Coveting Thy Neighbors’ Taxation Coveting Thy Neighbors’ Taxation

Abstract - Recent research has shown that a state’s overall burden is dependent on that of neighboring states. By disaggregat- ing a state’s tax burden into its individual components, this paper demonstrates that during the period of 1967–1996, state with a mobile tax base had positive response rates as high as 60 percent. Thus, a 10 percent increase in neighboring states’ rates was met by an increase of up to 6 percent in the home state’s rate. Taxes with relatively immobile tax bases exhibit negative responsiveness, mean- ing that states respond to rate increases in neighboring states by decreasing home rates.

INTRODUCTION

“Give me $3 worth so I can get to Maryland.” —a gasoline customer to a local service attendant in the District of Columbia (Richburg, 1980).

he quote above describes what was a common story in Tthe District of Columbia after the city implemented a 6 percent gasoline tax in August of 1980. The tax raised gaso- line prices by an average of 8 cents a gallon, causing many residents to cross the border into Maryland, where prices were lower. In the first 26 days of the tax’s existence, gasoline sales fell 27 percent in the District, resulting in lower revenues than had been expected (Richburg, 1980). The tax was repealed soon thereafter, replaced with a lower one–cent tax, which had a minimal impact on differentiating gasoline prices between the District and Maryland. This historical example shows that the results of state policy can be impacted by actions that occur beyond the border. This experience, however, is not limited to the realm of gasoline taxes. New Hampshire, for example, has state sanctioned li- Jonathan C. Rork quor stores located along Interstate 95 immediately over its Department of borders with both Massachusetts and Maine, and has re- Economics, minded Bostonians on local radio that the state has no sales Vassar College, tax as well. People are well aware of the lack of personal in- Poughkeepsie, NY come taxation in Florida promoted by local authorities. 12604 To date, most empirical research on has focused on the interdependence of property taxes set by lo- National Tax Journal cal governments. The anecdotal evidence above is sugges- Vol. LVI, No. 4 tive that tax competition is also alive at the state level. This December 2003 paper adds to the literature by estimating the degree of inter- 775 NATIONAL TAX JOURNAL dependence among five of the major tax AFDC expenditures are not the only area instruments used by state governments. in which states may compete in expendi- We find that for the time span 1967–1996, tures, however. Case, Hines, and Rosen state taxes on motor fuel, tobacco, and cor- (1993) found that a one–dollar increase in porate income, all of which have mobile neighboring state expenditures would raise tax bases, respond positively to rates set own–state expenditures by 70 cents. These in neighboring states. Hence, in- results remained strong once expenditures creases in neighboring states are matched were disaggregated into separate categories by a rate increase in the home state. How- such as education and highways. Using a ever, state taxes with relatively immobile different time span, Baicker (2001) finds bases, such as the personal and that a one–dollar increase in expenditures general , exhibit negative respon- will result in anywhere from a 33 to 88 cent siveness. This means that decreases in increase in home expenditures depending neighboring tax rates were matched by an on the definition of neighbor used. increase in the rate of the home state. Another large area of work has focused The remainder of the paper proceeds as on the adoption of various programs and follows. We begin by discussing earlier regulations. Alm, McKee and Skidmore literature on interstate competition in the (1993) found that after 1980, an important second section. The third section intro- determinant of whether a state had duces a model highlighting the role of adopted a lottery was whether or not a capital mobility within interstate tax com- neighboring state had already imple- petition. After describing the data, we ex- mented one. Anderson and Wassmer plain the estimation procedure and dis- (1995) estimated hazard rates for the adop- cuss the results. The last section summa- tion of tax abatements among Michigan rizes our main conclusions. municipalities and found positive duration dependence. The authors used this evi- PREVIOUS RESEARCH dence to conclude that cities were involved in a prisoner’s dilemma where abatements There are three main areas in which state were offered as a way of evening the play- policy may respond to the actions of other ing field. Brueckner (1998) performed a states: expenditure policy, program adop- similar study on the adoption of growth tions and taxation. On the expenditure side, control measures among California towns an early area of work focused on the deter- and found that cities were more likely to mination of state benefits for the program impose growth controls if their neighbors Aid to Families with Dependent Children had also done so. Moreover, these cities (AFDC). Studies by Gramlich and Laren implemented growth controls of the same (1984), Peterson and Rom (1989), and Smith intensity level as their neighbors. (1991) found support of a race to the bot- Fredriksson and Millimet (2002) found that tom: states responded to benefit cuts in in the case of state environmental policies, neighboring states with cuts of their own. states would match the stringency levels These studies excluded the cases of n–state of their policies to those of their neighbors. competition either by assuming neighbor With regards to the empirical work on benefits to be exogenously determined tax competition, most work has focused from own–state benefits or by looking at on property taxes. Heyndels and Vuchelen only bilateral interactions. Saavedra (2000), (1998), Brueckner and Saavedra (2001) and using techniques from the spatial econo- Revelli (2001) found evidence of property metrics literature to account for both these tax competition among municipalities in weaknesses, confirmed the finding of state Belgium, metro Boston, and the United interdependence of AFDC benefit setting. Kingdom respectively. Outside of local 776 Coveting Thy Neighbors’ Taxation property taxes, Buettner (2001) illustrated to the structural trends in state taxation competition in Germany with regards to we now witness.2 the tax local jurisdictions could place on business earnings. Finally, Besley and TAX COMPETITION FOR STATES Rosen (1998) showed that states would respond to a change in federal gasoline The Basics taxes by raising state gasoline taxes. The mobility of capital plays a vital role Goodspeed (2000, 2002) discovered a simi- in models of tax competition.3 In these lar responsiveness between national and models, capital flows between jurisdic- local income taxes among European states. tions as it searches for the highest after tax Case (1993) and Besley and Case (1995) rate of return. One way that a government argue that voters look to their neighbor- can encourage capital inflow, therefore, is ing states in order to get a sense of cur- by lowering its tax on capital to a level rent economic conditions. If neighboring below that of its neighbors. If the response states are changing their tax rates, voters of capital is elastic in nature, the jurisdic- will be more accepting of a tax change tion will increase overall revenue despite because national trends seem to dictate the lower tax rate. such a reaction. Governors are well aware The problem with this logic, however, is of this and are consequently more likely that there is nothing to prevent other juris- to adjust their tax rates when neighbors dictions from undercutting the first tax 1 are adjusting theirs. In empirically test- change. Under the assumption of perfect ing this conjecture, Besley and Case use a mobility, meaning capital can traverse bor- tax liability measure that is an aggregate ders costlessly, the only equilibrium that of sales, corporate and personal income can be maintained is the race to the bot- taxes paid as their measure of the tax rate; tom scenario in which both jurisdictions hence the authors say nothing about spe- have tax rates of zero. Should a positive cific tax rates in a given state. rate be chosen by one jurisdiction, the other This paper disaggregates the Besley and will respond by lowering their rate by ep- Case tax measure into its individual com- silon, thereby attracting all the capital. ponents, illustrating that certain state The assumption of perfect mobility is taxes, especially excise taxes, are very problematic, however, as some capital in- much susceptible to competition. This curs a cost of relocating. Hence, lowering finding may help explain why the relative a tax rate by a tiny bit may not entice capi- importance of excise taxes within state tax tal movement, as the tax savings will not structures has declined over time. For ex- offset the costs capital owners incur by ample, if states want forecasts of tax rev- relocating. This effectively prevents juris- enue to be accurate for purposes of bud- dictions from facing the “all or nothing” getary planning, they may lower their re- scenario above, as some capital will not liance on competitive tax sources, leading find moving advantageous.4 There is still

1 This notion of yardstick competition is a second avenue for state interdependence. For politicians, voters’ perception of a change in state taxation is likely to be as important as the fear of budgetary repercussions from direct competition. While the competition aspect should primarily affect mobile resources, the political aspect can affect immobile resources as well. Hence, different taxes may be affected differently. 2 For a discussion of the structural tax changes that have occurred within the American South, see Rork (2003). 3 Models of tax competition are obviously richer than the abbreviated version we describe in this section. See Wilson (1999) for a thorough survey of the tax competition literature. 4 The lack of perfect mobility can also result in the granting of tax holidays, under which capital owners are offered a large tax break to entice them to locate in a particular state. The state then slowly raises tax rates over time, knowing that capital will be unable to recoup the costs of relocating again. See Doyle and van Wijnbergen (1994) and Bond and Samuelson (1986) for a more thorough discussion. 777 NATIONAL TAX JOURNAL downward pressure on tax rates, but be- Because people are often tied down to cause all capital is not responsive, the a particular state because of employment, equilibrium will result in positive tax rates one does not expect movement from one for both jurisdictions. state to another on the basis of a change in the personal income tax. While the gain Mobility and State Level Competition from such a move is lower income tax payments, these savings are likely offset In thinking about competition in differ- by the costs of moving and the possibility ent forms of state taxation, the key com- of a longer commute, which would in- ponent remains mobility, although now crease both the time and monetary the mobility is that of the corresponding costs of employees.5 Firms are perhaps tax base. If a tax is placed on a specific more likely to move between states than good, for instance, consumers may re- residents, but a firm’s mobility is also lim- spond by searching out the area with the ited by relocation costs that can range lowest tax–inclusive price. As the District from those associated with finding new of Columbia example illustrated, people employees to building new factories. are willing to cross the border for lower Thus, a competitive effect may exist in cor- gasoline prices. It seems unlikely, how- porate taxation, but it is likely muted ever, that residents would consistently do when compared to excise taxes, for ex- all their shopping in another state to take ample. advantage of a difference in a broader This discussion suggests that the degree sales tax. Hence, a general sales tax may of mobility of the targeted tax base will exhibit less responsiveness to changes in determine the degree of competitiveness, neighbors’ tax rates than an excise tax on and hence excise taxes appear to be better a specific good. Essentially, the argument candidates to exhibit tax competition than boils down to the demand for a specific broad based tax instruments. This is borne good (e.g., cigarettes, gasoline) being more out by Table 1, which reports the percent- elastic than the demand for consumption age share of for each of the in general. five tax instruments explored in this

TABLE 1 AVERAGE SHARES OF STATE TAX REVENUE BY SOURCE AND YEAR Tax Source 1967 1972 1977 1982 1987 1992 1996 30–Year Average Motor Fuel average 17.51 14.48 10.88 8.17 8.10 8.33 7.67 10.44 maximum 37.58 23.02 20.00 18.40 14.92 14.69 14.01

Tobacco average 5.34 5.02 3.76 2.65 2.17 1.88 1.82 3.30 maximum 13.60 17.27 13.50 7.98 5.68 4.58 5.26

Corporate Income average 5.75 6.05 7.76 7.37 7.28 5.66 6.55 6.70 maximum 17.52 13.24 16.76 24.54 27.00 15.34 21.51

Personal Income average 13.58 18.89 22.18 25.49 27.35 29.97 30.44 23.78 maximum 47.51 49.41 57.76 62.37 65.41 67.04 63.93

General Sales average 26.67 29.50 30.71 31.11 32.48 31.89 31.85 30.87 maximum 54.90 51.74 55.86 54.41 58.24 58.40 58.40 SOURCE: State Government Finances, 1967–1996.

5 This assumes the individual remains employed at the same location. 778 Coveting Thy Neighbors’ Taxation study.6 Table 1 reports the highest share There are two goods in the model: X, a across all states for the year indicated, as private good, and G, the per–capita level well as the average of all 48 mainland of a public good provided by the state. states for that year. The theory of tax com- Assuming X and G are additively sepa- petition predicts that if competition is rable yields utility in the form of: present, revenue should decline over time. Thus, instruments with a decreasing share [1] U(X, G) = u(X) + v(G). of overall tax revenue are candidates for finding tax competition.7 Motor fuel and Let y represent an individual’s private tobacco exhibit steady declines in shares before–tax income, and Y represent total throughout the sample in both the aver- private income in state 1. The government age and maximum share, with motor fuel of state 1 finances expenditures through being the most pronounced. Personal in- a proportional income tax, t. Hence: come and general sales tax shares, on the other hand, demonstrate a steady increase [2] X = (1 – t) y. over time, perhaps indicating that as cer- tain state tax revenues are whittled away Individuals are allowed to move be- by competition, states respond by increas- tween states in response to a given tax rate. ing their non–competitive, less responsive This mobility within the tax base implies taxes. It is hard to conclude much from that the state’s total income will depend the movement of the corporate income tax on the tax rates of both states. Thus: share, as it increases and decreases dur- ing the period. However, the steady de- [3] G = tY(t, t*), cline witnessed after the late 1970’s sug- gests that if competition in corporate in- where t* represents the income tax of state come taxation was present, it did not take 2. The problem for the government is to root until that time. choose the tax rate that maximizes the utility of the median voter subject to the budget constraint of the individual [2] and Modeling the Impact of Mobility the government [3]. Substituting the con- In order to illustrate how neighbors may straints into the objective function [1] and influence a home state’s tax choice, we maximizing yields a first order condition of: model the case of a state using a personal income tax8 in a spirit similar to Goodspeed ∂Y [4] –u [y] + ν Y + t — = 0. (2002). For simplicity the paper focuses on x G[ ∂t ] the case of two states with an emphasis on the reaction function of state 1. We assume The term ∂Y/∂t captures the degree of re- residents in both states have single–peaked sponsiveness of the tax base to the tax preferences. Doing so allows us to invoke change. Should residents respond to tax the median voter theorem, which implies changes by relocating, one would expect that the government will satisfy the pref- the sign to be negative, as higher taxes erences of the median voter. would induce people to locate to state 2.

6 We use the tax share measure as a first pass at looking into the erosion of the corresponding base. We ulti- mately focus on the responsiveness of the actual tax rate. 7 A state may lower its tax rate and receive higher revenue, hence resulting in a higher percentage share than seen previously. If there is truly downward competition occurring along this instrument, it should be the case that another state would undercut the first state, resulting in lower revenue. Thus, while an increase in per- centage share is possible, it should be met by an eventual decrease if states are actually competing. 8 The model holds for other types of tax instruments; the personal income tax was chosen for simplicity. 779 NATIONAL TAX JOURNAL

Ultimately, we are interested in the im- is also advantageous because federal gov- pact of a change in state 2’s tax on the ernment initiatives such as Temporary of state 1. This can be achieved Aid for Needy Families (TANF), which by differentiating [4] with respect to t*: have the potential to alter state finances, are not present.10 ∂Y ∂Y ∂Y The financial data was matched with [5] ν Y + t —— t —— + ν —— GG[ ∂t ][ ∂t* ] G[ ∂t* demographic information compiled from The Statistical Abstract of the United States. ∂ 2Y + t —— The end result is a data set containing ∂ ∂ * t t ] thirty years of information for each of the 48 mainland states. The study focuses on As in Besley and Rosen (1998) and five main categories of state tax revenue: Goodspeed (2002), the slope of the reaction personal income tax, corporate income * function with respect to t can be either tax, general sales tax, motor and positive, negative, or zero. Note that if there tobacco tax. These sources account for a is no mobility, equation [5] simplifies to large percentage of all state tax revenue ∂ ∂ * zero as Y/ t will be zero. The remainder in any given year during the sample. of the paper sets out to estimate the slope of the reaction function for state level taxes on different goods and types of income. REGRESSION MODEL Defining the Competition DATA In order to model state tax competition, State financial data for this analysis has one must start by determining which been compiled from various issues of State states are competing with one another. Government Finances for the years 1967– The discussion in the previous section 1996. We start with 1967 since prior to that highlighted the importance of mobility year, various tax data (such as corporate within the tax base, suggesting that geog- income and personal income tax rev- raphy will play a large role in determin- enues) were aggregated together in some ing competitors. For example, individu- states. Using 1967 as a starting point has als from Vermont are not going to drive a nice historical advantage in that federal to Utah to purchase gasoline, but they restrictions on industrial bond issuances may travel to New Hampshire, Massachu- were implemented at the same time. States setts or New York. Thus, bordering states involved in industrial recruitment efforts appear to be the logical starting point.11 were forced to switch from the use of Once the bordering states for each state bonds to the use of as the basis are identified, the next step is to assign of their recruiting efforts. By starting in weights in order to capture the relative 1967, the full impact of this switch in importance one state may wield over an- policy focus is captured.9 Ending in 1996 other. This paper focuses on two common

9 The federal restrictions on industrial bond issuances were implemented in part because the federal govern- ment was losing potential revenue. See Rork (2003) for a further historical discussion. 10 Obviously there are other federal policies, such as the Act of 1986, that will have an impact during the time period being investigated, and the estimation technique will account for these policy changes. 11 There is no reason to limit the definition of neighbors to geography, as non–contiguous states may also be competitors for other reasons. Case, Hines and Rosen (1993), for example, defined neighbors as those states with a similar racial composition. In using a spatial approach, however, there is no way to estimate what these weights would be, leaving the modeler to assign the structure of competition. Rather than impose a structure that may not exist, the paper limits itself to the geographic definition, which is the common definition in most studies. 780 Coveting Thy Neighbors’ Taxation approaches: contiguity weights and popu- and the tax measure of its neighbor. De- lation–contiguity weights. A contiguity noting Tit to be the tax measure in state i weight treats each bordering observation at time t yields a linear relationship of: equally. In the example above, the obser- β θ ξ λ vations from New Hampshire, Massachu- [6] Tit = Xit + Tjt + i + t + uit , setts and New York would get equal weight, as they are assumed to wield where Tjt represents the neighbor’s tax 12 ξ λ equal influence. The observation from rate, i and t are state and year fixed ef-

Utah, on the other hand, would be given fects, respectively, and uit is a mean zero, a weight of zero since Vermont and Utah normally distributed random error. Since do not share a common border. A popula- most states have multiple neighbors, we Σ tion–contiguity weight still assumes that replace Tjt by j wij Tjt, where wij represents the relevant observations are only those the weight assigned to state j. With simple that border the observation. In this case, contiguity weights, Tjt is replaced by the however, weights are based on the over- average tax measure of the bordering states. all populations of the bordering observa- With the population measure, Tjt is replaced tions. Thus, New York would have a by a weighted average of bordering states. higher weight than Massachusetts, which Since each state has a set of weights in turn would have a higher weight than (which can be thought of as a 48x1 vec- New Hampshire. tor), equation [6] can be converted into a As is common in the literature, we use system of equations written as: row–standardized weights, meaning the θ β ξ λ sum of the weights equals one. Thus, in the [7] Tit = WTit + Xt + i + t + ut , case of the contiguity weights for Vermont, each neighboring state would be given a where W is a weighting matrix assigning weight of 1/3, as three states border Ver- neighbors to every state. mont. In the case of population–contiguity The inclusion of Tit on the right hand side weights, one takes the bordering observa- of equation [7] means that estimation by tion and divides it by the sum of the entire OLS will be inconsistent, due to correlation bordering population. Thus, New York’s with the error. We estimate equation [7] weight would be its population divided by using the instrumental variables approach the sum of the populations of New York, outlined in Kelejian and Prucha (1998) in Massachusetts and New Hampshire. which weighted values of the exogenous variables (WX) are used as instruments.13 The Basic Modeling Framework Determination of Exogenous Variables We follow Case, Hines and Rosen (1993) and assume that each state’s tax measure In order to conclude that interstate is a function of state characteristics (Xit) tax competition exists, it is important to

12 The standard approach is to assume tax rates are determined contemporaneously, meaning a state’s tax rate in a given year depends on the tax rate of its neighbors in that same year. One could argue that the neighbors’ rate from last year is more appropriate. While the estimation procedure used in each case is different (using a contemporaneous choice of tax rates involves the need to control for endogeneity, whereas using lagged tax rates does not), the results do not change substantially. Thus, we report results from the standard contempo- raneous approach. 13 Note that an alternative approach is to invert the system and estimate via maximum likelihood. However, should spatial error dependence also be present, maximum likelihood may provide false evidence of strategic interaction. (See Brueckner (2003) for a discussion.) While results remain unchanged using maximum likeli- hood, in the paper we present the instrumental variable technique since it has been shown (Kelejian and Prucha, 1998) to yield consistent estimates, even in the presence of spatial error dependence. 781 NATIONAL TAX JOURNAL control for other factors that may precipi- may be willing to adjust their tax rates in tate changes in a state’s tax rate. These order to encourage job creation through factors fall into three broad categories: fis- firm relocation. Per capita income by state cal, political and demographic. We discuss is included as it measures the general well each of these categories in turn. being of a state’s residents while also con- The first category is what Alm et al. trolling for potential income effects in con- (1993) describe as fiscal stress. Under this sumption. The percent of a state’s popula- scenario, a state is forced to change its tax tion that is elderly (over age 65) is included rate in order to make up for an unforcasted because this represents a vocal and active budget shortfall. To control for this sce- population within a state’s electorate. nario, measures of per capita outstanding As mentioned above, the final specifica- debt, as well as per capita federal trans- tion also includes year and state fixed ef- fers, have been included. If federal trans- fects. The time effect controls for variables fers decrease, a state may have to raise that affect all states in a given year, such as taxes in order to make up for the foregone business cycle conditions. This will also income. If a state has too much outstand- address federal tax changes, which Besley ing debt it may have to raise income in and Rosen (1998) and Goodspeed (2002) order to avoid a bond ranking crisis.14 found to play a role in state tax determina- A second argument is that the political tion. State fixed effects are included to con- environment plays a role in the determina- trol for unobserved state characteristics, as tion of tax rates (Besley and Case, 1995). To well as history. For example, New Hamp- control for this possibility, three variables shire voters have a mindset of opposing are included. The first is a dummy variable sales taxation, whereas Oregon and Wash- for whether or not the year of observation ington have historically adopted different was an election year in the state. Politicians tax structures. Including state fixed effects may be less likely to raise a tax in the face allow us to control for these important fac- of an election, and may be inclined to offer tors in the analysis. tax reductions if re–election was uncertain. We also include dummy variables indicat- Measuring the Tax ing whether or not the governor and the majority of the state legislature are from the Finally, the dependent variable in the same party, as tax changes may be easier to analysis is the state tax rate for each of the implement if all political bodies are from tax categories. The statutory rate for sales the same political party since an air of co- taxation (in terms of percentages), gasoline operation is more likely. Moreover, with taxation (in terms of cents per gallon) and this variable we make the distinction be- cigarette taxation (in terms of cents per tween all political bodies being Democratic package) are used. For both the corporate and all political bodies being Republican, income tax and personal income tax, an as party affiliation may signal one’s will- average tax rate was created.15 For corpo- ingness to change taxes. rate income, corporate income tax revenues Finally, state descriptive variables are were divided by adjusted state GSP, which also included. A state’s unemployment rate is simply state GSP with government expen- is added since states with high employment ditures and personal income subtracted.16

14 Massachusetts faced this situation in the early 1990’s. 15 Accounting for the deductibility of the state income tax at the federal level would make the measure closer to an effective tax rate, but given that this would involve additional assumptions on household structure and income, the average tax measure was used instead. 16 Ideally we would also like to net out investment income for corporations, but that data is unavailable at the state level. 782 Coveting Thy Neighbors’ Taxation

In the case of personal income, the mea- increase of ten cents in neighboring states’ sure was all revenues collected from the cigarette taxes would induce a home in- personal income tax divided by personal crease between four and six cents. The income in the state.17 results show cigarette taxes have a posi- tive correlation with state per capita in- RESULTS come and a negative correlation with the percentage of the population that is eld- Table 2 reports results of the instrumen- erly. There is some evidence of cigarette tal variables estimation for each of the five taxes being substitutes for federal trans- tax rates described in the previous section. fers, although the significance of the coef- Note that we include two sets of estima- ficient on per capita transfers is low. Both tion results for each tax. The first column political parties are associated with higher for each tax reports results using the con- cigarette taxes, although neither effect was tiguity weights. The second column re- statistically different from zero. ports estimates using population weights. Note that in general, the choice of weight- Gasoline Taxation ing scheme brings little change to the re- sults. We discuss each tax in turn. We The results for per gallon gasoline taxes tested the validity of the instruments for (in cents) are reported in column (2) of all five taxes under both weighting Table 2. As was the case for cigarette taxes, schemes. Although not reported in the gasoline taxation appears to have a com- text, both the Basmann and Sargan N*R2 petitive element. The measure of compe- overidentifying restriction tests support tition is positive and statistically signifi- the choice of using weighted values of the cant under both weighting schemes, rang- exogenous variables as instruments for all ing between 0.463 for the population specifications. weights to 0.600 with the contiguity weights. A large component of federal Cigarette Taxation transfers to states is money earmarked for highway improvements. Since state mo- Column (1) in Table 2 contains the re- tor fuel tax receipts are also often targeted sults using the per package state tax (in for infrastructure improvements, the cents) imposed on cigarettes by states. negative relationship between gasoline Under both weighting schemes, the mea- taxes and federal transfers indicates a de- sure of competition is statistically signifi- gree of substitutability between the two cant, with the coefficient ranging from revenue sources.18 The positive correlation 0.416 under population weights to 0.636 with state debt suggests states are willing with contiguity weights. Thus, an average to turn to gasoline taxes in a time of fiscal

17 In most models of tax competition, all capital in the state is assumed to face the same tax rate as there is no differentiation made between types of capital for tax purposes. Thus, the only method available to a state for altering its tax base is either to raise or to lower its tax rate. For certain types of taxes, such an assumption does not hold. For example, many states can alter their sales tax base by exempting necessities such as food and clothing. Similarly, changing the limits on income brackets allows states to change the tax base of income taxes. In situations like these, it may be the case that the use of one rate may not be the best means of capturing state competition. Thus, we re–estimated the model using tax shares for sales, personal income and corporate income taxation. Since the tax share measure is based on revenue, it better accounts for all the different man- ners in which a state may alter its tax base. The results using the tax share measure are consistent with those using the single rate. For brevity, we only report the results using the tax rate as the dependent variable. 18 The recent experience in Massachusetts is a prime example. The federal government’s reluctance to finance further cost overruns of the Big Dig has resulted in the Massachusetts legislature contemplating an increase in gasoline taxes to finance this shortfall. 783 NATIONAL TAX JOURNAL

0.165*** (0.055)

1.525** (0.635)

–0.015*** (0.005)

0.010 (0.009)

0.021** (0.009)

–0.014 (0.012)

–0.0001 (0.0002)

–0.020 (0.061)

0.005 (0.004)

0.150 (0.194)

(5)

INCOME

CORPORATE

0.163*** (0.050)

2.328*** (0.615)

–0.019*** (0.005)

0.008 (0.008)

0.022** (0.009)

–0.014 (0.012)

0.001 (0.003)

–0.034 (0.060)

0.003 (0.004)

0.172 (0.181)

Contiguity Population

0.012

–3.931

–0.047

(0.048)

(3.338)

(0.020)

(0.038)

(0.040)

(0.047)

(0.012)

(0.233)

(0.233)

(0.777)

0.085**

0.024**

–0.506*

8.806***

–0.164***

–0.107***

–1.300***

(4)

SALES TAX

0.028

0.020*

–3.064

–0.041

–0.047

(0.051)

(3.294)

(0.019)

(0.038)

(0.040)

(0.046)

(0.012)

(0.235)

(0.028)

(0.740)

0.094**

8.449***

–0.237***

–0.087***

–1.205***

Contiguity Population

0.032

0.030

0.001

d errors in parentheses. d errors

0.013*

–0.010

(0.044)

(1.967)

(0.013)

(0.021)

(0.024)

(0.029)

(0.007)

(0.016)

(0.010)

(0.421)

0.027**

–0.017*

–0.756*

8.169***

–0.097**

(3)

INCOME

PERSONAL

TION OF STATE TAX RATES TAX TION OF STATE

0.027

0.030

0.005

0.012*

–0.048

–0.017

–0.672

(0.411)

(0.051)

(1.925)

(0.012)

(0.021)

(0.024)

(0.030)

(0.007)

(0.016)

(0.010)

0.024**

–0.018*

7.069***

Contiguity Population

cent level, standar

ABLE 2

T

0.463*** (0.046)

26.036** (12.343)

–0.164 (0.108)

0.033 (0.140)

0.050 (0.160)

0.087 (0.189)

–0.056 (0.054)

–5.934*** (1.364)

0.325* (0.189)

12.634*** (3.865)

(2)

GAS TAX

0.021

0.039

0.054

0.327*

16.949

–0.017

(0.059)

(0.102)

(0.139)

(0.153)

(0.189)

(0.050)

(1.337)

(0.180)

(3.848)

(11.921)

0.600***

–0.225**

–6.595***

12.844***

Contiguity Population

cent level, * significant at 90 per

INSTRUMENTAL VARIABLE ESTIMA VARIABLE INSTRUMENTAL

0.404

0.758

–0.011

–0.424

–2.459

(0.041) (0.168)

(0.298)

(0.324) fects. (0.487)

(0.106)

(2.422)

(0.210)

(5.484)

(23.474)

0.416*** 1.507***

0.292***

–59.783**

–12.676**

Population

(1)

CIGARETTE TAX

0.711

0.356

(2.50)

–0.112

–0.389

(0.050) (0.177)

(0.294)

(0.310)

(0.489)

(0.098)

(0.143)

(5.819)

–4.442*

(24.107)

0.636*** 1.306***

0.288***

–11.725**

–56.094**

Contiguity

cent level, ** significant at 95 per

ranges from a low of 0.69 for the corporate income tax to a high of 0.90 for the personal income tax. ranges from

2

All regressions contain state and year fixed ef All regressions *** significant at 99 per

The adjusted R

Independent Variable

neighboring states’ tax rate

percent of population percent 65 years or older

state per capita income (in 1000’s)

election year dummy (YES = 1)

same party-democrat

same party-republican

state unemployment rate

state per capita federal transfers (in 1000’s)

state per capita debt (in 1000’s)

constant

weight

tax rate

784 Coveting Thy Neighbors’ Taxation trouble. There is also a positive relation- of competition is negative and statistically ship between gasoline taxes and the per- different from zero. The coefficients range cent elderly, which likely stems from the from –0.237 to –0.164, meaning that a 10 fact that the elderly travel shorter dis- percent increase in neighboring states’ tances by car than other segments of the sales tax rates would be met by an ap- population. proximately 2 percent decline in the home state’s rate. As was the case with the per- Personal Income Taxation sonal income tax, it appears as though states are recognizing the lack of mobility Column (3) in Table 2 illustrates the re- in their sales tax base in order to maintain sults for the personal income tax rate mea- revenues. sure discussed earlier. The coefficient on The estimated coefficients of the control the neighboring states’ personal income tax variables in the specification are as ex- is negative, with estimates ranging from pected. Consumption has a positive in- –0.048 to –0.097.19 This means that a 10 per- come elasticity, hence it makes sense that cent increase in neighboring states’ per- states with high per capita income levels sonal income tax rates would induce a would have lower sales taxes, since the home decrease of less than one percent. The sales tax functions like a . lack of mobility in the tax base may explain Both political parties are associated with this result. Only a small fraction of a state’s higher sales taxes, although the Republi- population may respond to a change in can dummy is never statistically signifi- neighboring personal income tax rates by cant. Finally, sales taxes appear to be an- relocating; most individuals are likely to other substitute for federal money, as remain in their home state of residence. By transfers from the federal government recognizing the inelastic response of its exhibit a negative relationship. residents, states can recover some of the revenue lost to relocation by simply rais- Corporate Income ing personal income tax rates at home. The positive relationship between per- Finally, column (5) of Table 2 reports the sonal income taxation and the percent eld- results for the measure of the corporate erly is consistent with the finding of income tax. The estimated coefficient is Conway and Houtenville (2001), who ar- the same under each weighting scheme: gue that the elderly’s lack of wage income 0.16. Thus, a 10 percent increase in the makes them more likely to lobby for high average corporate income tax rate of one’s personal income taxes in lieu of other neighbors results in an increase of 1.6 per- taxes. While insignificant, the coefficients cent at home. As expected, the political on the political party dummies have their dummies work in opposite directions, expected signs, with Republicans (Demo- with the Democrats raising and Republi- crats) associated with lower (higher) per- cans lowering corporate income taxes, al- sonal income taxes. though only the Democratic dummy is statistically significant. Since most of the General Sales Taxation elderly population are retired, the impor- tance of a corporate presence is lessened. The results for the general sales tax rate As a result, the elderly may lobby for are reported in column (4) of Table 2. Un- higher corporate income taxes to replace der both weighting schemes, the measure taxes that impact them more directly.

19 Note that the personal income tax is the only tax where the choice of weight matters, as the contiguity weights yield insignificant findings. 785 NATIONAL TAX JOURNAL

CONCLUSION Karen Smith Conway, Ronald Fisher, Rob- ert McMillan, Thomas Nechyba, John This paper began by arguing how the Pencavel and two anonymous referees mobility of the tax base can impact whether from this journal. An earlier version of this a particular tax instrument is set competi- paper benefited from seminar participants tively. Results indicate that taxes with rela- at the University of New Hampshire, tively mobile tax bases respond positively Hamilton College and the 2000 meetings to rates set in neighboring states. Response of the National Tax Association. Finally, I rates are found to range from 16 to 64 per- wish to thank Steve Fink for his excellent cent, meaning a 10 percent decrease in research assistance. neighboring states’ tax rates would be matched by a decline of between 1.6 and 6.4 percent in the home state’s tax rate, REFERENCES depending on the tax in question. Taxes Alm, James, Michael McKee, and Mark with relatively immobile tax bases, such as Skidmore. personal income taxation, are found to re- “Fiscal Pressure, Tax Competition, and the spond negatively. Moreover, the intensity Introduction of State Lotteries.” National of this response is muted when compared Tax Journal 46 No. 4 (December, 1993): 463– to the response of mobile taxes, as states 76. are found to have response rates between Anderson, John, and Robert Wassmer. 10 and 24 percent. Thus, a decrease in the “The Decision to ‘Bid for Business’: Munici- rates of neighboring states of 10 percent pal Behavior in Granting would be met by an increase of the home Abatements.” Regional Science and Urban state’s tax rate between 1 and 2 percent. Economics 25 No. 6 (December, 1995): 727– These results suggest that states relying 38. heavily on excise taxes do so at their peril, Baicker, Katherine. as these tax rates and their correspond- “The Spillover Effects of State Spending.” ing revenue streams are most susceptible NBER Working Paper No. 8383. Cambridge, to erosion by neighboring actions. In or- MA: National Bureau of Economic Research, der to ensure stable tax revenue, states will 2001. want to place more emphasis on those Besley, Timothy, and Anne Case. taxes that do not respond as dramatically “Incumbent Behavior: Vote–Seeking, Tax– to competition from beyond the border. Setting, and Yardstick Competition.” Ameri- These taxes will be characterized by a lack can Economic Review 81 No. 1 (March, 1995): of mobility within the tax base. The inelas- 25–45. tic nature of the tax base provides the state Besley, Timothy, and Harvey Rosen. with the ability to raise tax rates and to “Vertical Externalities in Tax Setting: Evi- generate additional tax revenue to offset dence from Gasoline and Cigarettes.” Jour- revenue losses attributable to decreasing nal of Public Economics 70 No. 3 (December, tax rates in neighboring states. Recent 1998): 383–98. trends of increasing state reliance on per- Bond, Eric W., and Larry Samuelson. sonal income and sales taxation while de- “Tax Holidays as Signals.” American Eco- creasing state reliance on excise taxation nomic Review 76 No. 4 (September, 1986): are consistent with this portrayal. 820–26. Brueckner, Jan. “Strategic Interaction Among Governments: Acknowledgments An Overview of Empirical Studies.” Univer- This paper benefited from comments by sity of Illinois at Urbana–Champaign. Rosanne Altshuler, Doug Bernheim, Mimeo, 2003. 786 Coveting Thy Neighbors’ Taxation

Brueckner, Jan. Gramlich, Edward, and Deborah Laren. “Testing for Strategic Interaction Among Lo- “Migration and Income Redistribution Re- cal Governments: The Case of Growth Con- sponsibilities.” Journal of Human Resources trols.” Journal of Urban Economics 44 No. 3 19 No. 4 (Fall, 1984): 489–511. (November, 1998): 438–67. Heyndels, Bruno, and Jef Vuchelen. Brueckner, Jan, and Luz Saavedra. “Tax Mimicking Among Belgian Municipali- “Do Local Governments Engage in Strate- ties.” National Tax Journal 51 No. 1 (March gic Property–Tax Competition?” National 1998): 89–101. Tax Journal 54 No. 2 (June, 2001): 203–29. Kelejian, Harry, and Ingmar Prucha. Buettner, Theiss. “A Generalized Spatial Two–Stage Least “Local Business Taxation and Competition Squares Procedure for Estimating a Spatial for Capital: The Choice of the Tax Rate.” Re- Autoregressive Model with Autoregressive gional Science and Urban Economics 31 No. 2– Disturbances.” Journal of Real Estate Finance 3 (April, 2001): 215–45. and Economics 17 No. 1 (July, 1998): 99–121. Case, Anne. Peterson, Paul, and Mark Rom. “Interstate Tax Competition After TRA86.” “American Federalism, Welfare Policy and Journal of Policy Analysis and Management 12 Residential Choices.” American Political Sci- No. 1 (Winter, 1993): 136–48. ence Review 83 No. 3 (September, 1989): 711– Case, Anne, James Hines, and Harvey Rosen. 28. “Budget Spillovers and Inter- Revelli, Federico. dependence.” Journal of Public Economics 52 “Spatial Patterns in Local Taxation: Tax No. 3 (October, 1993): 285–307. Mimicking or Error Mimicking?” Applied Conway, Karen Smith, and Andrew J. Economics 33 No. 9 (July, 2001): 1101–07. Houtenville. Richburg, Keith B. “Elderly Migration and State Fiscal Policy: “D.C. Sales Fall 27% In August Following 6% Evidence from the 1990 Census Migration Tax.” Washington Post (October 8, 1980): C1. Flows,” National Tax Journal 54 No. 1 (March, Rork, Jonathan. 2001): 103–23. “Getting What You Pay For: The Case of Doyle, Chris, and Sweder van Wijnbergen. Southern Economic Development.” Vassar “Taxation of Foreign Multinationals: A College. Mimeo, 2003. sequential Bargaining Approach to Tax Saavedra, Luz Amparo. Holidays.” and “A Model of Welfare Competition with Evi- 1 No. 3 (December, 1994): 211– dence from AFDC.” Journal of Urban Econom- 25. ics 47 No. 2 (March, 2000): 248–79. Fredriksson, Per, and Daniel Millimet. Smith, Patricia. “Strategic Interaction and the Determination “An Empirical Investigation of Interstate of Environmental Policy Across U.S. States.” AFDC Benefit Competition.” Journal of Urban Economics 51 No. 1 (Janu- 68 No. 1–3 (January, 1991): 217–33. ary, 2002): 101–22. U.S. Department of Commerce. Bureau of the Goodspeed, Timothy. Census. “Tax Competition and Tax Structure in Open State Government Finances. Washington, D.C., Federal Economies: Evidence from OECD Various years. Countries with Implications for the Euro- U.S. Department of Commerce. Bureau of the pean Union.” European Economic Review 46 Census. No. 2 (February, 2002): 357–74. Statistical Abstract of the United States. Wash- Goodspeed, Timothy. ington, D.C., Various years. “Tax Structure in a Federation.” Journal of Wilson, John. Public Economics 75 No. 3 (March, 2000): 493– “Theories of Tax Competition.” National Tax 506. Journal 52 No. 2 (June, 1999): 269–304. 787