Tax structure and state economic growth during the Great Recession

The Social Science Journal 51 (2014) 79–89

Contents lists available at ScienceDirect

The Social Science Journal journal homepage: www.elsevier.com/locate/soscij

Tax structure and state economic growth during the Great Recession Richard V. Adkisson a,∗ , Mikidadu Mohammed b a Department of Economics and International Business, Box 30001, MSC 3CQ, New Mexico State University, Las Cruces, NM 88001-8003, United States b Department of Economics, University of Utah, 260 South Central Campus Drive, Salt Lake City UT 84112-9150, United States

a r t i c l e

i n f o

Article history: Received 28 September 2012 Received in revised form 12 October 2013 Accepted 12 October 2013 Available online 8 November 2013

Keywords: Tax structure Taxes and growth Great Recession

a b s t r a c t Concern about the effect of taxes on economic growth and development in the United States is longstanding. While most studies are concerned with the growth impacts of tax burden, marginal rates, or the impact of a particular tax, there are few works that examine the impact of tax structure in the way it is defined in this work. Here, tax structure is defined as the shares of revenue collected by various taxes. Using a pool of data on the 50 states between 2004 and 2010, this paper explores the relationship between state and local tax structure and growth of real per-capita GDP through the Great Recession centered in 2008. The results are used to generate estimates of the growth impacts of revenue neutral changes in tax shares. © 2013 Western Social Science Association. Published by Elsevier Inc. All rights reserved.

1. Introduction Most residents of the American states are interested in the level of economic activity and rate of economic growth in their state. In partial response, states undertake activities aimed at increasing economic growth or other measures of economic development. As conditions are discussed and policies developed, it is nearly certain that one important topic will be the state’s tax system and its relation to economic activity. Are taxes too high or too low? Is the state tax system too progressive or too regressive? Is the state too reliant on one tax or the other and subsequently business “unfriendly?” The answers to such questions are seldom if ever definitive and perhaps more often based on conventional wisdom and/or ideology than fact (Reese & Rosenfeld, 2001; Mazcrov, 2013). “Anecdotal stories about the U.S. tax code can sometimes have a larger impact on

the policy debate than a stack of statistical studies” (Engen & Skinner, 1996, p. 622). Economic cycles are not equally felt across the fifty American states. Fig. 1 shows the growth impacts of the Great Recession at the national level and the business cycle trough in 2008.1 Fig. 1 also demonstrates that the growth experience of states varies substantially – before, after, and during the recession. The purpose of this paper is to explore the empirical relationship between the structure of state and local taxes and states’ relative economic performance through 2004 and 2010. Economic performance is measured as growth in real per-capita gross domestic product (GDP). The focus is on determining the impacts of tax structure, and in particular, the extent to which marginal, revenue-neutral changes in tax shares relate to state growth rates, through this recessionary period. The null hypothesis is that tax structure

∗ Corresponding author. Tel.: +1 575 646 4988; fax: +1 575 646 1915. E-mail addresses: [email protected] (R.V. Adkisson), [email protected] (M. Mohammed).

1 The National Bureau of Economic Research (NBER) identifies the trough as June 2009 however the annual data provided here masks this. See http://www.nber.org/cycles.html.

0362-3319/$ – see front matter © 2013 Western Social Science Association. Published by Elsevier Inc. All rights reserved. http://dx.doi.org/10.1016/j.soscij.2013.10.009

80

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

Year to year percentage growth of real per capita GDP

10 8

8.27

7.37

6

7.65

6.03

5.98

5.88

4.09

4 2.35

2

1.84

1.7

-1.27

-2

-2.19

-2.07

1.57

0.8

0

-1.6

-4

-4.1

-4.62

-6

-3.78

-6.4

-8 -9.26

-10 -12 2003-2004

2004-2005 U.S. growth

2005-2006

2006-2007

Lowest growth state

2007-2008

2008-2009

2009-2010

Highest growth state

Fig. 1. Range of real per capita GDP growth rates, U.S. and 50 states 2004–2010. Source: Bureau of Economic Analysis

does affect economic growth and performance through the business cycle,2 although the analysis begins with no a priori hypotheses as the nature of the effects. 2. Literature review 2.1. Fiscal impacts on growth and development: empirical models Several researchers have studied the relationship between fiscal policy and economic growth, and both neoclassical and endogenous growth models provide the theoretical foundations for these studies. Barro (1990, 1991), King and Rebelo (1990), Mendoza, Milesi-Ferretti, and Asea (1997), and Lucas (1990) use endogenous growth models to examine both the positive and normative taxation effects. To test predictions of these models with respect to the structure of both taxation and expenditure, Kneller, Bleaney, and Gemmell (1999) classify elements of the government budget into one of four categories: distortionary (taxes on income and property) or non-distortionary (taxes on consumption) taxation, and productive or nonproductive expenditures. They find that shifting revenue away from distortionary forms of taxation and toward nondistortionary forms has a growth-enhancing effect and switching expenditures from productive toward unproductive forms is growth-retarding. Others explore international differences in taxes and tax structure. Koester and Kormendi (1989) use cross-country data to examine the impact of average and marginal rates

2 A gap in data availability limits the choice of beginning year thus limiting the ability to infer long-term growth impacts in the present work.

on the level and growth of economic activities. Rabuska and Bartlett (1985) find that a reduction in marginal tax rates in excess of 50% improves the economies of developing countries. Koch, Schoeman and van Tonder (2005) find that decreased tax burdens are strongly associated with increased economic growth potential in South Africa. In general, cross-country studies on the effects of taxes on economic growth suggest that higher taxes impede growth. Miller and Russek (1997) find that for developing countries, debt-financed increases in government expenditure retard growth and tax-financed increases stimulate growth, while the opposite is the case for developed countries. Lee and Gordon (2005) examine the effects of statutory corporate tax rates on the growth of per capita gross domestic product (GDP), using a crosssection data set of countries and find that increased corporate tax rates retard future growth rates within countries. Similarly, Arnold (2008, 2011) observes 21 Organization for Economic Co-operation and Development (OECD) countries and finds property taxes to be the most growth-friendly, followed by consumption taxes and then by personal income taxes. Corporate income taxes appear to have the most negative effect on GPD per capita. Dye and Feiock (1995) study the impact of state personal income tax adoption on growth and find a negative relationship, although most changes are attributable to national conditions. Boskin (1988) analyze the effects of tax changes in the 1980s and draws several conclusions but finds no ‘smoking gun’ with respect the impact of tax policy on national growth. Phillips and Goss (1995) conduct a meta-analysis of empirical papers that examine the impact of state and local taxes on development. While they find evidence to suggest that tax policy can impact state

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

economic development/growth, serious questions remain as to the size of the effects. Obviously there is a substantial literature devoted to examining the effect of fiscal variables on economic growth and development. Few of these are specifically focused on tax structure. However, structure is important. As noted by Gentry and Ladd, (1994, p. 748), “economists have devoted considerable attention to the characteristics of individual taxes, but little attention to the broader question of the appropriate mix of taxes within a governmental jurisdiction.” The papers reviewed below address this concern. 2.2. Tax structure and growth Although there is great variation in method, several authors explore the impacts of tax structure across states and nations. Hettich and Winer (1984) provide a theoretical explanation for variation in tax structure concluding that tax structure decisions are driven by the decisions of self-interested politicos. Their empirical work defines tax structure as the share of revenue raised through the personal income tax share. Mullen and Williams (1994) refer to marginal and average tax rates when studying tax structure. Luna and Murray (2010) focus on the structural characteristics of state corporate taxes when explaining the choice of business organizational form. As in the current work, Howard (2003) defines tax structure as shares of revenue and finds a statistical relationship between a state’s tax portfolio and its economic well-being but admits several limitations to his study. In many industrialized countries, OECD countries at least, certain categories of taxes, such as property taxes positively impact growth. Other categories, such as personal taxes, negatively impact growth. Drawing on Creedy and Gemmell (1982, 1984), Gemmell (1985) argues that, given the tax systems of most European countries, there is an inherent tendency for the tax ratio, defined as the ratio of total tax revenue to national income, to rise simultaneously with an increase in the share of income tax revenue relative to that from consumption taxes. The rise in the total tax ratio and the increase in the share of tax revenues from different tax categories negatively impact growth. Utilizing neoclassical, optimal growth, and overlapping generation models, Bhattarai (2010) finds empirical support based on rank correlation and panel regression analysis to suggest that OECD countries with high tax GDP ratio generally have lower growth rates relative to those with a lower ratio. Arnold et al. (2011) examines the question of how to design tax policy that both speeds recovery from economic crisis and contributes to long-term economic growth. Using a panel of 21 OECD countries over 34 years, the analysis focuses on tax structure, such as tax mix or rates and bases of different taxes, rather than levels as measured, for instance, by the overall tax-to-GDP ratio. Their analysis suggests that economic growth can be increased by gradually shifting the tax base toward consumption and immovable property. It also argues that improvement in the design of individual taxes can be growth-enhancing. Comparing the Austrian tax structure against the tax structure in countries with the highest per capita GDP and growth rates, Pesendorfer (2008) finds that high level of

81

labor income taxes negatively affects the growth potential in Austria. Mamatzakis (2005) examines how, over time, output growth responds to shocks in the tax mix and the tax burden in Greece. He suggests that output growth responds negatively to an increase in tax burden, defined as the sum of direct and indirect taxation over output, while tax mix, defined as the ratio of indirect over direct taxation, indicates a positive impact on output growth. Similar to the present work, Widmalm (2001) examines the tax structure and growth relationship across 23 OECD countries to answer the question: are some taxes better than others? Among Widmalm’s conclusions are that overall tax burden is negatively correlated with economic growth and taxes on personal income have an especially negative effect, but consumption tax is growth enhancing. Nelson (1989) considers whether energy producing states substitute energy-related revenue for revenue from other forms of taxation and whether these shifts have a detectable impact on economic development. He analyze a 1963–1983, three-state panel of data and finds that tax switching occurs and reduces reliance on other taxes attracts inflows of capital and labor to the states. Phillips and Goss (1995) follow up on an earlier study by Bartik (1992). They conduct a meta-analysis of empirical papers that examine the impact of state and local taxes on development. While they find evidence to suggest that tax policy impacts state economic development, questions remain as to the size of the effect. The work discussed above covers many aspects of the tax/growth relationship. Most are concerned with the longer-term growth impacts of tax burden, marginal rates, or the impact of a particular tax. While some discuss tax structure, there is no obvious general agreement on how structure should be defined or measured although most begin with a modified growth model. A few define structure in ways similar to the way it is defined herein, the shares of revenue collected by various taxes, but operationalize structure differently. The model defined and tested below has been well informed by the literature, especially Arnold et al. (2011) and Peach and Starbuck (2011). Major differences are that the current paper has a short-term (business cycle) orientation and that tax structure is operationalized somewhat differently than in previous work.

3. The model, the data, and the method The empirical model used to analyze the impacts of tax burden and structure built upon the model used by Arnold et al. (2011) with modifications suggested by Peach and Starbuck (2011). Arnold et al. (2011) operationalize tax structure in a way similar the present work. Peach and Starbuck use their model to measure the influence of gas and oil production on county-level growth across New Mexico counties for several decades and across 925 counties in 13 energy producing states for the year 2000. The Peach and Starbuck model was emulated in part for its simplicity, in part for its effectiveness, and in part for its congruence with Reed (2009) findings regarding the appropriate specification of empirical growth models aimed at explaining variations in growth across states. The analysis is based on

82

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

the model in Eq. (1). GROW(1 or 2)i,t = ˇ0 + ˇ1 RPCGDPi,t + ˇ2 GROW0i,t + ˇ3 CAPITALi,t + ˇ4 EDUCATIONi,2000 + ˇ5 POPGROWTHit + ˇ6 LFPRi,t + ˇ7 MINING + ˇ8 FORECLOSURE + [public finance variables] + [fixed effect variables] + ∈ i,t The definitions and sources of all variables used in the model are provided in Table 1. 3.1. Growth and control variables The variables included in the model operationalize the modified growth model as discussed above. RPCGDP serves as a proxy for the level of economic development in the state. GROW0 is included because, through the business cycle, a state’s growth rate in one year, especially if it is unusually high or low (boom or bust), could be predictive of the direction of the following year’s growth as the state’s economy adjusts to a more normal growth rate or as the growth trend continues across time periods. CAPITAL is included to represent the marginal increase in the capital stock. It is an imperfect proxy for a change in the capital stock but few alternatives are available. Hypothetically CAPITAL should lead to growth, a positive effect, but the time frame used in this work may be too short for impacts to be realized. EDUCATION is a proxy for the quality of labor and a positive sign is expected. POPGROWTH and LFPR are included to account for change/dynamism in the state’s labor market. At least in the long term, one would expect both to be positively related to growth. MINING is included to account for growth impacts driven by world energy and mineral markets that can influence short term growth rates in states heavily dependent on these markets, but the sign on MINING cannot be predicted a priori. Finally, a major cause of the Great Recession was boom and bust in the housing market. FORCLOSURE is included to control for the housing crisis’ influence on growth. As FORCLOSURE identifies the four states that were most impacted by the housing crisis a negative sign is expected. 3.2. Public finance variables Empirical studies of tax impacts face many challenges. Theory suggests that tax burdens reduce economic incentives and thus impose a drag on economic activity. Alternatively, high quality public capital and services can enhance economic activity by complementing private activities. Thus when one attempts to estimate the impact of tax burden it is often difficult to distinguish whether the impacts identified are responses to the tax burden or to the public goods and services funded by the taxes. BURDEN is included here largely as a control variable. No attempt is made to deal with the tax/expenditure conundrum, so no prediction is made as to the sign on BURDEN. Similarly, high marginal tax rates are thought to alter economic

incentives but, in a situation where individuals and firms are subjected to a variety of taxes, marginal rates are difficult to identify with any accuracy. This is even more so when one attempts to identify aggregate marginal tax rates. Since marginal rates are not available, the variable TRATE, an approximation of average tax rates faced by residents of a state, is included instead. Theory predicts a negative relationship between TRATE although the weakness of the proxy suggests that a bold hypothesis is unwarranted. Tax structure is operationalized by including the shares of revenue states collect from various taxes, SALES, PROPERTY, INCOME, and CORPORATE. Although states rely on a wide variety of taxes these four taxes represent the major revenue generators for most states. Hypothetically, one would expect any tax increase to have a negative impact on growth. However, the analysis that follows asks not the impact of an increase in a particular tax but rather the marginal impacts of switching from one tax to another while raising the same level of revenue, which represents revenue-neutral tax switching. As emphasized above the focus is on short-term impacts, so care should be taken in making inferences to longer time periods. 3.3. Fixed effects Differences in state tax structures and other regional characteristics are subtle and difficulty to observe and quantify. Likewise, broader forces whose impacts fall across all or many states can influence growth. A simple model cannot be expected to capture everything. For these reasons the model allows for year and stateby-state fixed effects. The various YEAR dummy variables account for unobserved annual effects, such as changes in the national economy, national policies, etc. The year 2004 is the base year for the annual fixed effects. Cross-sectional fixed effects are calculated as part of the estimation routine but are not reported. 3.4. A note on multicollinearity and residualized variables The model posited here has two potential problems. First, economic growth could potentially impact tax structure thus causing an endogeneity problem. To minimize this potential tax shares are used here to explain future growth and current year growth rate is included as an explanatory variable. Secondly, there is substantial potential for multicollinearity among the some of the explanatory variables. Multicollinearity is not always considered problematic, especially if prediction is the goal, but in this case the relative contributions some of the collinear variables are of interest. For this reason, three of the variables mentioned above, LFPR, BURDEN, and TRATE have been residualized as described in Allen (1997) and Hansen (2013). Initial analysis using variance inflation factors (VIFs) indicates a positive collinear relationship between EDUCATION and LFPR. States with more educated populations have higher labor force participation rates. To break this relationship, a regression with LFPR as the dependent variable and EDUCATION as the independent variable is

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

83

Table 1 Data descriptions and sources. Variable

Variable description

Source(s), see below

One-year growth

State one-year future growth rate – ln of real per capita GDP in year t + 1 − ln of real per-capita GDP in year t

1*

State two-year future growth rate – ln of real per capita GDP in year t + 2 − ln of real per-capita GDP in year t

1*

Current state (year t) Real per-capita GDP expressed in 2005 dollars

1

State one-year future growth rate – ln of real per capita GDP in year t − ln of real per-capita GDP in year t − 1

1*

Capital expenditures in manufacturing as percent of state GDP, by state and year

1, 2*

Percentage of state population age 25 and over with at least a high school diploma in 2000. Series repeated for each year in panel

3

State population growth from previous year − ln population in year t − ln population in year t − 1

4*

Portion of labor force participation rate by state and year that cannot be accounted for by education level (residualized variable as described below)

5*

Dummy variable = 1 for states with higher than average mining as a share of GDP for all seven years (AK, LA, MT, NM, OK, TX, WV, WY)

1*

Dummy variable = 1 for states identified as having the most persistently high foreclosure rates

6

State and local capital tax revenues, per capita, by state and year, in thousands that cannot be accounted for by GDP per capita and tax structure (residualized as described below)

7*

Rate of National Tax Foundation State and Local Tax Burdens Estimates that cannot be accounted for by GDP per capita and tax structure (residualized variable as described below)

8*

Percentage of total state and local tax revenues collected by sales tax, general and selective

9

Percentage of total state and local tax revenues collected by property taxes

9

Percentage of total state and local tax revenues collected by personal income tax

9

Percentage of total state and local tax revenues collected by corporate income tax

9

GROW1 Two-year growth GROW2 Real per-capita GDP RPCGDP Current year growth GROW0 Capital expenditures CAPITAL Education level EDUCATION Population growth POPGROWTH Labor force part. LFPR Mining state MINING High foreclosure state FORCLOSURE Tax burden

BURDEN Average rate

TRATE Sales tax SALES Property tax PROPERTY Personal income tax INCOME Corporate tax CORPORATE 2005–10 1 2 3

4 5 6 7 8

9

Dummy variables to identify year of observation. 2004 is the base year Bureau of Economic Analysis, Regional Economic Accounts, http://www.bea.gov/itable/index.cfm U.S. Census Bureau, Annual Economic Survey of Manufactures, various tables, http://www.census.gov/manufacturing/asm/ U.S Census Bureau – Table 1 – Educational Attainment of the Population 25 Years and Over for the United States, Regions, and States, and for Puerto Rico: 1990 and 2000. http://www.census.gov/prod/2003pubs/c2kbr-24.pdf Census Bureau Population Estimates. http://www.census.gov/popest/data/intercensal/state/state2010.html U.S. Census Bureau, Statistical Abstract of the U.S. and Bureau of Labor Statistics, various tables. Steve, McLinden, Bankrate.com, Top 10 States for Foreclosure, FL, NV, CA, AZ http://www.bankrate.com/finance/real-estate/top-10-states-for-foreclosure-1.aspx U.S. Census Bureau, State & Local Government Finance http://www.census.gov/govs/local/ National Tax Foundation, “State and Local Tax Burdens: All Years, One State, 1977–2010.” http://taxfoundation.org/article/state-and-local-tax-burdens-all-years-one-state-1977–2010 U.S. Census Bureau, State & Local Government Finance, http://www.census.gov/govs/local/ Indicates that final variable was calculated by authors using the original data from the source(s) identified

84

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

Table 2 States without selected taxes. States with no corporate income tax

States with no general sales taxa

States with no personal income tax

Nevada Texas Washington Wyoming

Alaskab Delaware Montana New Hampshire Oregon

Alaska Florida Nevada South Dakota Texas Washington Wyoming

a Note: The sales tax variable used in the econometric model is the sum of general and selective sales taxes. All states have selective sales taxes but these have no general sales tax. b Note: Alaska has no state general sales tax but small values for local general sales tax were reported and used in the econometric estimates.

Table 3 Descriptive statistics, relevant continuous variables – N = 50, 50 states, 7 years. Variable GROW1 GROW2 RPCGDP GROW0 CAPITAL EDUCATION POPGROWTH LFPRa BURDENa TRATEa SALES PROPERTY INCOME CORPORATE

Mean 0.00 0.01 41,207.00 0.00 11.55 81.96 0.01 0.00 0.00 0.00 35.39 30.34 20.25 3.73

St. dev.

Minimum

Maximum

0.03 0.04 7578.10 0.03 5.72 4.33 0.01 3.34 927.80 0.76 12.19 9.32 10.92 2.74

−0.10 −0.15 27,820.00 −0.10 1.13 72.90 −0.06 −11.97 −1964.50 −1.92 5.62 10.64 0.00 0.00

0.08 0.16 64,900.00 0.08 31.04 88.30 0.04 7.46 6763.10 2.40 62.71 64.60 44.65 22.42

Sources: Provided in Table 1 a Note: Reminds the reader that the original values for these variables were replaced by instruments as described in the text.

estimated and the residuals are used as in instrument for labor force participation. Thus, LFPR only includes the variation that is independent of the state’s education level. Similarly, BURDEN and TRATE are potentially jointly influenced by the state’s income level and tax structure. Thus, both BURDEN and TRATE were regressed against per-capita GDP, and the four tax share variables (SALES, PROPERTY, INCOME, and CORPORATE). Again, the residuals from these regressions are used as instruments for the original values leaving only the variation in the original variables that are not explained by the potentially collinear variables. As further evidence that endogeneity is not a serious problem in this short-term exercise; GROW0 did not have any detectable simultaneous relationship with either BURDEN or TRATE. Because LFPR, BURDEN, and TRATE have been residualized, care must be taken in their interpretation. The variation remaining in these residualized variables is the variation not shared with the variables used in the auxiliary regressions used to separate the impacts. It is clear that including tax shares for multiple taxes in the same model may be problematic. As in the case of BURDEN and TRATE, tax shares may be marginally influenced by annual deviations in income and because the shares (including taxes included in the other category, part of the fixed effects) must sum to 100%. Additionally, if one share changes from one year to the next, by definition some other tax share(s) must change as well. Variance inflation factors, reported below, indicate that this not as severe

as one might expect. There is some evidence of moderate collinearity between SALES and INCOME but, for two reasons, no adjustments are made in this case. One reason is that the variance inflation factors do not provide evidence of severe multicollinearity among the tax share variables. Typically, VIF values greater than five are considered indicative of severe multicollinearity (Studenmund, 2001, p. 248) although some authors suggest a rule-ofthumb cutoff of ten (Kennedy, 2003, p. 213). The highest VIF, and the only one above five, is SALES with a value of 6.1. At first, this may seem odd but as only the tax shares of four (major) taxes are included and that many exogenous and historical factors influence state tax structures, weak collinearity seems less surprising. Secondly, if actual tax shares are replaced by instruments, it is unclear how the coefficients should be interpreted because the shares no longer sum to 100%.

4. Data and estimation method The model is estimated using a pool of data on the 50 American states for the seven-year period, 2004–2010. The dependent variable is alternatively the one-year forward growth in state real per capita GDP (GROW1) and the two-year forward growth rate in state real per capita GDP (GROW2). Growth is measured as the log differences between the future and current values of real per-capita GDP.

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

Table 1 provides information on the variables used in the model. For full disclosure, Table 2 identifies states that do not use particular taxes, therefore having zero shares. Table 3 provides descriptive statistics on the continuous variables used. The estimation of pooled and panel data models introduces special problems, in particular the possibility of cross-sectional heteroskedastic and serially correlated errors making the use of ordinary least squares estimation inappropriate. One long-accepted method to deal with these problems is to use the Parks–Kmenta model (Kmenta, 1986) wherein adjustments are made for these potential problems. Beck and Katz (1995) conduct Monte Carlo experiments and conclude that under some conditions the Parks–Kmenta model underestimate standard errors. Beck and Katz (1995) propose a method whereby panel coefficients are estimated by ordinary least squares and standard errors from a heteroskedastic model are used for hypothesis testing. Reed and Webb (2010) replicate Beck and Katz’s work and conduct further Monte Carlo studies and conclude that under some conditions the Parks–Kmenta feasible generalized least squares model is more appropriate than the Beck and Katz method. Later, Reed and Ye (2011) attempt to identify the conditions that favor the choice of model. While they provide useful guidelines, there is still not a definitive answer for the correct choice of model for pooled/panel estimation. As a compromise this paper reports the estimates from both models. The body of the paper uses the panel corrected standard errors model (PCSE) in the mode of Beck and Katz (1995) with state fixed effects estimated as described in Greene (2000) and programmed in the Shazam econometrics software. The state fixed effects account for unobserved state-by-state variation and, according to Kennedy (2003), allow the estimation of short run effects. Time fixed effects are estimated by the inclusion of year dummy variables with 2004 as the omitted base year. Appendix reports the parameters estimated by the Parks–Kmenta method for comparison.

85

may be caused by the weakness of the capital measure. Alternatively, the expected long-term benefits of investment may have the opportunity cost of short-term decline if investment expenditures displace some other type of expenditure that has a greater short-term growth impact. EDUCATION has no detectable relationship with one year growth but shows a positive impact on two-year growth. A one percentage point increase in average educational achievement in a state predicts a 0.9% increase in two-year growth. Educated labor was helpful in survival/recovery during the Great Recession. Similarly, LFPR shows a positive and detectable relationship with two-year growth, although the direction of causation is debatable. MINING and FORCLOSURE are included to account for differences in economic reliance on natural resource bases and differential impacts from the housing boom and bust. MINING shows no detectable relationship with growth, perhaps because the impacts, if any, are accounted for in the cross-sectional fixed effects. However, FORECLOSURE has a negative relationship with two-year growth. The model estimates that the four high-foreclosure states had, on average, a two-year growth rate that was 1.6% lower than the average for other states. High foreclosures appear to retard recovery.

5.2. Tax burden and marginal tax rate variables The estimated model shows tax burden, BURDEN, has a small but statistically detectable positive relationship with one-year growth and no relationship with two-year growth. It is likely that this relationship is clouded by the tax/expenditure conflation problem mentioned above. It is difficult to separate assumed negative impacts of tax burdens from the assumed positive impacts of the related public spending. Alternatively, perhaps states with larger public sectors, high values of BURDEN, are cushioned from the negative impacts of recession, at least in the short term. Tax rates show no relationship with either one or two year growth.

5. Results 5.1. Growth and control variables

5.3. Tax structure variables

The initial level of state economic development, proxied by RPCGDP, shows a negative and statistically detectable relationship with both one-year and two-year growth. This indicates that wealthier states, on average, had lower growth rates than did poorer states. Over longer time periods, this would suggest convergence. Alternatively, economic growth (GROW0) in one year, year t, shows a positive impact on growth in the following two years. It appears that growth begets growth, at least in the short term. A 1% change in current growth predicts a 0.109% increase in the following year’s growth rate and a 0.203% increase in growth over the following two years. Capital, properly measured, should relate positively to long-term economic growth. In this case CAPITAL shows no detectable relationship with one-year growth and a negative relationship with two-year growth. This unexpected outcome

The results on the tax share variables, SALES, PROPERTY, INCOME, and CORPORATE provide the key insights of this paper, and care must be taken in interpreting the results. One might be tempted to look at a single coefficient and conclude something about the marginal impact of a tax. One must remember that the variables are expressed in shares and thus a marginal change in one tax share implies an offsetting share to one or more of the other taxes. In other words, to decrease the sales tax share by one percentage point implies that the shares of the other variables jointly increase by one percentage point. These results must be interpreted for the case of a revenue neutral shift in tax shares. The marginal effects must be netted out. Also, because the growth impacts are potentially different for the one- and two-year growth, rates a separate interpretation for each year is in order.

86

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

Table 4 Empirical estimates using pooled data, 50 states, 7 years, n = 350. Panel corrected standard error model with state by state fixed effects. Variable

One-year growth (t-ratio)

Two-year growth (t-ratio)

Variance inflation factors

Real per-capita GDP RPCGDP Current year growth GROW0 Capital expenditures CAPITAL Education level EDUCATION Population growth POPGROWTH Labor force part LFPR Mining state MINING High foreclosure state FORCLOSURE Tax burden BURDEN Marginal rate TRATE Sales tax SALES Property tax PROPERTY Personal income tax INCOME Corporate tax CORPORATE 2005

−0.598E − 06 (−2.400)** 0.109 (1.838)*** −0.0001 (−0.428) 0.0003 (1.067) 0.263 (1.408) 0.0008 (1.640) −0.003 (−0.737) −0.006 (−1.104) 0.421E−05 (2.485)* −0.0001 (−0.089) −0.0007 (−3.497)* −0.0006 (−3.089)* −0.0005 (−2.632)* −0.0002 (−0.276) 0.044 (3.025)* 0.004 (0.187) −0.008 (−0.370) −0.044 (−1.507) −0.051 (−1.198) 0.058 (−1.365) .5485

−0.893E − 06 (−2.503)** 0.203 (2.100)** −0.008 (−2.086) 0.009 (1.927)*** 0.023 (0.083) 0.002 (2.368)** −0.003 (−0.360) −0.016 (−2.096)** −0.107E−05 (−0.444) 0.003 (1.439) −0.0009 (−3.147)* −0.0009 (−2.843)* −0.0007 (−2.238)** 0.0000 (0.047) −0.020 (−1.038) 0.078 (−3.115)* −0.109 (−2.937)* −0.130 (−2.626)* −0.150 (−2.538)** −0.145 (−2.085)** .5555

2.42

2006 2007 2008 2009 2010 R-Square * ** ***

1.08 1.65 1.74 1.47 1.18 1.77 1.49 1.97 1.68 6.31 2.87 4.06 2.00

99% confidence level. 95% confidence level. 90% confidence level.

5.4. Revenue neutral shifts in tax share Tables 4–6 report the predicted net effects of a revenue-neutral shift of 1% of tax revenue from one tax

source to another. The changes are based on the estimated coefficients on SALES, PROPERTY, INCOME, and CORPORATE. The estimated coefficients on SALES, PROPERTY, and INCOME are statistically significant with at least

Table 5 Estimated percentage change of growth rate from a revenue neutral shift in tax sharesone-year forward growth (one percentage point shift in revenue share).

Table 6 Estimated percentage change of growth rate from a revenue neutral shift in tax shares two-year forward growth (one percentage point shift in revenue share).

From → To ↓

Sales tax SALES [−0.07%]

Prop. tax PROPERTY [−0.06%]

Pers. inc. tax INCOME [−0.05%]

Corp. inc. tax CORPORATE [0.0%]

From → To ↓

Sales Tax SALES [−0.09%]

Prop. Tax PROPERTY [−0.09%]

Pers. Inc. Tax INCOME [−0.07%]

Corp. Inc. Tax CORPORATE [0.0%]

SALES [−0.07%] PROPERTY [−0.06] INCOME [−0.05%] CORPORATE [0.0%]

n/a

−0.01%

−0.02%

−0.07%

n/a

0.0%

−0.02%

−0.09%

+0.01%

n/a

−0.01%

−0.06%

0.0%

n/a

−0.02%

−0.09%

+0.02%

+0.01%

n/a

−0.05%

+0.02%

+0.02%

n/a

−0.07%

+0.07%

+0.06%

+0.05%

n/a

SALES [−0.09%] PROPERTY [−0.09%] INCOME [−0.07%] CORPORATE [0.0%]

+0.09%

+0.09%

+0.07%

n/a

Source: Coefficients reported in Table 4 and author calculations.

Source: Coefficients reported in Table 4 and author calculations.

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

95% confidence; therefore, their slightly rounded values are used in Tables 5 and 6. Because the hypothesis that the coefficient on CORPORATE is equal to zero cannot be rejected, a zero is used as the coefficient for CORPORATE. Tables 5 and 6 assume that a one-percent revenue shift is made from one specific tax to another of the four taxes specifically included in the model. To identify the predicted growth impact of a 1% revenue-neutral move across taxes one simply picks one of the taxes listed across the second row of the table and one of the taxes listed in the first column of the table. The number in the shared cell in the table indicates the net marginal predicted percentage impact on the growth rate from a one-percentage point revenue shift from the tax indicated in row two to the tax in column one. For example, if one percentage point of revenue collections is transferred from sales tax to personal income tax, the model predicts a 0.02% increase in both one-year and two year growth rates. For one-year growth, decreasing reliance on sales tax is predicted to increase growth by 0.07%, while increasing reliance on personal income is expected to reduce growth by 0.05%, leaving a positive difference of 0.02%. Although the coefficients are different for two-year growth, the revenue neutral impacts on growth are the same as for one-year growth. Alternatively, a revenue neutral switch from reliance on personal income taxes to property taxes is predicted to have a negative growth impact on both one- and twoyear growth throughout the Great Recession period. Given that CORPORATE has no detectable impact, higher reliance on corporate taxes should have been growth enhancing through the Great Recession. This may simply be a short-term phenomenon, but it would advise against cuts in corporate income taxes as a short-term response to recession.

6. Discussion/conclusions In recent decades, tax policy has been at the center of public debate and it has at times become difficult to conduct a dispassionate, evidence-based discussion of tax issues (Adkisson & Mohammed, 2012). At the same time, tax policies are important in that they alter economic signals and thus have potential impacts on the operation of the economy. As mentioned in the introduction, the general public, economic developers, and others regularly express concern as to how taxes impact their states. The concerns are genuine, but one hopes that tax structure policy decisions are guided by evidence on the impact of taxes rather than conventional wisdom or emotional appeal. For example, Mazcrov (2013) recently criticizes the Tax Foundation for asserting that the bulk of empirical studies on taxes suggest that higher taxes always reduce growth, when much of the literature tells a different story. To track, quantify, and explain state and local tax structures, let alone explain their impacts on growth, is not a particularly clean and easy task. Industrial mix, political factors, and tax competition, in combination with other potential influences, can shape a state’s tax structure (Porca, 2003). Similarly tax structure itself, measured as a

87

set of revenue shares, is potentially influenced by the very factors it might potentially explain. Still, it is a worthwhile exercise to try to sort out the relation between tax structure and economic growth. This paper explains part of the relationship between tax structure and short-run growth in real per capita GDP across the 50 American states during the 2004 to 2010 period, a period that includes the Great Recession. An empirical economic growth model is modified to suit the question and, to provide estimates as to how revenue neutral shifts in tax shares relate to short-term economic growth. To attempt to quantify the impact of revenue-neutral tax reassignment is unique in the literature. For reasons discussed above, care has been taken to eliminate some of the potential pitfalls in this type of exercise and the results are discussed largely in terms of relationships rather than causes. The authors recognize that further study is required to verify the relationships identified in this work, and particularly to find whether the identified relationships are evident over other and longer time periods. Still, at least one general conclusion can be drawn from this analysis and, if corroborated by subsequent research, may be useful to guide tax policy. Tax structure is shown to have a statistical relationship with short-term economic growth between 2004 and 2010. The marginal effects of various taxes differ and, although they are statistically detectable, they are not especially large. In all cases the ‘impact’ on the one- and two-year change in real per capita GDP is only plus or minus a small fraction of a percent. These impacts may be larger if longer term growth periods are examined. Ultimately, the evidence in this work suggests that minor state-bystate differences in tax structure are relatively benign in terms of short run economic growth and/or recovery from a recessionary period. One might be tempted to discount the findings of this paper as being statistically significant but economically insignificant. However, they are not. As mentioned in the introduction, tax policy, or at least proposals for changes in tax policy, is not always driven by objective evidence. Thus, some proposals for change may be made based on conventional wisdom rather empirical evidence. The empirical evidence here suggests that marginal differences in tax structure have detectable but very small impacts on growth rates, at least in the context of the Great Recession. Given this evidence, states facing economic downturns should not rush to adjust their tax structures if their goal is to enhance their growth prospects. At least they should not expect large changes if they do. The paper also suggests that short-term cuts in corporate taxes to promote recovery are ill advised. Having said this, further exploration of this topic is certainly warranted. This work faces some limitations that could possibly be overcome in future work. In particular, the years 2004–2010 were chosen for analysis in part because of data availability and unavailability. At the time of writing, the last year for which reliable tax share data are available is 2010. While data are available for earlier years there is a reporting gap, 2001–2003, that limits the choice of years

88

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89

observed. Without this gap, it would be possible to make stronger inferences as to the impact of tax structure on long-term growth. As new data becomes available or by using methods to work around the missing data problem, it might be possible to extend the study to include more years of data and thus extend the analysis to the longer term. Similarly, only four categories of taxes are specifically analyzed here. In future work it might be possible to take a more refined look at tax structure by operationalizing a wider array of tax shares.

Appendix

Table A2 Estimated percentage change of growth rate from a revenue neutral shift in tax shares one-year forward growth (one percentage point shift in revenue share). From → To ↓

Sales tax SALES [−0.04%]

Prop. tax PROPERTY [−0.03%]

Pers. inc. tax INCOME [−0.04%]

Corp. inc. tax CORPORATE [0.0%]

SALES [−0.04%] PROPERTY [−0.03] INCOME [−0.04%] CORPORATE [0.0%]

n/a

−0.01%

0.00%

−0.04%

+0.01%

n/a

−0.01%

−0.03%

0.00%

−0.01%

n/a

−0.04%

+0.04%

+0.03%

+0.04%

n/a

Source: Coefficients reported in Table 4 and author calculations.

Tables A1–A3. Table A1 Empirical estimates using pooled data, 50 years, 7 years, n = 350 Park–Kmenta method (FGLS) with state by state fixed effects. Variable

One-year growth (t-ratio)

Two-year growth (t-ratio)

Variance inflation factors

Real per-capita GDP RPCGDP Current year growth GROW0 Capital expenditures CAPITAL Education level EDUCATION Population growth POPGROWTH Labor force part LFPR Mining state MINING High foreclosure state FORCLOSURE Tax burden BURDEN Marginal rate TRATE Sales tax SALES Property tax PROPERTY Personal income tax INCOME Corporate tax CORPORATE 2005

−0.411E−06 (−2.466)** 0.160 (4.635)* −0.0001 (0.742) 0.0003 (1.349) 0.174 (1.379) 0.001 (2.821)* −0.004 (−1.421) −0.006 (−1.817)*** 0.261E − 05 (1.786)*** −0.0005 (−0.410) −0.00038 (−2.166)** −0.00034 (−2.310)** −0.00037 (−2.361)** 0.0002 (0.447) 0.041 (2.816)* −0.003 (−0.117) −0.013 (−0.5591) −0.043 (−1.359) −0.045 (−1.103) -0.055 (−1.108) .6121

−0.608E−06 (−2.180)** 0.214 (3.318)* −0.0005 (−1.605) −0.001 (2.798)* −0.097 (−0.483) 0.001 (1.982)** −0.001 (−0.284) −0.008 (−1.521) −0.125E − 06 (−0.050) 0.001 (0.435) −0.00059 (−1.854)*** −0.00057 (−2.240) −0.00049 (−1.755)*** −0.00005 (−0.0139) −0.020 (−1.007) −0.077 (-2.526)** −0.106 (−2.490)** −0.125 (−2.433)** −0.138 (−2.276)** −0.138 (−1.906)*** .6181

2.42

2006 2007 2008 2009 2010 Buse R-square * ** ***

99% confidence levels. 95% confidence levels. 90% confidence levels.

1.08 1.65 1.74 1.47 1.18 1.77 1.49 1.97 1.68 6.31 2.87 4.06 2.00

R.V. Adkisson, M. Mohammed / The Social Science Journal 51 (2014) 79–89 Table A3 Estimated percentage change of growth rate from a revenue neutral shift in tax shares two-year forward growth (one percentage point shift in revenue share). From → To ↓

Sales tax SALES [−0.06%]

Prop. tax PROPERTY [−0.06%]

Pers. inc. tax INCOME [−0.05%]

Corp. inc. tax CORPORATE [0.0%]

SALES [−0.06%] PROPERTY [−0.06%] INCOME [−0.05%] CORPORATE [0.0%]

n/a

0.00%

−0.01%

−0.06%

−0.00%

n/a

−0.01%

−0.06%

+0.01%

+0.01%

n/a

−0.05%

+0.06%

+0.06%

+0.05%

n/a

Source: Coefficients reported in Table 4 and author calculations.

References Adkisson, R. V., & Mohammed, M. (2012). Pragmatism to dogmatism: The laissez faire myth and the disabling of the American Fisc. Review of Social Economy, 70(4), 421–450. Allen, M. P. (1997). Understanding regression analysis. New York: Plenum Press. Arnold, J. (2008). Do tax structures affect aggregate economic growth empirical: Evidence from a panel of organization for economic co-operation and development countries. Organization for Economic Co-operation and Development (Working Paper) Arnold, J., Brys, B., Heady, C., Johansson, A., Schwellnus, C., & Vartia, L. (2011). Tax policy for economic recovery and growth. Economic Journal, 121, 59–80. Bartik, T. J. (1992). The effects of state and local taxes on economic development: A review of recent research. Economic Development Quarterly, 6(1), 102–111. Barro, R. J. (1990). Government spending in a simple model of endogenous growth. Journal of Political Economy, 98(5), S103–S125. Barro, R. (1991). Economic growth in a cross-section of countries. Quarterly Journal of Economics, 106(2), 407–443. Beck, N., & Katz, J. N. (1995). What to do (and not to do) with timeseries cross-section data. The American Political Science Review, 89(3), 634–647. Bhattarai, K. (2010). Taxes, public spending and economic growth in OECD countries. Problems and Perspectives in Management, 8(1), 14–30. Boskin, M. J. (1988). Tax policy and economic growth: Lessons from the 1980. Journal of Economic Perspectives, 2(4), 71–97. Creedy, J., & Gemmell, N. (1982). The built-in flexibility of progressive income taxes: A simple model. Public Finance, 37, 362–371. Creedy, J., & Gemmell, N. (1984). Income redistribution through taxes and transfers in Britain. Scottish Journal of Political Economy, 31, 44–59. Dye, T. R., & Feiock, R. C. (1995). State income tax adoption and economic growth. Social Science Quarterly, 76(3), 648-54. Engen, E., & Skinner, J. (1996). Taxation and economic growth. National Tax Journal, 49(4), 617–642. Gemmell, N. (1985). Tax revenue shares and income growth: A note. Public Finance, 40(1), 137–145. Gentry, W., & Ladd, H. F. (1994). State tax Structure and multiple policy objectives. National Tax Journal, 47(4), 747–772. Greene, W. H. (2000). Econometric analysis (4th ed.). Upper Saddle River, NJ: Prentice Hall. Hansen, B. E. (2013). Econometrics. University of Wisconsin. www.ssc.wisc.edu/∼bhansen Hettich, W., & Winer, S. (1984). A positive model of tax structure. Journal of Public Economics, 24(1), 67–87.

89

Howard, B. (2003). Does the mix of a state’s tax portfolio matter? An empirical analysis of United States tax portfolios and the relationship to levels and growth in real per capita gross state product. Journal of Applied Business Research, 19(3), 97–109. Kennedy, P. (2003). A guide to econometrics (5th ed.). Cambridge, MA: The MIT Press. King, R., & Rebelo, S. (1990). Public policy and economic growth: Developing neoclassical implications. Journal of Political Economy, 98(1), 126–151. Kmenta, J. (1986). Elements of econometrics (2nd ed.). Macmillan Press. Kneller, R., Bleaney, M. F., & Gemmell, N. (1999). Fiscal policy and growth: Evidence from organization for economic co-operation and development countries. Journal of Public Economics, 74(2), 171–190. Koch, S. F., Schoeman, N. J., & van Tonder, J. J. (2005). Economic growth and the structure of taxes in South Africa: 1960–2002. South African Journal of Economics, 73, 190–210. Koester, R. B., & Kormendi, R. C. (1989). Taxation, aggregate activity and economic growth: Cross country evidence on some supply-side hypotheses. Economic Inquiry, 2(3), 367–386. Lee, Y., & Gordon, R. H. (2005). Tax structure and economic growth. Journal of Public Economics, 89, 1027–1043. Lucas, R. (1990). Supply-side economics: An analytical review. Oxford Economic Papers, 42(2), 293–316. Luna, L., & Murray, M. N. (2010). The effects of state tax structure on business organizational form. National Tax Journal, 63(4), 995–1021. Mamatzakis, E. C. (2005). The dynamic responses of growth to tax structure for Greece. Applied Economics Letters, 12, 177–180. Mazcrov, M. (2013). Academic research lacks consensus on the impact of state tax cuts on economic growth. Washington, DC: Center on Budget and Policy Prioritie. July 17. Mendoza, E., Milesi-Ferretti., G., & Asea, P. (1997). On the effectiveness of tax policy in altering long-run growth: Harberger’s superneutrality conjecture. Journal of Public Economics, 66(1), 99–126. Miller, S., & Russek, F. (1997). Fiscal structures and economic growth: International evidence level. Economic Inquiry, 35(3), 603–613. Mullen, J. K., & Williams, M. (1994). Marginal tax rates and state economic growth. Regional Science and Urban Economics, 24(6), 687–705. Nelson, M. A. (1989). Taxation of natural resources and state economic growth. Growth and Change, 20(4), 13–30. Peach, J. T., & Starbuck, C. M. (2011). Oil and gas production and economic growth in New Mexico. Journal of Economic Issues, 45(2), 511–526. Pesendorfer, K. (2008). Tax and economic growth in Austria. In Monetary Policy & the Economy, 3rd Qtr. (pp. 21–40). Phillips, J. M., & Goss, E. P. (1995). The effect of state and local taxes on economic development: A meta-analysis. Southern Economic Journal, 62(2), 320–333. Porca, S. (2003). Important determinants of state tax revenue portfolios. In Proceedings: Ninety-fifth Annual Conference on Taxation, Orlando, Florida, November 14–16, 2002, and minutes of the annual meeting of the National Tax Association Thursday, November 14, 2002, (pp. 416–426). Rabuska, A., & Bartlett, B. (1985). Tax policy and economic growth in developing nations. Washington: U.S. Agency for International Development. Reed, W. R. (2009). The determinants of U.S. state economic growth: A less extreme bounds analysis. Economic Inquiry, 47(4), 685–700. Reed, W. R., & Webb, R. (2010). The PCSE estimator is good – Just not as good as you think. Journal of Time Series Econometrics, 2(1), 1–24. Reed, W. R., & Ye, H. (2011). Which panel estimator should I use? Applied Economics, 43, 985–1000. Reese, L. A., & Rosenfeld, R. A. (2001). Yes, but. . .: Questioning the conventional wisdom about economic development. Economic Development Quarterly, 15(4), 299–312. Studenmund, A. H. (2001). Using econometrics: A practical guide (4th ed.). Boston: Addison Wesley Longman. Widmalm, F. (2001). Tax structure and growth: Are some taxes better than others? Public Choice, 107, 199–219.