Net Interstate Population Growth Rates and The

0 downloads 0 Views 102KB Size Report
0The consumerMvoter may be viewed as picking that community which best satM ... maximum net discounted present value that can be expected by moving out ...
Net Interstate Population Growth Rates and The Tiebout-Tullock Hypothesis: New Empirical Evidence, 1990-2000 RICHARD J. CEBULA∗

Abstract This study empirically investigates the Tiebout-Tullock hypothesis as it might have applied to the pattern of net interstate population growth rates over the period 19902000. For the study period, it appears that the net state population growth rate has been an increasing function of the ratio of the total state plus local government outlays on public education in a state to that state’s total state plus local government tax burden. Additional variables in the study, including the previous-period median single-family housing-price inflation rate, a measure of previous-period growth in real personal income per capita and certain quality-of-life variables, also prove to be significant determinants of the net population growth rate in a state. In this context, it appears that, for the study period, the Tiebout-Tullock hypothesis played a significant role in determining state net population growth rates. (JEL H20,H30, H31); Atlantic Econ. J., 30(4): pp. 413-20, c Dec. 02. ° All Rights Reserved

Introduction There is a rich literature dealing with the determinants of geographic mobility and regional population growth rate differentials in the United States (see Greenwood [1975] and Cebula [1979] for surveys of earlier-period studies; some of the more recent or more frequently cited contributions relevant to the present study would include Cebula [1974], Cebula and Belton [1994], Conway and Houtenville [1998], Gale and Heath [2000], Glantz [1973], Lybbert and Thimany [2000], Pack [1973], Renas [1980, 1983], Saltz [1998], and Vedder [1976]). There are a number of these studies that have yielded potentially profound public policy implications, namely, those studies that investigate the so-called Tiebout-Tullock hypothesis. In general, the empirical Þndings regarding the Tiebout-Tullock hypothesis for the pre-1990 period are that: poor migrants have tended to be attracted to areas offering higher levels of public assistance (welfare); most migrant groups have been attracted to areas with higher per capita spending on public education; and most migrant groups, especially the elderly and higher income migrants, prefer areas with lower tax (especially property and income tax) burdens. The Tiebout [1956, p. 418] hypothesis argues that: “The consumer-voter may be viewed as picking that community which best satisÞes his preference pattern for public goods.” Tullock [1971] states the hypothesis in a way that stresses the choice consumer-voters make as assessing the bundle of local public goods and services and tax liabilities when ∗

Armstrong Atlantic State University–U.S.A.

414

CEBULA: POPULATION GROWTH RATES

415

voting with their feet. In any case, to the extent that the excess of the net value of these local goods and services over the associated tax liabilities is not capitalized into property values in a given area, in-migration should occur and out-migration should be diminished in order to reap the beneÞts of a perceived Þscal surplus. Whereas the Tiebout-Tullock hypothesis was empirically investigated at length for the periods of the 1950s, 1960s, 1970s, and 1980s, inquiries into the hypothesis have been lacking for the period of the 1990s. Accordingly, for the 1990-2000 period, this exploratory study seeks to examine the Tiebout-Tullock hypothesis empirically, thereby integrating the 2000 census information and providing updated and current analysis. The model includes Þscal variables and also reßects purely economic and quality of life considerations. This study of the Tiebout-Tullock hypothesis differs from other related studies in one or more ways. First, this study deals with net geographic population growth rate determinants (which include migration) for the period 1990-2000, a period that to date has not received attention in terms of the Tiebout-Tullock hypothesis. Second, unlike most related studies, geographically comparable living cost levels are included, although not as a separate variable, but rather in order to create a geographically comparable expected real future income variable. The focus on state-level data parallels numerous earlier studies (Cebula [1974]; Cebula and Belton [1994]; Gale and Heath [2000]; Gallaway and Cebula [1973]; Renas [1978]; and Saltz [1998]); however, in part, it also reßects the need to include considerations of geographic living-cost differentials on the one hand (to avoid omitted-variable bias related to money illusion) and the fortuitous availability of a series of high quality geographically comparable state-level living cost data from McMahon [1991]. Third, the previous-period median single-family housing price inßation rate in each geographic unit (state) is included in the model as a separate variable. This speciÞcation appears to be unique to the present study. As explained below, this variable is intended to account for the potential out-migration impact of higher capital gains beneÞts that might be experienced in higher median,singlefamily housing-price-inßation rate states. Fourth, as explained below, the Tiebout-Tullock Þscal variable for each of the 50 states is expressed as the ratio of total state plus local government outlays on public education in any state to that state’s total level of state plus local government taxes. This speciÞcation too is effectively unique within this literature. Hopefully, these differentiating traits will enable the present study to provide useful updated insights into the Tiebout-Tullock hypothesis. Focusing on total net population growth rates rather than migration reßects the argument that there may be greater (or at least different) public-policy implications (including statewide public policy implications such as state tax, education, and public assistance policies: Cebula [1974]; Gade and Adkins [1990]; and Nelson [2000]) related to net total population growth rates as opposed simply to observed migration per se; moreover, net total population growth rates arguably reßect not only net migration decisions but also (among other things) the decision to not migrate. A Simple Model In this study, we follow the models in Riew [1973], Cebula [1979], and elsewhere in treating the consumer-voter as regarding the location or migration decision as an investment decision. Accordingly, the decision to migrate from area i to area j requires that the net discounted present value of the move from area i to area j, DP V ij, be both (a) positive and (b) the maximum net discounted present value that can be expected by moving out of area i into any other known alternative. As suggested by the models in Riew [1973] and Cebula [1979], the net discounted present value of migration from area i to area j, DP V ij, consists of three major component parts:

416

AEJ: DECEMBER 2002, VOL. 30, NO. 4

1) Expected real income or income growth (I) in areas i and j and housing-price-inßation (HINF ) and related capital gains from housing sales considerations in areas i and j; 2) Expected beneÞts from publicly-provided goods and services such as public education (P E) and expected costs from tax liabilities (T ) in the areas; and 3) Expected quality-of-life (QOL) characteristics in the areas. Thus, it follows (based on Riew [1973] and Cebula [1979]) that migration will ßow from area i to area j if: DP V ij > 0; DP V ij = MAXf or j, j = 1, . . . , n

,

(1)

where n represents all of the feasible and known alternative locations to area i. Clearly, if DP V ij < 0, the resident of area i will remain in area i; indeed, a population ßow from area j to area i may well occur. The net population growth rate in an area (P OP GR) consists of net in-migration from other states, net in-migration from other nations, and net natural population growth among existing state residents. The decision to migrate from state i to state j implies that for some persons, DPV ij > 0 and that their DPV is maximized in state j. The decision for some immigrants to the U.S. to move from nation i to state j implies that their DP V ij > 0 and that it is maximized in state i. Finally, the decision for residents to remain in state j implies that DP V ji is not positive. Given the above framework, it follows that for state j: P OP GRj = f (Ij, HIN F j, P Ej, T j, QOLj)

.

(2)

Expressed in linear terms, the model in (2) initially becomes: P OP GRj = a + bIj + cHIN F j + dP Ej + eT j + f QOLj

.

(3)

Based on standard migration modeling and the Tiebout-Tullock hypothesis, it is expected that: b > 0, d > 0, e < 0, f > 0

.

(4)

It is further hypothesized in the present study that c < 0. This is because in areas having greater previous-period median single-family housing-price inßation rates, there are greater potential capital gains (which may also be exempt from income taxation) that can be experienced from selling one’s own home. Thus, those persons having the most inßated housing prices have a greater incentive, ceteris paribus, to sell their homes and move elsewhere. Two Simple Regression Models Based on the general framework summarized in (3) and (4) above, the following reducedform regressions are initially estimated: P OP GRj

= a0 + a1Ij + a2SU N SHIN Ej + a3JAN T EM P j + a4W EST j + a5HP Ij +a6P ET j + u0 , (5)

P OP GRj = b0 + b1Ij + b2SU NSHIN Ej + b3W EST j + b4HP Ij + b5P ET j + u00

, (6)

417

CEBULA: POPULATION GROWTH RATES

where P OP GRj is the net population growth in state j from April, 1990 to April, 2000, expressed as a percentage of state j’s April, 1990 total population; a0 and b0 are constants; Ij is the percentage growth rate of real per capita personal income in state j, 1981-90, as a proxy for expected future real income opportunities in state j; SUN SHINEj is the average annual percentage of possible sunshine in state j (that is, the average annual percentage of daylight in state j that sunshine is experienced); JAN T EM P j is the average January temperature in state j (in degrees Fahrenheit); W EST j is a dummy variable indicating whether state j is a western state, with western states being hypothesized as offering a higher overall quality of life [Gallaway and Cebula, 1979]; HP Ij is the average annual percentage inßation rate of the median price for single-family homes in state j, 1980-90, as an alternative measure of expected future living costs in state j; P ET j is the ratio of total state plus local government outlays on public education in state j to total state plus local government taxes in state j, 1990, expressed as a percentage; and u0 and u00 are stochastic error terms. The data source for variable P OP GRj is the Statistical Abstract of the United States, 2001 (Table 20). An alternative speciÞcation (to P OP GRj) of relative population growth (using the same raw data) is found in the conclusion to this study. As shown in the conclusion, the basic results are unchanged when using the alternative speciÞcation. For variables SU NSHINEj, JAN T EMP j, and W EST j, the sources are the Statistical Abstract of the United States, 1994, (Tables 379, 387) and Gallaway and Cebula [1973]. The data for variable HP Ij are obtained from Chao and Cebula [1996, Table 1], who in turn computed HP Ij for 1980-90 using data from the Statistical Abstract of the United States, 1991 and the Statistical Abstract of the United States, 1981. As shown below, the cost-of-living index (COLj) for deßating nominal income into real expected income is obtained from McMahon [1991, Table 3], although we Þnd that use of ACCRA living-cost indices in lieu of COLj leaves the results effectively unchanged. In any case, McMahon [1991] formulates reduced-form estimations for computing a geographically comparable state-level living cost index for each of the years from 1981 through 1990. The variable Ij is computed, as follows: Ij =

[(P IP Cj, 1990/COLj, 1990) − (P IP Cj, 1981/COLj, 1981)] (P IP Cj, 1981/COLj, 1981)

,

where P IP Cj, 1990; P IP Cj, 1981 is the nominal personal income per capita in state j in 1990 and 1981, respectively (Statistical Abstract of the United States,2001 (Table 727); Statistical Abstract of the United States,1985 (Table 731)) and COLj, 1990; COLj, 1981 is the cost of living for the average four-person family unit in state j in 1990, and in 1981, respectively, expressed as an index (average = 100.00). [McMahon, 1991, Table 3] The Ij ratio, expressed as a percentage, is taken as the measure of expected future growth in real personal income per capita in state j. In principle, this speciÞcation parallels that in Cebula and Belton [1994] and in Gale and Heath [2000] for interstate migration during the 1980s.The three variables SU N SHIN Ej, JANT EMP j, and W EST j are intended to reßect elements of the quality of life that have previously been found to affect migration patterns (Gallaway and Cebula[1973]; Cebula [1979; 1990]; Gale and Heath [2000]; and Renas [1978; 1980]). The P E and T variables from the previous section have been combined here into one variable, P ET j. This speciÞcation reßects the fact that P E and T are likely to be statistically related, especially since state budgets are generally required to be balanced. To address this issue, the variables P E and T are expressed as a ratio (that in turn is expressed as a percent).

418

AEJ: DECEMBER 2002, VOL. 30, NO. 4

Estimating (5) and (6) by OLS, using the White [1980] procedure to correct for heteroskedasticity, yields (7) and (8), respectively: P OP GRj

= −26.24 + 3.75Ij + 0.398SUN SHIN Ej − 0.08JAN T EMP j (+5.23) (+3.04) (−0.90) +8.94W EST j − 0.059HP Ij + 3.47P ET j (+3.44) (−3.10) (+2.61) RSQ = 0.71, adjRSQ = 0.66, F (6.43) = 17.20 ,

= −25.53 + 3.67Ij + 0.355SU N SHIN Ej + 8.51W EST j (+5.48) (+3.35) (+3.41) −0.062HP Ij + 3.18P ET j (−3.04) (+2.83) RSQ = 0.70, adjRSQ = 0.67, F (5, 44) = 20.5 ,

(7)

P OP GRj

(8)

where terms in parentheses beneath coefficients are t-values. Ten of the 11 estimated coefficients in (7) and (8) have the hypothesized signs and are statistically signiÞcant at the Þve percent level or beyond. In (7) and (8), the estimated coefficients on the expected real income variable are both positive and signiÞcant at the one percent level. Thus, it appears that the previous-period percentage growth rate of real personal income per capita as deÞned above, as a proxy for expected future real income growth and or employment opportunities, acts to signiÞcantly and positively act on net population growth. Next, in (7) and (8), the estimated coefficients on the SU N SHIN Ej and W EST j variables are all positive and statistically signiÞcant at the one percent level. Thus, the net population growth rate tends to be higher in states having a higher percentage of sunshine and in western states, where it has been argued that the quality of life is perceived to be higher than elsewhere (Gallaway and Cebula [1973]; Cebula [1979]; Saltz [1998]; and Vedder [1976]). On the other hand, whereas the JAN T EMP j variable fails to be statistically signiÞcant in (7), this insigniÞcance is not attributable to multicollinearity with the other variables in the system. Moreover, using annual average heating degree days rather than JANT EMP j, as in Renas [1978; 1980] and Gale and Heath [2000], does not provide signiÞcant coefficients. In (7) and (8), the estimated coefficients on the previousperiod median single-family housing-price inßation rate variable are negative and signiÞcant at the one percent level. Thus, the net population growth rate tends to be lower in those states where the previous-period median single-family housing-price inßation rate has been higher. Presumably, this may be because in those states having higher such housing-price inßation rates, the prospect of larger capital gains may induce increased rates of housing sales and hence to some extent increased out-migration, thereby reducing net population growth in those states. Finally, the focus is on the public-policy variable, P ET j. In (7) and (8), the estimated coefficients on variable P ET j are both positive and statistically signiÞcant at approximately the one percent level. Thus, the evidence indicates that states with higher ratios of state plus local government public education outlays to state plus local government tax burdens (as the proxy for perceived Þscal surplus) experience greater net population growth rates. These results provide strong empirical support for the Tiebout-Tullock hypothesis for the 1990-2000 period.

CEBULA: POPULATION GROWTH RATES

419

Two Alternative Estimates It has been suggested by Cebula [1990] for elderly migrants and by Saltz [1998] for younger migrants age 20-40 years that the very existence of a state income tax, of and in itself, may act as a deterrent to in-migration. Accordingly, following both of these studies, regression (5) and (6) are both amended to include an additional variable, T Dj, which is a dummy variable indicating whether state j has a state income tax system. T Dj = 1 for those states having a state income tax system, and T Dj = 0 otherwise. These OLS estimates, using the White [1980] correction for heteroskedasticity, are provided in (9) and (10) below: P OP GRj

= −23.08 + 3.59Ij + 0.415SU NSHIN Ej − 0.08JAN T EMP j + 7.48W EST j (+5.57) (−3.14) (−1.01) (+3.15) −0.05HP Ij + 3.70P ET j − 5.74T Dj (−2.46) (+2.38) (−1.71) RSQ = 0.73, adjRSQ = 0.69, F (7.42) = 16.43 , (9)

P OP GRj

= −22.37 + 3.52Ij + 0.37SUN SHINEj + 7.04W EST j − 0.05HP Ij (+5.68) (+3.55) (+2.95) (−2.67) +3.40P ET j − 5.69T Dj (+2.58) (−1.65) RSQ = 0.73, adjRSQ = 0.69, F (6, 43) = 18.98 .

(10)

In estimations (9) and (10), the estimated coefficients on the variables Ij, SU N SHIN Ej, W EST j, and HP Ij are signiÞcant at the 2.5 percent level or beyond with the expected signs; furthermore, the estimated coefficient in (9) on variable JAN T EMP j once again fails to be signiÞcant at an acceptable (Þve percent) level. Clearly, these results are entirely consistent with the corresponding results in (7) and (8) above. In (9) and (10), whereas the income tax dummy has the expected negative sign, it is barely statistically signiÞcant at only the ten percent level, so that the evidence regarding the signiÞcance of the existence of a state income tax system is relatively weak. It should be noted further that in two separate estimates, as an alternative to T Dj, the average state income tax rate in each state (T AV Ej) was used, with those states having no state income tax at all assigned a value of zero (T AV Ej = 0). In both of these estimates, the overall results were entirely consistent with those in (9) and (10). Finally, the estimated coefficients on the P ET j variable in (9) and (10) both are positive and signiÞcant at roughly the two percent level, once again providing empirical support for the Tiebout-Tullock hypothesis. Hence, the percentage net state population growth rate appears to be an increasing function of the previous-period real personal income per capita growth rate and certain qualityof-life variables (SU N SHIN Ej and W EST j), while being a decreasing function of the state previous-period median single-family housing price inßation rate. Furthermore, the percentage net state population growth rate does appear to follow a Tiebout-Tullock pattern insofar as it is an increasing function of the ratio of state plus local government outlays on public education to state plus local government tax burdens. On the other hand, the evidence appears to indicate that the existence of a state income tax system per se may not signiÞcantly affect the percentage state population growth rate pattern during the 1990-2000 period. It is interesting to observe that the latter Þnding is not the consequence of multicollinearity between the variables P ET j and T Dj; indeed, the correlation coefficient between these two variables is only +0.04.

420

AEJ: DECEMBER 2002, VOL. 30, NO. 4

Conclusions This empirical study has examined the impact on percentage net state population growth rates over the 1990-2000 period (P OP GR) of a variable reßecting the Tiebout-Tullock hypothesis. This study controls for the impacts of interstate previous-period real personal income per capita growth rate differentials, previous period median single-family housing price inßation-rate differentials, and quality-of-life factors. The results strongly suggest that the Tiebout-Tullock framework is alive and well. Indeed, the evidence consistently indicates that the percentage net state population growth rate has been an increasing function of the ratio of state plus local government outlays on public education to the state plus local government tax burden (as the adopted proxy for the perceived Þscal surplus) over the 1990-2000 period. In closing, it should be noted that, as an alternative to P OP GRj as speciÞed above, the left-hand-side variable measuring the net population growth rate might be restated as:

Log Population Change, 1990-2000 = Log Population, 2000 − Log Population, 1990 Adopting this approach, we obtain the following OLS results after adopting the White [1980] correction:

Log Population Change,1990-2000 = −1.63 + 3.516Ij + 0.0022SU NSHIN Ej (+3.74) (+2.07) +0.042W EST j − 0.004HP Ij + 5.74P ET j (+2.08) (−1.68) (+3.07) RSQ = 0.45, adjRSQ = 0.38, F (5, 44) = 7.07 (11) Although these results lack some of the robustness found in estimations (7)—(10) above, the results are nevertheless compatible with those estimations. The HP Ij variable does have the expected sign, although it is signiÞcant at only the ten percent level. On the other hand, the SU NSHIN Ej, W EST j, and Ij variables all remain signiÞcant at acceptable levels with the expected signs. Perhaps most relevant, in view of the focus of this study, the estimated coefficient on the P ET j variable is once again positive and signiÞcant (at beyond the one percent level), providing yet further support for the Tiebout-Tullock hypothesis. References Cebula, R. J. “A Brief Empirical Note on the Tiebout Hypothesis and State Income Tax Policies,” Public Choice, 67, 1990, pp. 101-5. –. The Determinants of Human Migration. Lexington, MA: Lexington Books, 1979. –. “Interstate Migration and the Tiebout Hypothesis: A Disaggregated Analysis,” Journal of the American Statistical Association, 69, 1974, pp. 876-9. Cebula, R. J.; Belton, W. J. “Voting with One’s Feet: An Analysis of Public Welfare and Migration of the American Indian,” American Journal of Economics and Sociology, 53, 1994, pp. 273-80. Chao, R. Y.; Cebula R. J. “Determinants of Geographic Differentials in the Savings and Loan Failure Rate: A Heteroskedastic TOBIT Estimation,” Journal of Financial Services Research, 10, 1996, pp. 5-25. Conway, K. S.; Houtenville, A. J. “Do the Elderly Vote with Their Feet?,” Public Choice, 97, 1998, pp. 663-85.

CEBULA: POPULATION GROWTH RATES

421

Gade, M. N.; Adkins, L. C. “Tax Exporting and State Revenue Structures,” National Tax Journal, 43, 1990, pp. 39-52. Gale, L. R.; Heath, W. C. “Elderly Internal Migration in the United States Revisited,” Public Finance Review, 28, March, 2000, pp. 153-70. Gallaway, L. E.; Cebula, R. J. “Differentials and Indeterminacy in Wage Rate Analysis: An Empirical Note,” Industrial and Labor Relations Review, 26 , 1973, pp. 991-5. Glantz, F. B. “Migration and Economic Opportunity: The Case of the Poor,” New England Economic Review, 40, 1973, pp. 14-9. Greenwood, M. J. “Research on Internal Migration in the United States,” Journal of Economic Literature, 13, 1975, pp. 397-433. Lybbert, T. J.; Thimany D. D. “Migration Effects of Olympic Siting: A Pooled Time Series CrossSectional Analysis of Host Regions,” Annals of Regional Science, 34, 2000, pp. 405-20. McMahon, W. W. “Geographic Cost of Living Differentials: An Update,” American Real Estate and Urban Economics Association Journal, 19, 1991, pp. 426-50. Nelson, M. A. “Electoral Cycles and the Politics of State Tax Policy,” Public Finance Review, 28, 2000, pp. 540-60. Ostrosky, A. L. “A Further Look at Family Budgets and Mobility,” Social Science Quarterly, 67, 1986, pp. 212-3. Pack, J. R. “Determinants of Migration to Central Cities, 1955-1960,” Journal of Regional Science, 13, 1973, pp. 155-64. Renas, S. M. “The Cost of Living, Labor Market Opportunities, and the Migration Decision,” Annals of Regional Science, 12, 1978, pp. 95-104. –. “An Empirical Note on the Tiebout-Tullock Hypothesis: Comment,” Quarterly Journal of Economics, 94, 1980, pp. 619-23. –. “The Cost of Living, Labor Market Opportunities, and the Migration Decision: More on problems of MisspeciÞcation and Aggregation Bias,” Annals of Regional Science, 17, 1983, pp. 98-110. Riew, J. “Migration as Investment,” Journal of Regional Science, 12, 1973, pp. 65-73. Saltz, I. S. “State Income Taxation and Geographic Labour Force Mobility in the United States,” Applied Economics Letters, 5, October, 1998, pp. 599-601. Tiebout, C. M. “A Pure Theory of Local Expenditures,” Journal of Political Economy, 64, 1956, pp. 416-24. Tullock, G. “Public Decisions as Public Goods,” Journal of Political Economy, 79, 1971, pp. 913-18. Vedder, R. K. The American Economy in Historical Perspective. Belmont, CA: Wadsworth, 1976. White, H. “A Heteroskedasticity-Consistent Covariance Matrix and a Direct Test for Heteroskedasticity,” Econometrica, 48, May, 1980, pp. 817-38.