Taxes, Growth and the Convergence of Incomes among U.S

Taxes, Growth and the Convergence of Incomes among U.S. States: 1959-99

Juan Feng Department of Agricultural and Resource Economics 2200 Symons Hall University of Maryland College Park, MD 20742 and Douglas J. Young* Department of Agricultural Economics and Economics Montana State University Bozeman, MT 59717 Voice: (406) 994-5622 FAX: (406) 994-4838 [email protected] * Correspondence should be addressed to Young.

February, 2003

The authors thank Robert Barro, Robert Fleck, Andrew Hanssen, Paul Polzin, Thomas Power and Christine Stoddard for helpful comments. Any errors remain the responsibility of the authors.

Taxes, Growth and the Convergence of Incomes Among U.S. States: 1959-99

Page 2

Abstract We examine the evolution of per capita income among the U.S. states between 1959 and 1999. Inequality of state incomes diminished significantly during the first two decades, but remained constant or grew slightly during the last two decades. There is strong evidence of slow conditional convergence (2% - 3% per year) during both periods, but growth in the last two decades favored states that were already relatively wealthy. Demographic structure and urbanization have significant effects on income levels and – less strongly – growth. There is little evidence that overall tax levels affect the level or growth of per capita income, and statistically significant but small effects of tax structure.


Page 3

I. Introduction Are taxes “too high” in high tax states? Do high state and local taxes reduce the growth of per capita income? Certainly, income levels vary considerably among the contiguous 48 states of the U.S. 1 For example per capita personal income was $38,505 in Connecticut in 1999, while Mississippi’s income was only $20,180 (see Table 1). Income growth has also varied a great deal over the last 40 years: The highest average annual real growth rate was achieved by South Carolina at 2.9 percent versus a national average of 2.2 percent. Nevada had the slowest growth at 1.7 percent per year. The central purpose of this paper is to examine tax policies and other factors that may account for this diversity in levels and growth of income. Our empirical approach owes much to studies of cross-country and cross-state growth by Robert Barro and his co-workers, in which incomes depend on tax policies, demographics and the structural composition of a state’s economy (Barro, 1997a, Barro and Sala-I-Martin, 1992, 1995). In addition, Barro’s work has documented the role of convergence – the tendency for lower income economies to grow faster and thus reduce inequality, at least among economies having similar institutional environments. In recent years, however, the unconditional convergence of incomes among states stopped (see Figure 1 and Bernat, 2001). Our results indicate that conditional convergence has continued, while other determinants of growth – mainly urbanization and human capital – have increased inequality.

1

This study excludes Alaska and Hawaii, where economies have been subject to somewhat different forces. For Alaska, the economy and tax revenues have been dominated by the construction of the pipeline and the subsequent flow of oil. The performance of Hawaii’s economy has been strongly affected by Japan: boom in the 1970s and 1980s, and stagnation in the 1990s. In addition, migration between Alaska, Hawaii, and the other states is more costly than within the contiguous states.

Taxes, Growth and the Convergence of Incomes Among U.S. States: 1959-99 Figure 1a Convergence: 1959-79 Growth Rate of Income

4.5% 4.0% 3.5% 3.0% 2.5% 2.0% 6000

8000

10000

12000

14000

16000

Per Capita Income in 1959

Figure 1b Convergence? 1979-99 Growth Rate of Income

2.5% 2.0% 1.5% 1.0% 0.5% 0.0% 14000

16000

18000

20000

22000

Per Capita Income in 1979

24000

26000

Page 4


Page 5

Studies of growth and taxes can be divided along three important conceptual lines. First, cross-country studies mostly assume that labor is immobile (e.g. Barro, and Mendoza, et al., 1997). As a result, levels of income and, implicitly, utility may differ a great deal between countries. In contrast, studies of state- level growth, including this one, often make the opposite assumption: Households are so mobile that any opportunities for improvement through migration are quickly eliminated. In short, utilities are equalized across states, at least for the marginal migrant (see, e.g., Gyourko and Tracy, 1989, 1991). Second, some studies examine the effects of taxes conditional on the level of public services, while other studies examine the unconditional effects of taxes, i.e. the consequences of simultaneous changes in taxes and (often implicitly) spending. 2 From a conditional perspective, taxes lower utility because they reduce the returns to work, saving and/or investment. But taxes also finance the provision of public services that may substitute for goods that could be provided privately (public schools, health care, law enforcement), and that may enhance private productivity (highways and other infrastructure). Thus, tax increases are always “bad” from a conditional perspective, but taxes together with the public services they finance may be either too low or too high from an unconditional perspective. Our basic approach is to apply the logic of compensating differentials to the unconditional level of taxes. We assume that utility is maximized at some positive level of taxes and spending. If taxes are either too low OR too high, incomes are expected to be higher in order to prevent households from moving to another state with a better tax-spending package (Tiebout, 1956). The model yields distinct implications for the slope of the tax- income relationship

2

Barro and Mendoza et al. fall into the first category, while Gyourko and Tracy are in the latter. Helms (1985) pioneered the technique of explicitly incorporating the government budget constraint.


Page 6

depending on the level of taxes: If taxes are too high in most states, then higher tax states are expected to have higher incomes, since higher taxes require higher compensation. On the other hand, if taxes are too low in most states, the model predicts a negative relationship between incomes and taxes, since higher taxes are closer to optimal levels. Finally, if taxes are about right, or if the distribution of state tax levels is roughly symmetric about the optimal level, the model predicts little or no relationship between taxes and income levels. The third conceptual distinction is between studies that aggregate all taxes into a single measure (e.g. percentage of income), and those studies that examine the structure of taxation. Theoretically, the exact effects of taxes depend on their structure, i.e. the details of income, sales, property and other taxes (Gyourko and Tracy, 1989, 1991, Mendoza et al., 1997). The “optimal” – from the standpoint of economic efficiency – tax structure may differ from state to state, depending on the state’s economic structure as influenced by geography, natural resources, demographics, etc. In addition, a state’s optimal tax structure may differ from the actual structure because of political effects. Thus the structure of taxes may be inefficient, with distinct incentives for migration by people with different levels or sources of income. Tax effects have been scrutinized in many previous studies, with mixed findings. 3 The present study examines the relationship between income levels and growth on the one hand, and several measures of taxes. The broadest measure is total tax revenues as a percentage of personal income. Separate measures of property, sales and income taxes are also considered. The empirical results indicate that there is almost no relationship at all between the level of income and the level of taxes or their structure. Income growth is also unrelated to the level of taxes, but

3

For reviews, see Bartik, 1992, Becsi, 1996, and Wayslenko, 1997. Some of these studies consider the growth in total income, which is the product of growth in per capita income and population growth.


Page 7

slower growth is associated with rising tax rates. In contrast, urbanization, age structure and education explain a great deal of the cross-state variation in income levels, and some of the variation in growth rates. Theoretical perspectives on the relationship between public finances and incomes are considered in the next section. The third section presents the econometric models and data, while the fourth discusses empirical results. II. Theory A. Public Finances in a Closed Economy Consider first an isolated (closed), one-period economy populated by a representative individua l subject to a proportional income tax at rate t. The individual maximizes utility by choosing hours of work, h, at real wage rate w, and goods consumption, c, subject to the budget constraint that consumption equals after tax income. 4 c = (1-t) wh => dc/dh = (1-t)w A proportional income tax is formally equivalent to a reduction in wages, creating opposing income and substitution effects, so that the net effect on labor supply is uncertain a priori. As noted by several authors, however, government services financed by taxes also have income effects (Gwartney and Stroup, 1983, Barro, 1997). If the value of the services provided is approximately equal to the taxes paid, then the income effects of the taxes and services are largely offsetting. The remaining substitution effect unambiguously reduces labor supply. 5

4

“Hours of work” are shorthand for all the dimensions of labor supply that may be influenced by taxation, including effort, occupational choice, investments in human capital, entrepreneurship, etc. 5 In the case of transfer payments, the income effects of the taxes and payments accrue to different individuals. Thus, the taxpayer is subject to the usual (opposing) income and substitution effects, while work by the transfer recipient is unambiguously discouraged. If transfers are mean-tested, there is an additional substitution effect discouraging labor supply.


Page 8

The effects of other taxes differ according to their exact structure, but the basic principle of offsetting income effects from taxation and expenditure remains. Thus, a progressive income tax that raises the same amount of revenue as a proportional tax differs in the distribution of the income and substitution effects among the populace: Higher income individuals experience larger effects. For the special case of a single individual, a progressive tax results in unambiguously lower labor supply than a proportional income tax that raises the same amount of revenue, because of the larger substitution effect. 6 Thus, there is some presumption that progressive taxes discourage labor supply more than proportional taxes, although the result is ultimately ambiguous because of the distributional effects. Broad based consumption taxes are equivalent to proportional income taxes in static models of labor supply: Both taxes reduce the after tax rate at which leisure can be transformed into goods. 7 Let t’ denote a sales tax applied to purchases of consumption goods. c(1+t’) = wh => dc/dh = w/(1+t’) The main difference is that income taxes also alter the relative price of goods today versus goods tomorrow, while broad based consumption taxes do not. Thus, income taxes discourage saving and investment more than sales taxes. Property taxes levied on equipment and structures also reduce the returns to investment and thus, ultimately saving. These considerations suggest that in an isolated economy, income levels and growth rates may be negatively related to taxes: Higher tax rates generally have larger substitution effects that reduce incentives to work and save, to invest in physical and human capital, and to engage in 6

See, e.g. Rosen (2002), pp. 374-6. If a consumption tax exempts various goods and services– as all real-world sales taxes do – there are additional considerations of whether the untaxed goods are more or less complementary to leisure than the taxed goods. Additionally, most retail sales taxes include significant amounts of productive inputs, affecting firm location 7


Page 9

entrepreneurial activities that increase wealth. Indeed, Barro (1997) concludes that countries with higher tax rates experience slower growth in per capita income, ceteris paribus. On the other hand, Stokey and Rebelo (1995) and Slemrod (2000) conclude that tax effects on income growth in the United States – while present – are small in magnitude. B. One State Among Many People are relatively free to move among states, in comparison with movements between countries. Thus, equilibrium requires that individuals maximize utility with respect to location, in addition to making their optimal choices of work, consumption, saving, etc. First suppose that all individuals are identical in tastes and productivity. Then equilibrium requires that utility be equalized in every state. That is, if state i offered higher utility than state j, people would move from state j to state i , until there is no marginal gain from further migration. If there is some fixed factor resulting in diminishing returns to labor, or crowding produces negative externalities, long run equilibrium will generally result in positive populations in each state. 8 Equal utility does not imply that states will be identical in all respects, however. Rather, the advantages a state may offer in one dimension may be offset by disadvantages in other respects. Specifically, some states may offer greater amenities (a better climate, beautiful scenery, a lower crime rate, etc.), more or higher quality public services, or a lower cost of living, but have lower incomes. Conversely, states having an unpleasant climate, high crime rates, a high cost of living or few or low quality public services would be expected to have higher incomes.

decisions (Gyourko and Tracy, 1989, 1991). 8 A contrary view suggests that increased density promotes greater specialization and technical advance (Becker and Glaeser, 1999).


Page 10

Suppose the utility of a representative individual in a state depends on per capita income, y, taxes, and other characteristics of the state, z. ui = u ( yi , taxi , zi ) If utility is equalized across states, then ui = u j = u , i.e. du = 0 across states. Differentiating the utility function with respect to income and taxes, one obtains a predicted relationship between the variables:

∂u ∂u dy + dtax = 0 ∂y ∂tax dy ∂u ∂tax ⇒ =− dtax ∂u ∂y du =

It is plausible that utility is positive related to per capita income, ceteris paribus, i.e., ∂u >0 ∂y Thus, dy > ∂u < 0⇔ 0 < > dtax ∂tax

If taxes lower an individual’s utility at the margin, then the model predicts that the depressing effect of taxes is compensated by higher per capita income. Conversely, if taxes raise an individual’s utility, then per capita income is lowered to balance out the “amenity.” As illustrated in Figure 2, long run equilibrium requires that per capita incomes be higher in states with higher tax burdens, if at the margin higher taxes lower utility, i.e. if taxes are “too high.” Conversely, if taxes are “too low” in low tax states, then income and taxes are expected to be negatively related across states. Finally, if taxes are – in the words of Goldilocks – “just right”


Page 11

or at least approximately right, then there is expected to be little or no relationship between taxes and income levels, ceteris paribus. C. Ceteris paribus The predicted relationship between taxes and incomes depends crucially on what is held constant. In particular, higher taxes clearly reduce utility, conditional on the quantity and quality of public services. But in fact most empirical studies – including this one – do not hold public services constant. The main problem is that while expenditures on various public services are easily measured, the quality and even quantity of public services are not. For example, expenditures on K-12 education are the largest single component of state and local government spending, but the connection between spending and student learning is not very tight (Hanushek, 1986). In addition, issues of causality inevitably arise: Do high transfer payments cause low incomes, or do high transfer payments result from low incomes? From an unconditional perspective, high tax states are also generally high expenditure states, and the question is whether the high taxes levied in some states – together with the expenditures they finance – lower utility at the margin. If taxes are “too high” in this broader sense, then the model predicts that taxes and per capita incomes will be positively related across states – higher per capita incomes compensate for excessively high taxes. The utility model is based upon two crucial assumptions—migration is costless and people are identical. Reality does not perfectly meet the theoretical assumptions. First, if migration is costly, it takes time to adjust to the long run equilibrium, so that utilities are not exactly equalized in any given year. This study uses observations ten years apart, i.e. 1959, 1969, 1979, 1989, and 1999. As an empirical matter, most state policies have been fairly stable over time.


Page 12

Thus, ten years is a fairly long period during which a great deal of adjustment would be expected to take place. On the other hand, previous work has indicated slow rates of convergence of per capita incomes – about two and one- half per cent per year – so that adjustments may well be incomplete (Barro and Sala-I-Martin, 1992). A second violation of the assumptions is that people are not identical. They may sort themselves according to preferences, or economic status. Some people may choose a lowincome state, because they value fresh air, safe communities, and pleasant work conditions more than good pay. Others, with different tastes, choose to live in Los Angeles. Thus, the analysis must be interpreted as applying to the marginal individual who is just indifferent between moving and staying, while inframarginal individuals earn location rents. A more serious issue is that state tax systems differ significantly in their structure. Steeply rising marginal rates of income tax may discourage higher income individuals from residing in a state, which lowers average income. Similarly, businesses may choose to locate in lower tax states, reducing the demand for labor and workers wages. Thus, high marginal tax rates and taxes levied on businesses may lower average income, even if the overall level of taxation is not especially high. III. Empirical Specification and Data We examine the tax-income relationship via two empirical specifications. The first relates taxes to the level of income and the second to the growth in income. The model in levels relates income per capita to tax and other variables. yit = α ttaxit + βt xit + uit


Page 13

where yit is the natural logarithm of the real per capita income level of state i in year t, tax it is one or more tax variables of state i in year t, x it represents other factors affecting incomes, and uit is a disturbance term. Separate regressions are estimated for each year, 1959, 1969, 1979, 1989, and 1999, as well as a pooled regression that restricts the coefficients to be the same in every decade (except the intercepts). Data are sampled at ten- year intervals for several reasons: First, the paper’s focus is on the determinants of long run income levels as opposed to short run fluctuations. The years included in the sample are close to business cycle peaks, eliminating some cyclical effects. 9 The decades also correspond at least roughly to distinct periods of growth and change in the U.S. economy. Food and energy prices rose sharply in the 1970s before declining in the 1980s, with significant effects on incomes in some states. Finally some of the tax data for the early years are only available for years near the decennial censuses. 10 Income growth is measured as the decade-long change in the natural logarithm of real per capita personal income, multiplied by ten. g it = ( yit − yit−10 )*10 Thus, the growth rate, g, is (approximately) the annual percentage rate of change of real per capita income. The empirical model relates growth to prior income, tax and other variables. g it = λ yit−10 + α 1taxit + α 2 zit + vit ; t = 1969, 1979, 1989, 1999 The first term on the right hand side is the logarithm of per capita income at the beginning of the decade. The coefficient λ measures the rate of convergence, conditional on the other variables

9

A partial exception is 1959, which is in the recovery phase from the 1958 recession. The relevant peaks are February 1961, November 1970, July 1980, March 1991, and March 2001. 10 See appendix A.


Page 14

in the equation. Lambda is expected to be negative, indicating that higher income states have slower rates of growth than lower income states, ceteris paribus. Income at the beginning of the decade is likely to be correlated with the disturbance term for several reasons. First, it is directly involved in the definition of the dependent variable. Thus any measurement errors in yit −10 will create a spurious negative correlation with the dependent variable. Second, growth may depend on unobserved factors that affect both the level and growth of incomes. For example, a favorable regulatory regime may positively affect both income levels and growth, and thus bias the estimator of lambda upward. Consequently, the growth models treat initial income, yit −10 , as endogenous. Instrumental variables are drawn from the income determinants identified in the level models. An additional complication is that the disturbance terms are correlated over time. c (vit , vit −10 ) ≠ 0 The appropriate estimation technique is then three stage least squares (3SLS). Descriptive statistics are presented in Table 2; here we briefly review the data. Income is measured as per capital personal income, adjusted to dollars of 1999 purchasing power using the CPI –UX1. 11 Because personal income contains large amounts of transfer payments, and because capital income can be earned anywhere almost independently of state of residence, we also experiment with per capita earnings – wage and salary income plus self-employment income – as an alternative measure of income. 12 Two types of tax variables are employed, tax revenues as a percentage of income (average tax rates), and statutory or effective tax rates. Our measures include the average tax rate 11

Economic Report of the President, 2001, Table B-62.


Page 15

including taxes from all sources (Total Tax), as well as average tax rates based on property, sales and income revenues separately. Average tax rates are the broadest measures of taxes, but they may depend on income levels and growth as well as affect them. The statewide statutory general sales tax rate is a less comprehensive measure of sales taxes, because it omits local sales taxes and cannot capture the breadth of goods and services that are subject to tax. However, statutory sales tax rates are less dependent on incomes than are revenues. 13 Similarly, the data set includes the top statutory marginal tax rate from the state’s personal income tax, and the average effective property tax rate on residential housing. The latter variable measures the property tax rate as the ratio of property taxes paid to the market value of housing for an average or representative household. 14 These statutory and effective rates are sometimes used as instrumental variables for the average tax rates, and sometimes entered directly as right hand side variables as alternative measures of tax policy. Incomes depend on a variety of other factors besides taxes. Metropolitan areas have higher incomes and higher cost of living than non- metro areas (Glaeser and Mare, 2001). States also differ in their energy costs, which are measured by heating plus cooling degree days for the largest city in each state, multiplied by the producer price index for energy divided by the CPIUX1. States with a large percentage of children in the population would be expected to have lower per capita incomes, ceteris paribus. At the beginning of the period under study, people over 65 years of age had above average poverty rates; by the end of the study period, the elderly

12

In addition, Gyourko and Tracy (1989) show that taxes may affect wages and rents differently. It would be nice if one could say that tax rates are exogenous. But they are clearly jointly determined with income levels in the larger scope of things. Most recently, states cut tax rates in the late 1990s as their treasuries swelled with the proceeds of strong economic growth and a booming stock market. Tax rates are again rising in the early 2000s in response to revenue shortfalls and expenditure increases stemming in part from slow growth. 14 Prior to 1999, the average effective property tax rate is estimated from FHA mortgages (Advisory Commission on Intergovernmental Finances, various years). Estimates from 1999 are from the Minnesota Taxpayers Association. 13


Page 16

have lower poverty rates than the general population. Education has strong effects on earnings, and the rewards to higher education declined in the 1970s before rising strongly in the last two decades (Freeman, 1976). The economic structure of a state – together with fluctuations in world prices and technological change – has a strong influence on incomes in at least some states. In general farm and natural resource states did well in the 1970s when commodity prices were unusually high, but the long term downward trend in relative prices resumed in the 1980s. 15 Employment in manufacturing grew in the 1960s before stagnating in the 1970s and declining in the last two decades. The “Rust Belt” states of the Midwest were most strongly affected, but manufactur ing is diverse and spread widely across the country. These effects are represented in the level regressions by variables measuring the percentage shares of real income in each state that are derived from farming, mining, and manufacturing. INCitj TOTINCit Share = 100* / , PPI tj CPIt j it

where INCitj is state i’s personal income originating in sector j in year t, TOTINC it is its total personal income in that year, PPI tj is the producer price index for sector j in year t, and CPIt is the consumer price index for year t. 16 Thus, a state which derives a larger percentage of its income from, say, farming will have a larger value for Share far min g . The share variables are included in the models of income levels.

15

Caselli and Coleman (2001) examine the transformation of the south eastern states out of low wage agriculture, and the convergence of wages in agricultural and nonagricultural employment. 16 PPI-Farm Products, PPI-Fuels & Power, and PPI-Finished Goods (excluding consumer goods): Economic Report of the President, 2001, Tables B-65 and 67. PPI-Finished Goods for 1959 is extrapolated from 1969 using the proportional change in PPI for all finished goods.


Page 17

As the relative prices of agricultural, mining and manufacturing products change over time, income growth in the states is differentially affected, depending on the industries’ importance in the various states. To capture these effects we construct a Terms of Trade variable based on changes in relative prices and the shares. 3

Termsof Tradeit = ∑ ∆Pt j *Shareitj −10 , j =1

where ∆Pt j =

PPItj CPIt − PPItj−10 CPI t−10 . PPIt j−10 CPIt −10

Thus, the ∆Pt j measure the percentage changes in the relative prices of farm, mining and manufacturing outputs over the previous decade, and the Terms of Trade variable is a weighted average of these with weights equal to the importance of the respective industries in each state at the beginning of the decade. IV. Empirical Results Table 3 displays estimates of the relationship between the level of per capita income on the one hand, and the average tax rate (“Total Tax”) and other variables on the other. The first five columns refer to separate OLS estimation for each year of the sample. The tax coefficients are uniformly small and insignificant. 17 The last column refers to pooled estimation – the coefficients are restricted to be the same in every period, except for separate intercepts for each year. The pooled estimates are more efficient if the coefficients are in fact equal, but this may not be true because the effects of age, education and economic structure have varied over time. In 17

Recall that taxes are measured as a percentage of income and have a mean of about ten percent. Thus, a ten percent increase in taxes amounts to a unit change in the tax variable. The point estimate for 1959, for example,


Page 18

any case, the tax coefficient from the pooled results is also small and insignificant. There is no relationship between income and taxes – neither positive nor negative – apparent in this specification. Most of the other coefficients in Table 3 are consistent with theoretical predictions. Incomes are higher in more metropolitan states, and lower when greater percentages of the population are less than 18 years of age or over 65. State with high proportions of adults with less than a high school education experience lower incomes. College is associated with higher income (except in 1979), although the effect is not always statistically significant. The farming and mining variables display significant, positive coefficients at the end of the 1970s, but are otherwise insignificant. Manufacturing is never significant. These results are consistent with a long run equilibrium in which farming and mining earn normal rates of return, but are subject to substantial fluctuations over shorter periods of time. The energy cost variable is always positive and significant four times, consistent with the idea that incomes compensate for higher living costs. Table 4 displays two stage least squares estimates of the same model. Statutory sales and income tax rates and effective property tax rates are used as instrumental variables for the average tax rate, Total Tax. The tax coefficients remain small, of mixed sign, and mostly insignificant. The point estimate from the pooled sample implies that a ten percent increase in taxes is associated with lower income of about nine-tenths of one percent. The coefficients of the other variables are also largely unaffected by the use of two stage least squares. Taxes are disaggregated into property, sales and income components in Table 5. The property tax coefficients are mostly positive and one is actually statistically significant at the five

implies that income would be lower by about one-half of one percent.


Page 19

percent level. Of course one would expect about one in 20 estimated coefficients to be significant, even if all of the true values are zero. The sales tax coefficients are mostly negative but none is significantly different from zero. The income tax coefficients are all negative, but none of these is statistically significant either. A joint test of the null hypothesis that all of the tax coefficients are zero cannot be rejected in any of the annual cross sections or in the pooled sample (see Table 8). The point estimates of both the sales and income tax coefficients suggest that an increase in the average tax rate of one percent of income is associated with a decrease in income of .6 to .9 percent. The coefficients of the other variables in the model are little affected by disaggregating taxes into the three components. The estimates of the disaggregated model are also similar when two stage least squares is used (results available upon request). Table 6 reports estimates when the effective property tax rate, the statutory general sales tax rate, and the top statutory marginal tax rate are entered directly as right hand side variables. None of the estimated tax coefficients from the annual regressions are statistically significant, but in the pooled sample the effective property tax rate is positive and significant at the ten percent level, while the income tax rate is negative and significant at the five percent level. The null hypothesis that all of the tax coefficients are zero is rejected in the pooled sample, but not in any of the individual cross sections (Table 9). Effective property tax rates average about one percent of market value. Thus the estimated coefficient in the pooled sample implies that a doubling of the property tax rate is associated with about a two percent increase in per capita income. A one percentage point increase in the top marginal income tax rate is associated with a decline in per capita income of about .3%.


Page 20

Table 7 displays the results when the dependent variable is per capita earnings, which may be more affected by taxes than the broader personal income. The estimated coefficients of Total Tax are mostly negative but none are significantly different from zero. In the pooled sample, the estimated coefficient implies that a 10 percent increase in taxes is associated with a 1.5 percent decline in income. The remaining coefficients are consistent with earlier results in terms of sign and statistical significance. Interestingly, the age and education variables have larger percent age effects on earnings than on personal income, which is consistent with their interpretation as productivity measures, and with transfer payments flowing to people with low earnings. Table 10 displays results from estimating the growth model. It turns out that it is more difficult to explain growth than levels of income, and following previous studies, the coefficients are restricted to be the same in each period apart from the constant term. The coefficients in the first column are estimated using the Seemingly Unrelated Regressions (SUR) method, which takes account of the fact that state- level growth is correlated over time but not the endogeneity of initial income or taxes. The estimated coefficient on initial income is highly significant and implies a rate of convergence of about three percent per year. When the same equation is estimated by three stage least squares, the rate of convergence declines slightly to about 2.7 percent per year. 18 These rates imply that while per capita incomes do display a tendency toward convergence, the speed is quite slow and thus incomes may remain depressed for a significant period of time after an adverse shock. For example, about one- half the effect of a shock to income remains after 25 years. 18

The instrumental variables for lagged income include most of the lagged values of the exogenous variables from the level regressions, namely metro, less than 18, over 65, not a high school, and energy cost. The effective property


Page 21

The tax coefficient is small, positive and insignificant, whether estimation is by SUR or 3SLS. The point estimate implies that a ten percent increase in taxes would increase growth by only about .03 percent per year. Only a subset of the variables from the level regressions have much explanatory power for growth. For example, the percentage of the population that is under 18 or over 65 does not contribute much to explaining growth in per capita income, and these variables are dropped from the growth regressions. 19 However, a 10 percentage point increase in the population living in metro areas is associated with faster growth of about .05 percent per year, while a 10 percentage point increase in college is associated with faster growth of about .5 percent per year. Finally, the variable measuring changes in the terms of trade has a statistically significant and positive coefficient: Rising relative prices for food, energy and manufactures increase growth in the states which earn larger shares of income from those industries. Column three of Table 10 includes an additional tax variable, the change in the average tax rate over the previous decade. 20 The estimated coefficient is statistically significant and implies that an increase of one percentage point in the average tax rate is associated with a decline in the annual growth rate of income of about .3 percent during the decade. Vedder’s (1996) estimate from a similar specification for 1960-93 yields an association of about minus .1 percent per year. The final two columns of Table 10 enter the property, sales and income tax rate variables directly into the growth equation. The coefficient of the property tax rate is significant and

tax rate, general sales tax rate and top personal income tax rate are instruments for Total Tax. 19 Changes in these variables are not significant either. 20 Two additional instrumental variables are included in this specification, since the change in tax rate is treated as endogenous: the change in the percentage of population less than 18, which is expected to influence school


Page 22

positive but very small: A 10 percent increase in property taxes implies an increase in growth of about .02 percent per year. A one percentage point increase is the top marginal tax rate, on the other hand, is associated with lower growth of about .03 percent per year. The null hypothesis that the property, sales and income tax coefficients are all zero is rejected at the one percent level in a Wald test (χ2 3 = 18.0) Table 11 splits the sample between 1959-79 and 1979-89, in order to determine whether the factors affecting growth changed. There is no evidence that the rate of conditional convergence declined: in both periods the rate is 2-3 percent per year. The average tax rate variable switches sign from one period to the next and remains insignificant. The relationship between changes in tax and growth is negative for both periods. While the point estimate is considerably larger in the second period, the difference is not statistically significant. When the three tax rates are entered directly, the property tax coefficients continue to be positive and the income tax coefficients negative, although the latter lose their statistical significance. The point estimates for the other variables provide some evidence that the determinants of growth did change between the two periods. The coefficients of Metro and College are larger and more likely to be significant in the second period, while the terms of trade variable loses its statistical significance. This is consistent with the view that the energy and food “crises” of the 1970s had important effects on state level growth, but that growth in the last twenty years has been concentrated in industries located in metropolitan areas and employing college educated labor. Overall, however, growth in the last 20 years is more difficult to explain than growth in the earlier period, as indicated by the lower R2 s.

spending, and an indicator for the presence of tax or expenditure limitation legislation (New, 2001).


Page 23

V. Conclusion A commonly held view is that excessively high taxes cause lower incomes. But if taxes are “too high” in a state – in the sense that the combination of taxes and public services lowers the utility of the residents – then people have an incentive to move to another state. In long run equilibrium, excessively high taxes must be offset by some compensating differential, for otherwise a state will be entirely depopulated. The empirical analysis has sought to uncover a relationship between taxes and income levels, but has failed to find any significant relationship at all, neither negative or positive. The finding of no relationship appears to be robust to a variety of measures of taxes, corrections for endogeneity, and definitions of income. In contrast, urbanization, age structure and education levels explain a great deal of the cross-state variation in incomes. One explanation for these results is that overall state tax levels are neither too low nor too high but rather just about right. This sort of equilibrium could emerge as a result of “tax competition” among the states, which prevents any state from departing too far for too long from near optimal levels (Benson and Johnson, 1986). Taxes need not be equalized across states for this to be true, since states vary considerably in their economic bases (tourism, natural resources, etc), and people may sort themselves among states according to their preferences for public services versus private goods (Tiebout, 1956). Another explanation is that state tax policies are determined by similar political and economic institutions, and thus the outcomes are roughly the same, but not necessarily optimal. Thus, for example, representative democracy may result in excessive – from the standpoint of


Page 24

economic efficiency – redistribution from the rich to the poor, but if this is true in every state, there are no incentives for migration. Taxes are also commonly thought to reduce economic growth, with particularly strong effects from taxation of capital income and high marginal income tax rates. The evidence presented here is mixed in this regard. There is no evidence of a relationship between growth in per capita income and the level of taxes, but increases in average tax rates are associated with lower growth. Unfortunately, the direction of causality is not clear: Political pressures for government expenditures and/or Federal matching formulas for programs targeted at the low income population may have caused taxes to increase faster in states with slower income growth. State tax cuts in the 1990s certainly appeared to be a response to growth, rather than a cause, and currently (2003) many states are considering tax increases in response to budget shortfalls resulting in part from slow growth. The mix of property, sales and income taxes appears to have only small effects on income levels and growth. An increase in the top statutory personal income tax rate of one percentage point is associated with a statistically significant reduction in annual growth of about three onehundredths of one percent. On the other hand, a ten percent increase in effective property tax rates is associated with an increase in growth of about the same magnitude. One explanation for these results is that property taxes are largely devoted to local expenditure for schools and infrastructure, while income taxes are more likely to be devoted to redistribution. Thus, higher property taxes together with the associated spending could increase a state’s attractiveness, while increased income taxes discourage the wealthy.


Page 25

There is strong evidence that state incomes tend to slowly converge, conditional on other determinants. Although state incomes ceased becoming more equalized during the 1980s and 1990s, this is because other determinants of income changed. Specifically, growth in the last twenty years has been concentrated in metropolitan areas and in states with high proportions of college educated population. Since these groups already had higher than average incomes, recent growth has tended to increase income inequality among the states, offsetting the effect of conditional convergence.


Page 26

REFERENCES Advisory Commission on Intergovernmental Relations. (1990). Significant Features of Fiscal Federalism, Vol. 1. “Budget Processes and Tax Systems”. M-169. ______. (1991). Significant Features of Fiscal Federalism, Vol. 2. “Revenues and Expenditures”. M-176-II. Barro, R. J. (1997a). Determinants of Economic Growth: A Cross-Country Empirical Study. Cambridge: MIT Press. ______. (1997b). Macroeconomics (5th ed.). Cambridge: MIT Press. ______, and Xavier Sala- i-Martin. (1992). “Convergence.” Journal of Political Economy, 100, 223-51. ______, and Xavier Sala- i-Martin. (1995). Economic Growth. McGraw-Hill Inc., New York, NY. Bartik, T. J. (1992) “The Effects of State and Local Taxes on Economic Development: A Review of Recent Research.” Economic Development Quarterly, 102-11. ______. (1991). Who Benefits from State and Local Economic Development Policies? Kalamazoo: W. E. Upjohn Institute. Becker, G. S., Glaeser, E. L., & Murphy, K. M. (1999). “Population and Economic Growth.” American Economic Review, 89-2: 145-49. Becsi, Z. (1996). “Do State and Local Taxes Affect Relative State Growth?” Federal Reserve Bank of Atlanta Economic Review, 81-2: 18-36. Benson, B. L., & Johnson, R. N. (1986). “The Lagged Impact of State and Local Taxes on Economic Activity and Political Behavior.” Economic Inquiry, 24, 389-401. Caselli, F., & Coleman, W. J., II. (2001). “The U. S. Structural Transformation and Regional Convergence: A Reinterpretation.” Journal of Political Economy, 109, 584-616. Council of State Governments. The Book of the States: 1959-99. Lexington: Council of State Governments. Department of Commerce, Bureau of the Census. (1981). 1980 Census of Population, Supplementary Report, Metropolitan Statistical Areas, PC80-S1-18. DC: U.S. Dept. of Commerce, Bureau of the Census.


Page 27

______. (1993). 1990 Census of Population and Housing, Supplementary Reports (1990 CPH-S-1-1). DC: U.S. Dept. of Commerce, Bureau of the Census. ______. (2000). Current Population Reports, Series P-20, No. 528. DC: U.S. Dept. of Commerce, Economics and Statistics Administration, Bureau of the Census. ______. (1960,1970,1980, and 1990). Current Population Reports, Series P-25, No. 214, 427, 437, 875, 1039, and 1058. DC: U.S. Dept. of Commerce, Economics and Statistics Administration, Bureau of the Census. ______. Statistical Abstract of the United States: 1959-2000. DC: Government Printing Office. ______. U.S. Census of Population: 1960, 1970, 1980, and 1990, Vol. I, Chapter C. DC: U.S. Dept. of Commerce, Bureau of the Census. Fisher, R. C. (1997). “The Effects of State and Local Public Services on Economic Development.” New England Economic Review, 53-67. Glaeser, E. L. and D. C. Mare (2001). “Cities and Skills.” Journal of Labor Economics, 192: 316-342. Gwartney, J., & Stroup, R. (1983). “Labor Supply and Tax Rates: A Correction of the Record.” The American Economic Review, 73, 446-451. Gyourko, J. and J. Tracy (1989) “The Importance of Local Fiscal Conditions in Analyzing Local Labor Markets.” Journal of Political Economy, 97-5: 1208-1231. ______. (1991) “The Structure of Local Public Finance and the Quality of Life.” Journal of Political Economy, 99-4: 774-806. Helms, L. Jay. (1985) “The Effect of State and Local Taxes on Economic Growth: A Time Series-Cross Section Approach.” Review of Economics and Statistics 67-4: 574-582. Koester, R.B., & Kormendi, R.C. (1989). “Taxation, Aggregate Activity, and Economic Growth: Cross-country Evidence on Some Supply-side Hypotheses.” Economic Inquiry, 27: 367-386. Mankiw, N. G. (2000). Macroeconomics (4th ed.). New York: Worth Publishers. Mendoza, Enrique G., Gian Maria Milesi-Ferretti, and Patrick Asea. (1997) “On the ineffectiveness of tax policy in altering long-run growth: Harberger’s superneutrality conjecture.” Journal of Public Economics 66: 99-126.


Page 28

Minnesota Taxpayers Association. (1999). 50-state Property Tax Comparison Study, Payable Year 1998. Mullen, J. K., & Williams, M. (1994). “Marginal Tax Rates and State Economic Growth.” Regional Science and Urban Economics, 24, 687-705. National Center for Policy Analysis. (2001, November 14). “Cities and Human Capital.” NCPA Policy Digest. National Oceanic and Atmospheric Administration. Comparative Climatic Data, annual. New, M. J. (2001, December 13). “Limiting Government through Direct Democracy: The Case of State Tax and Expenditure Limitations.” Policy Analysis, No. 420. Cato Institute. Phillips, J., & Goss, E. (1995). “The Effect of State and Local Taxes on Economic Development: A Meta-analysis.” Southern Economic Journal, 62, 320-333. Pindyck, R. S., & Rubinfeld, D. L. (1997). Econometric Models and Economic Forecasts (4 ed.). New York: Irwin/McGraw-Hill. th

Power, T., & Barrett, R. (2001). Post-Cowboy Economics: Pay and Prosperity in the New American West. Washington: Island Press. Roger, S. (1982). Making Business Location Decisions. Englewood Cliffs: Prentice-Hall. Rosen, H. S. (2002). Public Finance (6th ed.). New York: McGraw-Hill. Solow, R. M. “A Contribution to the Theory of Economic Growth.” Quarterly Journal of Economics, 70, 65-94. Slemrod, J. (2000) Taxing Ourselves (2d Ed.) Cambridge: MIT Press. Stokey, N.L. and S. Rebelo (1995) “Growth Effects of Flat-Rate Taxes.” Journal of Political Economy, 103-3:519-550. Tiebout, C. (1956) “A Pure Theory of Local Expenditures.” Journal of Political Economy, 84: 1145-1160. U.S. Census Bureau. [Online] Available: http://www.census.gov/population/estimates/metro-city/ma99-06.txt [2000, October 20], http://www.census.gov/population/estimates/state/st-99-08.txt [2000, March 9], http://www.census.gov/population/estimates/state/st-99-09.txt [2000, March 9].


Page 29

_____. (1960). Government Finances in 1959, Series GF-59-No. 2. DC: U.S. Government Printing Office. _____. (1970). Government Finances: 1968-69, Series GF-69-5. DC: U.S. Government Printing Office. _____. (1980). Government Finances: 1978-79, Series GF-79-5. DC: U.S. Government Printing Office. _____. (1991). Government Finances: 1988-89, Series GF-89-5. DC: U.S. Government Printing Office. _____. (1960). State Tax Collections in 1959. DC: U.S. Government Printing Office. _____. (1970). State Government Finances in 1969, Series GF69-No. 3. DC: U.S. Government Printing Office. U.S. Council of Economic Advisors. Economic Report of the President, 2001, Tables B-62, 65 and 67. DC: U.S. Government Printing Office. U.S. Department of Commerce, Bureau of Economic Analysis. [Online] Available: http://www.bea.doc.gov/bea/regional/spi/ [2001, April 14]. U.S. National Oceanic and Atmospheric Administration, Comparative Climatic Data. Vedder, R. K. (1996). State and Local Taxation and Economic Growth: Lessons for Federal Tax Reform. A Study Prepared for the Use of the Joint Economic Committee, Congress of the United States, 104th Congress, 1st Session, December. Washington: U.S. G.P.O.: For Sale by the U.S. G.P.O., Supt. of Docs., Congressional Sales Office. Wasylenko, M. (1997). “Taxation and Economic Development: The State of the Economic Literature.” New England Economic Review, 37-52. Yamarik, S. “Can Tax Policy Help Explain State-level Macroeconomic Growth?” Economics Letters, 68: 211-215.

Taxes, Growth and the Convergence of Incomes Among U.S. States: 1959-99 Figure 2a Utility and the Level of Taxes/Public Services

Utility

T*

Taxes (and Services)

Figure 2b An Indifference Curve

Income

y1 y0 T*

T1

Taxes (and Services)

Page 1


State AL AZ AR CA CO CT DE FL GA ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT

Table 1: Per Capita Personal Incomes: Levels and Growth 1959-1999 (in constant 1999 dollars) Rank Growth Rank 1959 1999 1999 59-99 State 1959 1999 1999 8,045 22,987 39 2.6% NE 10,660 27,049 20 10,407 25,189 35 2.1% NV 14,493 31,022 14 7,444 22,244 45 2.6% NH 11,198 31,114 6 14,446 29,910 10 1.8% NJ 14,024 35,551 2 11,920 31,546 8 2.4% NM 9,991 21,853 46 14,931 39,300 1 2.4% NY 14,208 33,890 4 14,361 30,778 12 1.8% NC 8,335 26,003 30 10,597 27,780 21 2.3% ND 8,852 23,313 36 8,752 27,340 22 2.8% OH 12,200 27,152 19 9,970 22,835 41 2.0% OK 9,790 22,953 40 13,929 31,145 7 2.0% OR 11,831 27,023 25 11,235 26,143 28 2.1% PA 11,810 28,605 15 10,681 25,615 31 2.1% RI 11,509 29,377 17 11,103 26,824 23 2.2% SC 7,339 23,545 37 8,457 23,237 38 2.5% SD 8,246 25,045 33 8,884 22,847 43 2.3% TN 8,425 25,574 32 9,753 24,603 34 2.3% TX 10,270 26,858 24 12,252 32,465 5 2.4% UT 10,386 23,288 42 12,811 35,551 3 2.5% VT 9,659 25,889 29 12,432 28,113 16 2.0% VA 9,938 29,789 13 10,887 30,793 9 2.6% WA 12,548 30,392 11 6,564 20,688 48 2.8% WV 8,420 20,966 47 11,209 26,376 27 2.1% WI 11,731 27,390 18 10,666 22,019 44 1.8% WY 12,010 26,396 26 United States 11,678 28,542 -

Page 2

Growth 59-99 2.3% 1.7% 2.6% 2.3% 1.9% 2.1% 2.8% 2.4% 2.0% 2.1% 2.0% 2.2% 2.2% 2.9% 2.8% 2.7% 2.4% 1.9% 2.5% 2.7% 2.2% 2.3% 2.1% 1.9% 2.2%


Page 3

Table 2a: Descriptive Statistics for 1959 N

Mean

Std Dev

Minimum

Maximum

48

9.27

0.20

8.79

9.61

48

9.07

0.19

8.61

9.42

N.A.

N.A.

N.A.

N.A.

N.A.

48 N.A. 48

8.67 N.A. 3.83

1.28 N.A. 1.34

6.61 N.A. 1.45

12.01 N.A. 6.67

Sales Tax Income Tax Unbalance

48 48 48

1.10 0.72 0.24

0.88 0.68 0.16

0.00 0.00 0.02

3.53 2.68 0.67

Property Rate Sales Rate Income Rate

48 48 48

1.27 1.79 3.94

0.46 1.35 3.66

0.48 0.00 0.00

2.21 4.00 11.00

N.A. N.A. N.A.

N.A. N.A. N.A.

N.A. N.A. N.A.

N.A. N.A. N.A.

N.A. N.A. N.A.

Metro Pop Less Than 18 65 And Over

48 48 48

51.99 36.87 8.80

25.82 2.93 1.64

0.00 31.52 5.67

97.40 44.87 11.62

Not A High School College And Higher Sharefarming

48 48 48

58.23 7.52 4.39

7.24 1.56 3.23

44.39 4.77 0.34

72.42 10.73 12.59

48 48 N.A.

4.39 20.02 N.A.

6.06 8.90 N.A.

0.11 3.38 N.A.

28.82 35.79 N.A.

48

4298.77

970.05

1514.29

6553.37

Dependent Variable log y log PC Earning Growth Rate Independent Variable Total Tax ?Total Tax Property Tax

?Property Rate ?Sales Rate ?Income Rate

Sharemining Sharemanufacturing ∆Terms of Trade Energy Cost


Page 4

Table 2b: Descriptive Statistics for 1969 N

Mean

Std Dev

Minimum

Maximum

48

9.61

0.17

9.23

9.93

log PC Earning

48

9.40

0.16

9.04

9.70

∆Growth Rate

Dependent Variable log y

48

3.37

0.59

2.12

4.88

Independent Variable Total Tax

48

9.76

1.15

7.90

12.65

?Total Tax

48

1.10

0.82

-0.39

3.38

Property Tax

48

3.81

1.26

1.35

6.71

Sales Tax Income Tax

48 48

1.66 1.26

0.86 1.00

0.00 0.00

3.89 3.43

Unbalance

48

0.14

0.11

0.00

0.62

Property Rate Sales Rate

48 48

1.91 3.23

0.67 1.30

0.56 0.00

3.15 6.00

Income Rate

48

5.42

4.43

0.00

17.94

?Property Rate

48

0.64

0.35

-0.49

1.33

?Sales Rate ?Income Rate

48 48

1.44 1.48

1.18 2.51

0.00 -2.00

5.00 10.44

Metro Pop

48

63.12

23.04

15.70

100.00

Less Than 18

48

35.68

2.13

32.62

42.15

65 And Over Not A High School

48 48

9.73 47.37

1.72 7.88

6.13 32.52

13.31 62.23

College And Higher

48

10.43

2.08

6.62

14.90

Share

farming

48

3.65

3.70

0.21

17.07

Sharemining Sharemanufacturing

48 48

3.90 20.99

5.85 8.89

0.28 3.66

26.00 38.11

∆Terms of Trade

48

-2.65

0.76

-6.19

-1.32

Energy Cost

48

3904.30

881.04

1375.33

5952.00


Page 5

Table 2c: Descriptive Statistics for 1979 N

Mean

Std Dev

Minimum

Maximum

48

9.88

0.13

9.60

10.11

log PC Earning

48

9.62

0.14

9.34

9.86

Growth Rate

48

2.76

0.65

1.38

4.51

Total Tax

48

9.80

1.23

8.00

13.81

?Total Tax

48

0.04

0.95

-2.55

2.12

Property Tax

48

3.12

1.14

0.98

5.79


Independent Variable

Sales Tax

48

2.19

1.01

0.00

4.19

Income Tax

48

2.10

1.22

0.00

4.83

Unbalance

48

0.07

0.08

0.00

0.39

Property Rate

45

1.26

0.58

0.28

2.75

Sales Rate

48

3.63

1.50

0.00

7.50

Income Rate

48

6.48

4.66

0.00

17.50

?Property Rate

45

-0.61

0.46

-1.44

0.72

?Sales Rate

48

0.40

0.71

-1.00

2.50

? Income Rate

48

1.06

2.54

-3.50

13.30

Metro Pop

48

65.68

22.40

24.00

100.00

Less Than 18

48

29.07

2.06

25.10

37.00

65 And Over

48

11.13

1.81

7.70

18.10

Not A High School

48

32.98

7.34

20.00

46.90

College And Higher

48

15.90

2.87

10.40

23.00

Sharefarming

48

1.60

1.70

0.07

8.68

Sharemining

48

2.62

4.27

0.22

23.79

Sharemanufacturing

48

17.48

6.94

4.10

32.10

∆Terms of Trade

48

6.74

6.46

1.81

31.81

Energy Cost

48

5507.07

1242.72

1939.93

8395.39


Page 6

Table 2d: Descriptive Statistics for 1989 N

Mean

Std Dev

Minimum

Maximum

48

10.05

0.16

9.73

10.45

log PC Earning

48

9.71

0.19

9.37

10.15

Growth Rate

48

1.66

0.97

-0.78

3.47

48

10.05

1.08

7.82

13.87

Property Tax

48

3.06

1.09

1.01

5.50


Independent Variable Total Tax ?Total Tax Sales Tax

48

2.37

1.03

0.00

5.00

Income Tax

48

2.40

1.23

0.00

5.12

Unbalance

48

0.07

0.09

0.00

0.43

Property Rate

47

1.18

0.52

0.22

2.38

Sales Rate

48

4.45

1.68

0.00

8.00

Income Rate

48

5.45

3.34

0.00

14.57

?Property Rate

45

-0.10

0.37

-1.59

0.70

?Sales Rate

48

0.83

0.85

-1.00

3.00

?Income Rate

48

-1.02

2.93

-10.50

7.07

Metro Pop

48

67.57

21.26

26.90

100.00

Less Than 18

48

26.19

2.39

22.61

36.97

65 And Over

48

12.64

1.73

8.60

18.00

Not A High School

48

24.01

5.51

14.90

35.70

College And Higher

48

19.62

3.77

12.30

27.20

Sharefarming

48

1.69

1.78

0.14

7.47

Sharemining

48

2.04

3.64

0.04

21.15

Sharemanufacturing

48

15.23

5.77

3.58

27.39

∆Terms of Trade

48

-3.07

1.04

-7.19

-0.90

Energy Cost

48

4717.12

1064.46

1661.66

7191.14


Page 7

Table 2e: Descriptive Statistics for 1999 N

Mean

Std Dev

Minimum

Maximum

48

10.20

0.15

9.94

10.58

log PC Earning

48

9.84

0.18

9.49

10.27

Growth Rate


48

1.50

0.40

0.70

2.36

Independent Variable Total Tax

48

10.48

1.07

8.47

13.63

?Total Tax

48

0.43

0.70

-1.88

2.23

Property Tax

48

3.11

1.09

1.13

5.60

Sales Tax Income Tax

48 48

2.52 2.63

1.11 1.29

0.00 0.00

5.02 5.05

Unbalance

48

0.07

0.09

0.00

0.45

Property Rate Sales Rate

48 48

1.18 4.76

0.56 1.72

0.38 0.00

2.93 7.00

Income Rate

48

5.72

3.22

0.00

12.00

?Property Rate

47

0.00

0.45

-1.49

1.00

?Sales Rate ?Income Rate

48 48

0.31 0.27

0.61 1.26

-2.00 -2.80

2.00 4.50

Metro Pop

48

68.25

20.73

27.90

100.00

Less Than 18

48

25.71

1.78

22.36

33.19

65 And Over Not A High School

48 48

12.81 15.51

1.68 4.21

8.70 8.80

18.10 24.90

College And Higher

48

24.64

4.65

17.30

38.70

Share

farming

48

1.52

1.78

-0.06

9.23

Sharemining Sharemanufacturing

48 48

1.75 14.46

3.19 5.27

0.02 4.02

18.79 27.41

∆Terms of Trade

48

-2.75

0.79

-4.60

-1.05

Energy Cost

48

3968.89

895.61

1398.09

6050.48


Page 8

Table 3. Total Taxes: OLS Estimation D. Dependent Variable: the Natural Logarithm of Real Per Capita Personal Income Explanatory Variable

Regression Coefficients 1979 1989

1959

1969

1999

Pooled

Total Taxit

-0.00481 (-0.34)

0.00662 (0.55)

0.00103 (0.16)

0.00128 (0.15)

0.00879 (0.83)

-0.00326 (-0.70)

Metro Popit

0.00230** (2.31)

0.00382*** (4.61)

0.00329*** (6.69)

0.00343*** (5.30)

0.00682*** (6.91)

0.00379*** (10.87)

Less Than 18it

-0.0390*** (-3.74)

-0.0319*** (-3.60)

-0.0405*** (-7.85)

-0.0365*** (-6.72)

-0.0417*** (-4.24)

-0.0301*** (-9.11)

65 And Overit

-0.0330* (-1.91)

-0.0260** (-2.16)

-0.0306*** (-4.94)

-0.0132* (-1.71)

-0.0246** (-2.38)

-0.0153*** (-3.46)

Not A High School it

-0.0111*** (-3.38)

-0.0083*** (-3.31)

-0.0104*** (-6.65)

-0.00384 (-1.45)

-0.00653* (-1.75)

-0.0086*** (-8.40)

College And Higherit

0.0158 (0.98)

0.0098 (0.98)

-0.00843* (-1.88)

0.0138*** (3.46)

0.00434 (1.22)

0.0074*** (3.49)

Shareitfar min g

-0.00585 (-0.71)

0.00477 (0.76)

0.0195*** (3.07)

0.00662 (0.90)

0.0205** (2.19)

-0.00299 (0.98)

Shareitmin ing

0.00125 (0.45)

-0.00184 (-0.71)

0.00572*** (2.60)

0.00215 (0.62)

0.00344 (0.84)

0.00145 (1.09)

Shareitmanufacturing

-0.00070 (-0.28)

0.00246 (1.17)

-0.000178 (-0.12)

0.00189 (0.97)

0.000413 (0.17)

0.00069 (0.73)

Energy Costit

0.00004** (2.05)

0.00004*** (2.57)

0.00001 (1.31)

0.00002 (1.34)

0.00005*** (2.64)

0.00003*** (4.07)

Intercept

11.333*** (16.93)

10.745*** (21.14)

11.552*** (41.29)

10.632*** (33.81)

10.784*** (28.75)

10.776*** (64.35)

Year Dummies R2

Included 85%

84%

89%

Notes: t statistics are in parentheses. * statistically significant at the .10 level for two-tailed test. ** statistically significant at the .05 level for two-tailed test. *** statistically significant at the .01 level for two-tailed test.

90%

83%

96%


Page 9

Table 4. Total Taxes: 2SLS Estimation Dependent Variable: the Natural Logarithm of Real Per Capita Personal Income Explanatory Variable


1959

1969

1999

Pooled

Total Taxit

-0.0194 (-0.69)

0.00307 (0.17)

0.00198 (0.19)

-0.00752 (-0.58)

-0.02557 (-1.35)

-0.00889 (-1.06)

Metro Popit

0.00227** (2.25)

0.00382*** (4.60)

0.00339*** (5.86)

0.00334*** (4.35)

0.00622*** (5.42)

0.00376*** (10.23)

Less Than 18it

-0.0365*** (-3.21)

-0.0315*** (-3.52)

-0.0403*** (-7.61)

-0.0361*** (-6.41)

-0.0372*** (-3.28)

-0.0298*** (-8.96)

65 And Overit

-0.0280 (-1.44)

-0.0254** (-2.06)

-0.0296*** (-4.53)

-0.0128 (-1.60)

-0.0188 (-1.57)

-0.0145*** (-3.26)


-0.0114*** (-3.39)

-0.00840*** (-3.30)

-0.0100*** (-6.06)

-0.00425 (-1.55)

-0.00562 (-1.32)

-0.00860*** (-8.17)


0.0162 (0.99)

0.0103 (1.01)

-0.00840 (-1.64)

0.0144*** (3.39)

0.00600 (1.46)

0.00793*** (3.62)

Shareitfar min g

-0.00554 (-0.67)

0.00512 (0.80)

0.0187** (2.64)

0.00586 (0.75)

0.0136 (1.24)

0.00273 (0.88)

Shareitmin ing

0.00113 (0.40)

-0.00170 (-0.64)

0.00571** (2.47)

0.00288 (0.77)

0.00282 (0.60)

0.00153 (1.12)


-0.00146 (-0.52)

0.00244 (1.16)

-0.00029 (-0.19)

0.00202 (1.01)

0.000154 (0.06)

0.000603 (0.62)

Energy Costit

0.00004** (2.11)

0.00004** (2.52)

0.00001 (1.36)

0.00002 (1.38)

0.00006*** (2.78)

0.00003*** (4.33)

Intercept

11.339*** (16.69)

10.763*** (20.97)

11.502*** (38.80)

10.706*** (32.15)

10.910*** (25.43)

10.802*** (60.14)

Year Dummies R2

Included 85%

84%

90%

90%

Notes: t statistics are in parentheses. * statistically significant at the .10 level for two-tailed test. ** statistically significant at the .05 level for two-tailed test.

*** statistically significant at the .01 level for two-tailed test.

79%

96%


Page 10

Table 5. Property, Sales and Income Taxes: OLS Estimation E. Dependent Variable: the Natural Logarithm of Real Per Capita Personal Income Explanatory Variable


1959

1969

1999

Pooled

Property Taxit

-0.000304 (-0.02)

0.00291 (0.18)

0.0153 (1.62)

0.00888 (0.78)

0.0323** (2.57)

0.00545 (0.89)

Sales Taxit

-0.00558 (-0.25)

-0.0154 (-0.84)

-0.0111 (-1.14)

-0.00698 (-0.62)

0.00316 (0.26)

-0.00880 (-1.32)

Income Taxit

-0.00479 (-0.17)

-0.00494 (-0.34)

-0.00547 (-0.78)

-0.00360 (-0.40)

-0.00184 (-0.18)

-0.00642 (-1.18)

Metro Popit

0.00231** (2.20)

0.00389*** (4.36)

0.00354*** (6.90)

0.00358*** (5.34)

0.00677*** (7.27)

0.00385*** (10.90)

Less Than 18it

-0.0389*** (-3.49)

-0.0298*** (-3.27)

-0.0376*** (-6.99)

-0.0351*** (-6.17)

-0.0385*** (-3.93)

-0.0284*** (-8.25)

65 And Overit

-0.0338* (-1.88)

-0.0253** (-2.05)

-0.0322*** (-5.22)

-0.0137* (-1.73)

-0.0247** (-2.54)

-0.0153*** (-3.46)


-0.0110*** (-2.90)

-0.00863*** (-2.95)

-0.00984*** (-6.37)

-0.00303 (-1.09)

-0.00450 (-1.25)

-0.00807*** (-7.50)


0.0161 (0.84)

0.00794 (0.67)

-0.0104** (-2.28)

0.0129*** (3.02)

0.00409 (1.14)

0.00709*** (3.19)

Shareitfar min g

-0.00571 (-0.65)

0.00581 (0.86)

0.0225*** (3.55)

0.00691 (0.93)

0.0222** (2.51)

0.00321 (1.04)

Shareitmin ing

0.00132 (0.45)

-0.00103 (-0.38)

0.00672*** (2.99)

0.00156 (0.45)

0.00417 (1.04)

0.00156 (1.16)


-0.00031 (-0.12)

0.00257 (1.14)

-0.00003 (-0.02)

0.00169 (0.79)

0.00108 (0.45)

0.00107 (1.09)

Energy Costit

0.00003 (1.58)

0.00004* (1.92)

0.000002 (0.24)

0.00001 (1.06)

0.00004** (2.06)

0.00002*** (2.98)

Intercept

11.296*** (15.86)

10.801*** (20.32)

11.516*** (42.32)

10.621*** (34.14)

10.703*** (30.04)

10.733*** (65.75)

Year Dummies R2

Included 85%

84%

91%


91%

85%

96%


Page 11

Table 6. Property, Sales and Income Tax Rates: OLS Estimation Dependent Variable: the Natural Logarithm of Real Per Capita Personal Income Explanatory Variable


1959

1969

1999

Pooled

Property Rateit

-0.0278 (-0.65)

0.0169 (0.57)

0.0219 (1.43)

0.0189 (0.94)

0.0284 (1.25)

0.0207* (1.87)

Sales Rateit

-0.0106 (-0.94)

-0.00697 (-0.74)

-0.000008 (-0.00)

0.00483 (0.83)

0.00208 (0.34)

-0.00265 (-0.75)

Income Rateit

-0.00131 (-0.29)

0.00043 (0.13)

-0.00152 (-0.73)

-0.00420 (-1.52)

-0.00565 (-1.54)

-0.00282** (-2.00)

Metro Popit

0.00258** (2.45)

0.00399*** (4.27)

0.00308*** (4.86)

0.00278*** (3.52)

0.00578*** (5.60)

0.00354*** (9.15)

Less Than 18it

-0.0387*** (-3.68)

-0.0300*** (-3.27)

-0.0393*** (-7.29)

-0.0357*** (-6.36)

-0.0377*** (-3.84)

-0.0285*** (-8.47)

65 And Overit

-0.0322* (-1.88)

-0.0265** (-2.10)

-0.0292*** (-4.39)

-0.0143* (-1.82)

-0.0220** (-2.12)

-0.0149*** (-3.33)


-0.0111*** (-3.16)

-0.00814*** (-3.10)

-0.0103*** (-6.04)

-0.00428 (-1.62)

-0.00508 (-1.37)

-0.00791*** (-7.70)


0.0160 (0.87)

0.00725 (0.63)

-0.00784 (-1.53)

0.0144*** (3.59)

0.00697* (1.94)

0.00828*** (3.80)

Shareitfar min g

-0.00563 (-0.65)

0.00528 (0.85)

0.0176** (2.59)

0.00497 (0.66)

0.0165* (1.80)

0.00218 (0.71)

Shareitmin ing

0.00061 (0.21)

-0.00098 (-0.34)

0.00636*** (2.74)

0.00128 (0.37)

0.00218 (0.50)

0.00141 (1.02)


-0.00083 (-0.34)

0.00224 (1.03)

-0.00019 (-0.13)

0.00196 (1.03)

0.00021 (0.09)

0.00082 (0.87)

Energy Costit

0.00004* (1.85)

0.00004** (2.08)

0.000008 (1.00)

0.00001 (0.80)

0.00004** (2.26)

0.00002*** (3.31)

Intercept

11.310*** (16.25)

10.762*** (20.50)

11.509*** (38.60)

10.693*** (34.44)

10.752*** (29.13)

10.701*** (66.01)

Year Dummies R2

Included 85%

84%

90%


91%

85%

96%


Page 12

Table 7. Total Taxes: 2SLS Estimation Dependent Variable: the Natural Logarithm of Real Per Capita Personal Earning Explanatory Variable


1959

1969

1999

Pooled

Total Taxit

-0.0214 (-0.77)

0.00005 (0.00)

-0.0045 (0.16)

-0.0128 (-0.79)

-0.0300 (-1.39)

-0.0152 (-1.64)

Metro Popit

0.0019* (1.92)

0.0036*** (4.39)

0.0036*** (5.88)

0.0035*** (3.63)

0.0071*** (5.39)

0.0038*** (9.31)

Less Than 18it

-0.0355*** (-3.17)

-0.0351*** (-4.00)

-0.0430*** (-7.59)

-0.0417*** (-5.92)

-0.0412*** (-3.17)

-0.0321*** (-8.73)

65 And Overit

-0.0367* (-1.92)

-0.0455*** (-3.77)

-0.0508*** (-7.28)

-0.0370*** (-3.69)

-0.0404** (-2.94)

-0.0322*** (-6.54)


-0.0122*** (-3.67)

-0.0083*** (-3.32)

-0.0091*** (-5.13)

-0.0026 (-0.76)

-0.0058 (-1.18)

-0.0075*** (-6.43)


0.0096 (0.60)

0.0058 (0.58)

-0.0076 (-1.39)

0.0190*** (3.58)

0.0084* (1.79)

0.0122*** (5.02)

Shareitfar min g

-0.0083 (-1.02)

0.0053 (0.85)

0.0234*** (3.07)

0.0093 (0.96)

0.0174 (1.38)

0.0063* (1.84)

Shareitmin ing

0.0003 (0.09)

-0.0024 (-0.93)

0.0071*** (2.86)

0.0014 (0.30)

0.0017 (0.31)

0.0015 (1.01)


-0.0010 (-0.35)

0.0029 (1.40)

0.0011 (0.68)

0.0042* (1.68)

0.0036 (1.14)

0.0022** (2.09)

Energy Costit

0.00003* (1.79)

0.00005*** (2.83)

0.00001 (1.41)

0.00002 (1.33)

0.00007*** (2.82)

0.00003*** (4.19)

Interceptit

11.344*** (16.96)

10.937*** (21.76)

11.515*** (36.31)

10.681*** (25.62)

10.767*** (21.89)

10.634*** (53.53)

Year Dummies R2

Included

84%

84%

89%


88%

80%

94%


Page 13

Table 8. Summary of F-tests on OLS Estimates Null Hypothesis: The joint effect of property taxes, sales taxes, and income taxes on income levels is zero.

Model 1959 1969 1979 1989 1999 Pooled

Statistic

Result, .05 significant level

F(3, 35) = 0.0232

Can’t reject

F(3, 35) = 0.287

Can’t reject

F(3, 35) = 1.417

Can’t reject

F(3, 35) = 0.447

Can’t reject

F(3, 35) = 2.53

Can’t reject

F(3, 223) = 1.193

Can’t reject

Table 9. Summary of F-tests on OLS Estimates Null Hypothesis: The joint effect of property tax RATES, sales tax RATES, and income tax RATES on income levels is zero.

Model 1959 1969 1979 1989 1999 Pooled

Statistic

Result, .05 significant level

F(3, 35) = 0.392

Can’t reject

F(3, 35) = 0.347

Can’t reject

F(3, 32) = 1.146

Can’t reject

F(3, 34) = 1.665

Can’t reject

F(3, 35) = 1.876

Can’t reject

F(3, 219) = 3.633

Reject

At .05 significant level, Fc(3, 8) = 2.60 Fc(3, 30) = 2.92 Fc(3, 40) = 2.84


Page 14

Table 10. Dependent Variable: Average Annual Growth Rate of Real Per Capita Personal Income over Four Periods: 1959-69, 1969-79, 1979-89, and 1989-99 Explanatory Variable

SUR

3SLS

3SLS

SUR

3SLS

log yt-10

-3.037*** (-9.297)

-2.628*** (-6.820)

-2.025*** (-4.844)

-3.213*** (-9.060)

-3.042*** (-7.204)

Total Tax t-10

0.0282 (1.006)

0.0293 (0.659)

-0.0168 (-0.339)

Property Ratet-10

0.2019*** (2.872)

0.1913*** (2.685)

Sales Ratet-10

0.0169 (0.7260)

0.0189 (0.8057)

Income Ratet-10

-0.0275*** (-2.799)

-0.0267*** (-2.723)

-0.2913*** (-3.880)

?Total Tax

Metro Popt-10

0.0060** (2.482)

0.0045* (1.757)

0.0046* (1.731)

0.0042* (1.788)

0.0036 (1.446)

Colleget-10

0.0644*** (4.541)

0.0529*** (3.218)

0.0240 (1.221)

0.0664*** (4.444)

0.0636*** (4.092)

?Terms of Tradet

0.0339*** (4.669)

0.0378*** (4.746)

0.0530*** (6.333)

0.0423*** (5.043)

0.0422*** (5.022)

R2

.29

.30

.38

.35

.35

Observations

192

188

188

188

188

Notes: t statistics are reported in parentheses. * statistically significant at the .10 level for two-tailed test. ** statistically significant at the .05 level for two-tailed test. *** statistically significant at the .01 level for two tailed test. Each regression also includes a decade-specific constant. R2 computed at weighted average of R2 from decade-specific regressions, with weights equal to decade-specific share of total variance in growth rates.


Page 15

Table 11. Dependent Variable: Average Annual Growth Rate of Real Per Capita Personal Income Split Sample: 1959—79 and 1979—99 1959-79 -2.75*** (5.4)

1979-99 -2.72*** (3.8)

Property Ratet-10

.149* (1.7)

.258** (2.5)

Sales Ratet-10

-.008 (.2)

.044 (1.3)

Income Ratet-10

-.019 (1.6)

-.015 (1.0)

log yt-10 Total Tax t-10

1959-79 -2.20*** (5.5)

1979-99 -2.47*** (3.4)

1959-79 -1.72*** (4.2)

1979-99 -2.18** (2.5)

-.0055 (.1)

.116 (1.5)

-.064 (1.2)

.062 (.6)

-.160* (1.8)

-.520*** (2.8)

?Total Tax

Metro Popt-10

.0020 (.7)

.0109** (2.5)

.0006 (.2)

.0089* (1.7)

.0026 (1.0)

.0079** (2.0)

Colleget-10

.0056 (.2)

.0301 (1.1)

.00001 (.0)

.001 (.3)

.019 (.7)

.054** (2.2)

?Terms of Tradet

.035*** (4.1)

.042 (.5)

.045*** (5.5)

.103 (1.1)

.037*** (4.4)

.001 (.0)

R2

.55

.19

.61

.18

.57

.27

Observations

96

92

96

92

96

92

Notes: All Estimates by 3sls. t statistics are reported in parentheses. * statistically significant at the .10 level for two-tailed test. ** statistically significant at the .05 level for two-tailed test. *** statistically significant at the .01 le vel for two tailed test. Each regression also includes a decade-specific constant. R2 computed at weighted average of R2 from decade-specific regressions, with weights equal to decade-specific share of total variance in growth rates.


Variables

1959

Appendix A Actual Years of Data 1969 1979

Page 16

1989

1999

y

X

X

X

X

X

PC Earning

X

X

X

X

X

Total Tax

X

X

X

X

97

Property Tax

X

X

X

X

97

Sales Tax

X

State Level

X

X

97

Income Tax

X

State Level

X

X

97

Property Rate

58

71

81

87

98

Sales Rate

X

70

80

X

98

Income Rate

60

70

80

X

98

Metro Pop

60

70

80

90

99

Less Than 18

58

X

X

X

X

65 And Over

58

X

X

X

X

Not A High School

60

70

80

90

X

College And Higher

60

70

80

90

X

Sharefarming

X

X

X

X

X

Sharemining

X

X

X

X

X

Sharemanufacturing

X

X

X

X

X

Notes: “X” means that data is of the desired year as well as on the state and local level. Actual years are given if data is not of the desired years. “State Level” means that data is on the state level.

Taxes, Growth and the Convergence of Incomes among U.S

Recommend Documents