal resource because it is partially and internally determined by state-highway investment in the previous years.

For the second point, from state governments’ perspectives, federal grants, measured through either budget obligation or reimbursement, may not influence state-highway spending because it was simply not available at the time that highway-spending decisions were made. This is similar to a retailer’s discount, in which a purchaser must pay for the cost up front and mail in the rebate form to receive the discounted portion in cash later. Consumer-psychology studies found that the rebates do not significantly influence purchasers to buy the discounted products as the savings from rebates do not occur at the time of the transaction, especially in the case where rebate amount is relatively low or when the rebate-processing time is relatively long (Kim, 2006). Another example is a product that is a big-ticket item, for example, automobile (Beltramini & Chapman, 2003), where the purchaser is aware of the full price before rebate as opposed to those who view the net price after the rebate (Ong, 2008).

For the third point, resources from elsewhere, including EAR and general revenue, may be used to finance highways because state governments can shuffle resources (under some budgeting rules) and must balance the budget. Thus, in a dynamic setting, federal grants will not influence highway investment. This is because the grants are internal resources resulting from the previous year’s decisions by state-government institutions, the grants are not available at the time the decision is made, and states must find some resources to pay for the costs up front. Furthermore, there will be an inverse relationship between general revenue and federal grants due to fiscal substitution between general-revenue programs and highway programs.

Method

To understand state governments’ dynamic-investment behaviors, the vector auto regression (VAR) technique was used. VAR treats all variables in the model as endogenously determined by the other variables in the model for a certain lagged period. VAR is, thus, a system equation in which each variable in the estimating model is determined by the lagged values of all variables in the model, including itself. The system equation controls for (a) the simultaneity among the variables in the model and (b) the autocorrelation or the lagged effect of dependent variables on independent variables. For the causality tests, the three following issues should be addressed.

Endogeneity of the Model Variables

The VAR model must be built based on specific assumptions or institutional knowledge to determine which variables are endogenously determined and which are not (Stock & Watson, 2007). Based on the theoretical discussion in the above section, a government highway investment decision is endogenously determined by its internal-revenue availability and public demand for public infrastructure. That is, revenue availability and spending size in one year determines the revenue availability and spending size for the following year. Major state revenue sources, including earmarks, federal grants, general revenue and bond proceeds, and state economies, as well as existing highway stocks, endogenously determine each other in the following years.

The political variables used in previous studies included democratic governor, democratic senate, and democratic house (Nesbit & Kreft, 2009). Democratic governor, house, and senate were found insignificant for highway spending, but significant for federal-grant obligations (Nesbit & Kreft, 2009). Since these variables represent state-political institutions which are partially determined by taxing and spending levels in the previous years (Merrifield, 2000; Reed, 2006), these variables, in theory, are partially endogenous in the sense that highway-spending levels in the past influences a specific government party to win elections in the future (Merrifield, 2000; Reed, 2006). They thus should be included in the VAR model. Because the number of variables in the VAR model determines the number of equations in the system, the total number of variables should be limited (Stock & Watson, 2007). Each variable included in the model should be efficient in terms of carrying important concepts and being powerful in determining other variables (Stock & Watson, 2007). For this condition, each of the three political variables was alternately included in the model to examine the effect of democratic government. The results were compared across the three estimations. These results from the three estimations were not significantly different from those that are reported in the next section. The basic testing model is stated below:

E X P_{i, t} = f (P I_{i, t - 1}, L M U_{i, t - 1}, L M R_{i, t - 1}, E A R_{i, t - 1}, F E D_{i, t - 1}, G E N_{i, t - 1}, B O N D_{i, t - 1}, G O V_{i, t - 1}),

where EXP_i,t is per capita state-administered highway disbursement in US$2,000 value in state i during year t; PI_{i,t − 1} is per capita PI in US$2,000 value in state i during year t − 1; LMU_{i,t − 1} is per capita state highway LMU in state i during year t − 1; LMR_{i,t − 1} is per capita state highway LMR in state i during year t − 1; EAR_{i,t − 1} is per capita state-highway EAR in US$2,000 value in state i during year t − 1; FED_{i,t − 1} is per capita federal aids granted to state government as reimbursement for the previous project contracts in US$2,000 value in state i during year t − 1; GEN_{i,t − 1} is per capita general revenue funds transferred to finance state-administered highways in US$2,000 value in state i during year t − 1; BOND_{i,t − 1} is per capita long-term bond proceeds used to finance state-administered highways in state i during year t − 1; and GOV_{i,t − 1} is the dummy variable for democratic government in state i during year t − 1.

According to Congleton and Bennett (1995), the testing model for state-highway investment includes gas taxes to control for infrastructure price. Given that the above testing model is quite complex and that most states’ gas taxes are fixed effect; gas tax was excluded in the basic testing model. However, in an alternative model, the author included state gas taxes in the basic model and found that the main results for the fiscal variables do not change very much compared to those in the original model and that gas tax is not statistically significant in this alternative model. The insignificance of the state gas taxes may be due to fixed effects, that is, state motor-fuel tax per gallon does not change dramatically from year to year from 2001 to 2006. For example, during this testing period, the average change for state gas price is 0.00000993 cents per gallon per year with standard deviation of 2.3 cents per gallon.

Previously Omitted Variables

Previous empirical studies found that state-highway spending is significantly determined by public demands measured by PI, lane miles in urban areas (LMU), lane miles in rural areas (LMR; Ghamkar, 2003; Nesbit & Kreft, 2009), internal-revenue sources measured by EAR (Dye & McGuire, 1992; Nesbit & Kreft, 2009), federal grant expenditure (FED; Dye & McGuire, 1992; Ebert & Fox, 1992), bond proceeds (BOND; Ebert & Fox, 1992; Ghamkar, 2003), and political variables measured by government (GOV; Knight, 2002). All of these variables are measured in per capita and constant dollar value to standardize government size across country and to separate inflation effects from government size. The internal revenue source, general revenue (GEN), was not used in previous empirical models but is included in this study’s model based on the three following reasons.

First, as demonstrated by the highway-finance statistics presented in Figures 1 and 2, EAR and general funds are transferred in and out at various degrees across states and times. The level of general revenue used in state-highway finance cannot be treated as a fixed-state effect since the amount of the general revenue used in highways changes across periods within a state. Furthermore, if EAR is partially used, instead of fully used, to finance the earmarked program, the level and composition of revenue sources should be considered because the effect of EAR on public spending will be more complex (Dye & McGuire, 1992, p. 546). Second, existing institutional budgeting knowledge suggests that EAR is used by subnational governments as a fiscal substitution for general revenue that would otherwise have to be appropriated (Borg & Morson, 1988; Deran, 1965; Michael, 2008; Netzer, 1992; Rubin, 2008). Some studies have found no significant linkage between EAR and earmarked spending (Deran, 1965) and that the effects of EAR comprises a ratio of less than one-to-one in increasing spending (Dye & McGuire, 1992). Last, based on institutional-highway knowledge, state governments use their own source revenue to pay for highway project costs up front for projects eligible for federal grants and receive a reimbursement once the project is completed (Ghamkar, 2003). If all of these are correct, general revenue will alter the roles of EAR and federal grants given that they are substitutable.

Interpretations

The last condition involves interpreting the VAR results according to the structure of the system equation. Since VAR treats all variables in the model, including fiscal variables, as endogenously determined, each coefficient of the fiscal variables presents an inertia effect of each revenue source on state-government investment when other revenue variables, demand variables, and the political variable are controlled for (Love & Zicchino, 2006). The common way to interpret the results is to understand the impulse response of a variable to another variable relative to the others’ in the model. For example, as shown in the above equation, if a federal grant does not result in a higher-investment level (which represents price effect) but lower reliance on other highway-revenue sources such as EAR, GEN, and BOND (which represents income effect), EXP would not pick up the FED effects, but other revenue variables (i.e., EAR, GEN, BOND) would. If a US$1 increase in EAR (EAR) does not result in an increase in investment (revenue constraint hypothesis) but a decrease in other revenue sources (or lower reliance on other revenue sources, the fiscal-substitution hypothesis), the EXP would not pick up the EAR effects, but others revenue variables, including GEN, would. The same interpretation is true for GEN and BOND. Thus, an advantage of VAR is that it not only allows us to understand the effect of inertia-revenue capacity on current expenditure, but also allows us to understand how revenues interact against one another before determining the next year’s spending.

Table 1 presents the summary statistics. Data were derived from the 48 U.S. continental states during the period from 2001 to 2006. This yields 288 observations (48 states *6 years). All financial data were derived from federal-highway statistics in various years. State-administered highway disbursement data were used to measure state-government highway investment. All fiscal capacity variables are in per capita and constant dollar value. State highway rural lane miles and urban lane miles are in per capita and were derived from federal-highway statistics. Per capita PI is in constant dollar value. Per capita PI data were derived from the Bureau of Economic Analysis. All political variables are dummy variables, where 1 represents democratic ideology and 0 otherwise.

Table 1.

Summary Statistics.

Variable	Mean	Standard deviation	Minimum	Maximum
Fiscal availability
EXP (per capita real state-administered highway disbursement)	344.3	152.9	149.1	1,067.1
EAR (per capita real highway earmarked revenue)	133.7	65.5	10.8	449.5
FED (per capita real federal-grant expenditure)	117.0	66.7	40.0	424.5
GEN (per capita real general revenue used in state-administered highway programs)	15.0	29.4	0.0	251.4
BOND (per capita real long-term bond proceeds used in state-administered highway program)	30.6	52.1	0.0	363.7
Highway demand
PI (per capita real PI)	28,573	4,186	21,391	41,812
LMR (per capita state highway lane miles in rural areas)	11.5	1.0	7.8	13.0
LMU (per capita state highway lane miles in urban areas)	10.2	1.0	8.0	12.2
Political institution (each was alternately entered into the model)
GOV (democratic government = 1, 0 otherwise)	0.5	0.5	0.0	1.0
UPCOUT (democratic senate = 1, 0 otherwise)	0.5	0.5	0.0	1.0
LOWCOUT (democratic house = 1, 0 otherwise)	0.5	0.5	0.0	1.0
Total observation: 288

Empirical Results

Table 2 presents the VAR results. Each row presents each individual variable’s response to the lagged variables of the model.

Table 2.

VAR Results.

Response to	EXP (t − 1)	PI (t − 1)	LMU (t − 1)	LMR (t − 1)	EAR (t − 1)	FED (t − 1)	GEN (t − 1)	BOND (t − 1)	GOVPARTY (t − 1)
Response of
EXP	0.276***	0.011**	−77.628***	145.448***	0.037	−0.113	−0.646***	−0.235*	−8.085
	(2.557)	(1.980)	(−2.560)	(2.509)	(0.203)	(−0.593)	(−3.247)	(−1.771)	(−0.701)
PI	−0.035	0.542***	964.356***	−1682.393***	−1.429	6.073***	0.647	0.953**	−38.825
	(−0.076)	(9.195)	(5.897)	(−5.959)	(−1.073)	(4.045)	(0.845)	(2.351)	(−0.544)
LMU	0.000	0.000	0.662***	−0.197	0.003	−0.003*	0.001	0.001**	−0.019
	(−1.046)	(0.813)	(3.408)	(−0.978)	(1.581)	(−1.675)	(1.590)	(1.908)	(−0.234)
LMR	0.000*	0.000	0.082	0.385***	0.002	−0.001	0.000	0.000*	−0.044
	(−1.934)	(0.846)	(0.707)	(2.380)	(1.536)	(−1.549)	(−0.942)	(2.032)	(−0.987)***
EAR	−0.005	−0.003	−14.912*	5.849	0.110	0.052	−0.042	−0.035	−6.133
	(−0.274)	(−1.194)	(−1.859)	(0.492)	(1.188)	(0.540)	(−0.938)	(−1.427)	(−1.467)
FED	−0.024	−0.005**	−3.086	11.330	0.026	0.182***	−0.037	−0.016	−1.993
	(−1.559)	(−2.316)	(−0.669)	(1.722)	(0.473)	(2.806)	(−1.287)	(−1.192)	(−0.629)
GEN	0.026	0.001*	−3.157	5.527	−0.148**	−0.178*	−0.359***	0.054**	−1.663
	(0.530)	(0.190)	(−0.340)	(0.314)	(−2.150)	(−1.842)	(−3.259)	(2.131)	(−0.425)
BOND	−0.055	0.004	−16.059	29.611	0.148	−0.167	0.144	−0.007	28.421**
	(−0.740)	(0.846)	(−0.834)	(0.898)	(0.907)	(−0.908)	(0.551)	(−0.049)	(2.100)
GOVPARTY	0.000	0.000	−0.050	0.182	0.002**	0.001	0.001	0.000	0.556***
	(0.821)	(0.172)	(−0.398)	(0.924)	(2.072)	(0.351)	(1.449)	(0.728)	(5.219)

Note: EXP = per capita real state-administered highway disbursement expenditure; PI = per capita real personal income; LMU = per capita state highway lane miles in urban areas; LMR = per capita state highway lane miles in rural areas; EAR = per capita real highway; EAR earmarked revenue; FED = per capita real federal grant expenditure; GEN = per capita real general revenue used in state-administered highway programs; BOND = per capita real long-term bond proceeds used in state-administered highway program; GOVPARTY = democratic government party. The VAR model was estimated by General Moment Methods (GMM), time, and state fixed effects were removed prior to estimation. As suggested by Stock and Watson (2007) and Love and Zicchino (2006), state fixed effect is removed by entity-demeaned method—that is, all values in each state’s variable value is subtracted by the mean values of the variable across periods in the same state. Time-fixed effect is removed by time-demeaned method—that is, each year’s variable value is subtracted by the mean value of that variable across states in the same year (Stock & Watson, 2007). Heteroskedasticity adjusted t-statistics are in parentheses. Total observations are 288. The results shown in this table are from the model in which democratic governor is the political variable. The results from the models in which democratic senate or democratic house are the political variables are not different from the results in this table. The results are available upon request.

p < .10. **p < .05. p < .01.

For example, the first row represents the response of state-highway investment from the current year (EXP_t) to the past year’s investment (EXP_{t − 1}), past year’s highway demands (PI_{t − 1}, LMR_{t − 1}, LMR_{t − 1}), and past year’s internal-revenue capacity (EAR_{t − 1}, FED_{t − 1}, GEN_{t − 1}, and BOND_{t − 1}). Each column presents a list of variables in the model picking up the effect of each individual variable. For example, column 1 indicates that two variables, current year expenditure (EXP_t) and rural lane miles (LMR_t), pick up the effects of past year’s investment (EXP_{t − 1}), while the rest of the variables do not. The results in column 1 also indicate that public investment does not significantly increase PI or the government revenue base (EAR, FED, GEN, BOND) within a year. The results in the same column also suggest that highway investment does not significantly influence vote on democrat governor at least within a year after investment spending. The main results are discussed below.

First, unlike the previous results estimated by the static framework, the results in the first row indicate that EAR and federal grants do not significantly explain state-highway investment in a dynamic setting, especially when the portion of general revenues used in highway spending are controlled for.² Furthermore, general revenue strongly and significantly influences state-government investment: a US$1 increase in general revenue transferred to finance highway investment decreases highway investment by about 65 cents. The negative sign on the coefficient of general revenue indicates that highways and other government programs marginally financed by general revenue are competing for common-pool resources: the higher reliance on general revenues to finance highways, the lower the spending on highways for the future.

Second, as presented in column 7 of the table, state-highway investment (EXP_t) does not pick up federal-grant effects, but PI (PI_t) and general revenues (GEN_t) do so. A US$1 increase in federal-grant results in a US$6 increase in per capita PI in the following year. A US$1 increase in federal-grant results in an 18 cent decrease in the use of general-revenue fund for highway investment in the following year. These results suggest fiscal-substitution roles of federal grants.

Third, as presented in column 6, earmarked highway revenue does not constrain highway-investment spending (fiscal-constraint hypothesis) but substitutes fiscal resources in the general-revenue fund (fiscal substitution). The results in this column indicate that highway investment does not pick up the effect of EAR, but the general-revenue fund does. A dollar increase in EAR results in a 15 cent decrease in diverting resources from the general-revenue fund to the highway fund in the following year, everything else equal. This is evidence supporting the fiscal-substitution hypothesis for the dynamic framework.

Last, as shown in the second row of the table, per capita income responds directly to highway stocks, both in rural and urban areas, as well as federal grants. A 1,000 per capita lane mile increase in urban areas will result in about a US$964 dollar increase in economy, while a 1,000 per capita lane mile increase in rural areas will withdraw the resources from state economy by about US$1,682. This result is possible given that infrastructure stocks in urban areas will serve a greater number of people than when they are located in rural areas.

Conclusions

This article investigates the effect of internal-revenue capacity, measured through inertia-fiscal variables, on state-highway investment, controlling for the lagged effects of highway demands and government-partisan ideology. The theoretical framework is that in a dynamic framework, a state government will invest according to its internal-revenue capacity forecasted through the baseline budget, existing tax laws, and other constitutional and legal commitments. Since the lagged revenues interact against one another, they not only determine future revenue, but also future expenditure size, and as a result, altering the amount of major revenues used to finance highway programs from year to year. This is because the state government shuffles resources at the marginal level to plan for future balanced budgets. The main argument is that when other revenue sources are included in the model, the roles of EAR and federal grants diminish because EAR and federal grants are substitutable. Furthermore, these two revenue sources, in a dynamic framework, will be inversely related to general revenues since they are fungible. It should be noted that this argument is specifically for state-investment behaviors in a dynamic framework, where the investment response to each fiscal factor is observed moment by moment on the assumption that state governments divert fiscal resources to comply with balanced-budget rules and that they need to pay highway costs upfront prior to receiving the federal grants as reimbursement once the project is completed.

The VAR results provide evidence that state-highway investment is both revenue- and demand driven, but not politically driven. Earmarked highway revenue is used to substitute the resources in the general-revenue funds that would otherwise be appropriated. The same is true for federal grants: federal aid (in terms of reimbursement) is used to substitute the resources in the general-revenue funds that would otherwise be appropriated for highway programs. Thus, the VAR results do not support fiscal constraints or the flypaper effects for EAR and grants, respectively, as the static results do. Furthermore, there is resource competition between highway programs and other public programs financed by general revenue since the empirical results indicate that for every US$1 of general revenue transferred, highway investment is reduced by more than half (65 cents). Thus, the implication is that since there are diminishing returns in using general revenues for financing highways, state government may consider reforming their earmarking policies and tax rates to regain a good alignment between users’ benefits and highway costs. In addition, federal government may reconsider its budget institutional practices—that the grants award credits rather than cash to the state government. This may be the reason that, in a dynamic framework, federal grants exhibit an income effect on state-government spending and such effects are delayed. Finally this study is somewhat limited in that the testing model does not include exogenous variables. Future study should extend the testing model by incorporating exogenous variables in the PVAR model.

Footnotes

Acknowledgements

I wish to acknowledge Inessa Love and Lea Zicchinoan for providing me the codes of PVAR and anonymous referees for the valuable comments and suggestions to improve this paper. I also thank Dr. Kenneth A. Kriz who encouraged me to utilize PVAR as an appropriate analytical tool for this type of study. Of course, any error that remians is my responsibility.

Declaration of Conflicting Interests

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

Funding

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

Notes

Author Biography

Arwiphawee Srithongrung, DPA, is an assistant professor at the School of Public Administration, University of Nebraska at Omaha. Her research interests include public capital budgeting, economic development, state-local and international fiscal policies, and transportation finance.

References

Becker

(1996). The illusion of fiscal illusion: Unsticking the flypaper effect. Public Choice, 86, 85-102.

Beltramini

R. F.

Chapman

P. S.

(2003). Do customers believe in automobile industry rebate incentives? Journal of Advertising Research, 43, 16-24.

Borg

M. O.

Mason

P. M.

(1988). The budgetary incidence of a lottery to support education. National Tax Journal, 41, 75-85.

Borg

M. O.

Mason

P. M.

Shapiro

S. L.

(1993). The cross effects of lottery tax on alternative state tax revenue. Public Finance Review, 21, 123-140.

Browning

(1975). Collective choice and general fund financing. Journal of Political Economy, 83, 377-390.

Buchanan

(1963). The economics of earmarked taxes. Journal of Political Economy, 71, 457-469.

Congleton

R. D.

Bennett

R. W.

(1995). On the political economy of state highway expenditures: Some evidence of the relative performance of alternative public choice models. Public Choice, 84, 1-24.

Deran

(1965). Earmarking and expenditures: A survey and a new test. National Tax Journal, 25, 223-228.

Dye

R. F.

McGuire

T. J.

(1992). The effect of EARs on the level and composition of expenditures. Public Finance Review, 20, 543-558.

10.

Eberts

Fox

(1992). The effects of federal policies on local infrastructure investment. Public Finance Review, 20, 557-571.

11.

Evans

W. N.

Zhang

(2007). The impacts of EAR on K-12 educational expenditure. Education Finance and Policy, 2, 40-73.

12.

Gamkhar

Oates

(1996). Asymmetries in the response to increases and decreases in intergovernmental grants: Some empirical findings. National Tax Journal, 49, 501-512.

13.

Ghamkar

(2000). Is the response of state and local highway spending symmetric to increases and decreases in federal highway grants? Public Finance Review, 28, 3-25.

14.

Ghamkar

(2003). The role of federal budget and trust fund institutions in measuring the effects of federal highway grants on state and local government highway expenditure. Public Budgeting and Finance, 239(1), 1-21.

15.

Goetz

(1968). Earmarked taxes and majority rule budgetary process. American Economic Review, 58(1), 128-136.

16.

Heyndels

Van Driessche

(1998). Mental accounting in local public sector budgeting: An empirical analysis for the Flemish municipalities. Eastern Economic Journal, 24, 381-394.

17.

(2008). State highway capital expenditure and the economic cycle. International Journal of Public Administration, 31, 101-106.

18.

Holtz-Eakin

(1991). Bond market conditions and state local capital spending. National Tax Journal, 44, 105-120.

19.

Jung

(2002). The effect of local earmarking on capital spending in Georgia counties. State and Local Government Review, 34(1), 29-37.

20.

Kim

H. M.

(2006). Consumers’ responses to price presentation formats in rebate advertisements. Journal of Retailing, 82, 309-317.

21.

Knight

(2000). The flypaper effect unstuck: Evidence on endogenous grants from the federal highway aid program. Washington, DC: Federal Reserve Board.

22.

Knight

(2002). Endogenous federal grants and crowd-out of state government spending: Theory and evidence from the Federal highway Aid Program. American Economic Review, 92(1), 71-92.

23.

Love

Zicchino

(2006). Financial development and dynamic investment behavior: Evidence from panel VAR. Quarterly Review of Economics and Finance, 46, 190-210.

24.

Merrifield

(2000). State government expenditure determinants and tax revenue determinants revisited. Public Choice, 102(1-2), 25-50.

25.

Meyers

H. G.

(1987). Displacement effects on federal highway grants. National Tax Journal, 40(2), 221-235.

26.

Michael

(2008). Earmarking state tax revenues. St. Paul: Minnesota House of Representatives Research Department.

27.

Miller

D. E.

Pierce

P. A.

(1997). Lotteries for education: Windfall or hoax? State and Local Government Review, 29(1), 34-42.

28.

Nesbit

Kreft

(2009). Federal grants, EARs, and budget crowd-out: State highway funding. Public Budgeting and Finance, 29(2), 94-110.

29.

Netzer

(1992). Differences in reliances on user charges by American state and local governments. Public Finance Review, 20, 449-511.

30.

Ong

B. S.

(2008). The impact of consumer perceptions of, and attitudes toward mail-in rebates on effectiveness of rebates. Journal of Promotion Management, 14(1-2), 45-58.

31.

Phelps

(2001). Real and monetary determinants of state and local highway investment, 1951-66. American Economic Review, 59, 507-521.

32.

Puentes

Prince

(2005). Fueling transportation finance: A primer on the gas tax. In Katz

Puentes

(Eds.), Taking the high road: A metropolitan agenda for transportation reform (pp. 45-76). Washington, DC: Brooking Institution.

33.

Reed

W. R.

(2006). Democrats, republicans, and taxes: Evidences that political parties matter. Journal of Public Economics, 90, 725-750.

34.

Rubin

(2008). The politics of public budgeting (7th ed.). Thousand Oaks, CA: Sage.

35.

Ryu

J. E.

(2006). Fiscal impacts of an innovative highway financing alternative on state highway expenditure: The case of federal assistance funds in the state infrastructure bank (SIB) programs. Journal of Public Works Management and Policy, 11(1), 33-48.

36.

Spindler

(1995). The lottery and education: Robbing Peter to pay Paul? Public Budgeting and Finance 15(3), 54-62.

37.

Stanley

M. T.

Block

S. B.

(1984). A survey of multinational capital budgeting. Financial Review, 19(1), 36-54.

38.

Stock

Watson

(2007). Introduction to econometrics (2nd ed.). Boston, MA: Addison-Wesley.

39.

Stotsky

J. G.

(1991). State fiscal responses to federal government grants. Growth and Change, 22(3), 17-31.

40.

U.S. Department of Transportation, Federal Highway Administration. (2001). Highway Statistics 2001. Retrieved June 11, 2010, from http://www.fhwa.dot.gov/ohim/hs01/

41.

U.S. Department of Transportation, Federal Highway Administration. (2007). Financing Federal-aid Highways. Retrieved June 15, 2010 from http://www.fhwa.dot.gov/reports/financingfederalaid/index.htm

42.

U.S. Department of Transportation, Federal Highway Administration. (2010). Highway Statistics Series. Retrieved June 11, 2010 from http://www.fhwa.dot.gov/policyinformation/statistics.cfm

43.

Wachs

(2005). Improving efficiency and equity in transportation finance. In Katz

Puentes

(Eds.), Taking the high road: A metropolitan agenda for transportation reform (pp. 77-100). Washington, DC: Brooking Institution.