The robust relationship between taxes and U.S. state income growth.

A somewhat different picture emerges when the specification is broadened to allow lagged effects. Column 5 reports the results of adding lagged values of the tax burden variable to the specification of column 1. While the contemporaneous relationship between tax burden and income growth remains positive, lagged values of tax burden are estimated to be negatively associated with state income.

This suggests that previous studies may have failed to identify a negative relationship between taxes and income growth because they relied on specifications that used annual data and did not allow for lagged tax effects. My analysis suggests that tax policies take time to work its full effects on the economy. When the specification is sufficiently general to pick up these effects, a negative relationship between taxes and income growth emerges.

The use of annual data may also have contributed to previous findings of coefficient instability. Though they are positively correlated, tax burden has been demonstrated to substantially mismeasure state tax policy (Reed and Rogers, 2006, 2007). Annual data are more vulnerable to measurement error bias than five-year interval data. Consequently, it would not be surprising if estimates of tax effects based on annual data were prone to instability depending on the particular distribution of measurement errors in the sample. This may be an additional reason why previous studies have had difficulty identifying robust tax effects.

CONCLUSION

Using five-year data from 1970-1999 and the 48 continental states, I find that both (1) contemporaneous changes and (2) lagged levels of taxes are negatively and significantly related to income growth. The estimated effects vary depending on variable specification; estimation procedure; time period, region, and state; and the manner in which the data are organized into five-year intervals. Nevertheless, the finding of negative and statistically significant tax effects is generally robust across all of these dimensions, with one exception: State--specific estimates of tax effects widely vary. This latter result may be explained by the narrow parsing of the data. At this level of analysis there are only six observations per state-specific tax coefficient.

These results are surprising given that previous studies have had difficulty identifying a robust relationship between state taxes and incomes. My analysis suggests that this may be because previous studies of state income growth have tended to use annual data, have differed in their variable specifications, and have not allowed for lagged tax effects. When I use annual data and restrict the analysis to contemporaneous tax effects, I estimate positive tax effects, but the sizes and significances of the tax coefficient greatly vary depending on variable specification. When I include lagged values of the tax variable, a negative relationship between taxes and growth emerges. This lack of robustness is not apparent in the five-year interval data. This may be because the variables interact over time in complex ways that are difficult to model. It may also be that the data--for definitional and measurement reasons--are not well-suited to relating to each other at the annual level.

It needs to be emphasized that my claim for robustness should be understood as applying only within the context of U.S. state income growth. It should not be interpreted as being more widely applicable to other contexts, such as employment growth, manufacturing activity, plant locations, etc., or to the relationship between taxes and income growth outside the U.S.

Much work remains to be done before reliable estimates of tax effects can be obtained. (29) However, this study establishes that there is a durable empirical relationship between taxes and U.S. state income growth that warrants further investigation. Obtaining a better understanding of the nature and cause of that relationship is a potentially fruitful avenue for future research. It is hoped that this study will stimulate efforts towards that end. APPENDIX STATISTICAL SUMMARY OF DATA

Std. Variable Mean Deviation Minimum Maximum DLNY (1) 8.23 5.20 -9.38 40.45 DLNK (2) 7.42 7.81 -26.92 55.43 DLNL (3) 4.66 3.99 -7.22 14.97 DLNN (4) 4.63 4.48 -8.63 21.45 Tax Garden (5) D 0.13 0.88 -5.52 5.91

L 10.87 1.37 7.92 19.27 Education (6) D 1.77 0.55 0.34 3.21

L 16.41 4.92 6.66 30.21 Working population (6) D 0.97 0.93 -1.22 2.93

L 55.84 3.18 47.54 62.26 Nonwhite (6) D 0.56 0.51 -0.98 2.42

L 11.75 8.76 0.36 37.35 Female (6) D -0.02 0.15 -0.57 0.75

L 51.23 0.77 48.77 52.76 Population (6) L 14.93 1.00 12.72 17.27 Population density (6) D 4.93 6.68 -8.44 37.26

L 162.25 230.78 3.44 1089.83 Urban (6) D 0.75 1.13 -1.97 3.96

L 67.18 14.43 32.16 93.54 Agriculture (6) D -0.06 2.46 -16.72 18.85

L 3.28 3.98 -8.92 29.06 Manufacturing (6) D -0.81 1.68 -6.09 3.37

L 20.93 8.42 3.73 40.49 Service (6) D 1.47 1.25 -3.22 6.40

L 19.51 5.65 10.93 41.55 Mining (6) D -0.19 0.76 -3.29 4.27

L 2.15 3.53 0.02 24.98 Union (6) D -1.47 2.36 -10.60 5.00

L 18.48 8.12 3.30 41.70 Diversity (6) D -0.06 0.77 -5.42 4.66

L 17.36 2.05 13.84 23.56 LNY_1 (7) 2.53 0.20 1.96 3.06 NonTaxRevenues (8) D 0.33 0.93 -2.13 6.26

L 8.16 2.29 3.44 20.13 Welfare (9) D 0.19 0.52 -1.70 2.88

L 2.15 0.83 0.75 5.30 Variable Descriptions: (1) DLNY is the percent change in real Per Capita Personal Income (1984 dollars). (2) DLNK is the percent change in net private Capital Stock created through 1-digit SIC industries (measured in millions of chained 1996 dollars). These data were provided by Steve Yamarik (cf. Garofalo and Yamarik, 2002). (3) DLNL is the percent change in total employment (source: BEA). (4) DLNN is the percent change in total population (source: Census). (5) Tax burden is the ratio of total state and local tax revenues over total state personal income. (6) These variables are described in Table 2. "D" denotes the five-year difference in the variable over the period (t-4, t). "L" denotes the value of the variable at the beginning of the five-year period. (7) LNY_1 is the value of the log of real Per Capita Personal Income (1984 dollars) at the beginning of the five-year period. (8) "NonTaxRevenues" is defined as General Revenues (state + local) minus Total Taxes (state + local) divided by Personal Income at the start of the fiscal year (source: Census). (9) "Welfare" is defined as Direct General Expenditure of State and Local Governments on Public Welfare divided by Personal Income at the start of the fiscal year (source: Census).

Acknowledgments

I am very appreciative to Steve Yamarik for providing data on state capital stocks (cf. Garofalo and Yamarik, 2002). I acknowledge helpful comments from Steve Yamarik, Bob Tannenwald, seminar participants at the Motu Economic and Public Policy Research Institute, the National Tax Association meetings, two anonymous referees, and the editor. Culpability for remaining errors is mine alone.

REFERENCES

Aim, James, and Janet Rogers. "Do State Fiscal Policies Affect Economic Growth?" Department of Economics, Andrew Young School of Policy Studies, Georgia State University. Mimeo, 2005.

Aschauer, David Alan. "Do States Optimize? Public Capital and Economic Growth." The Annals of Regional Science 34 No. 3 (September, 2000): 343-63.

Bania, Neil, Jo Anna Gray, and Joe A. Stone. "Growth, Taxes, and Government Expenditures: Growth Hills for U.S. States." National Tax Journal 60 No. 3 (June, 2007): 193-204.

Bartik, Timothy J. Who Benefits from State and Local Economic Development Policies? Kalamazoo, MI: W.E. Upjohn Institute for Employment Research, 1991.

Bassanini, Andrea, Scarpetta, Stefano, and Philip Hemmings. "Economic Growth: The Role of Policies and Institutions. Panel Data Evidence from OECD Countries." Economics Department, Organization for Economic Co-operation and Development (OECD). OECD Working Paper No. 283. OECD Economics Department, 2001.

Beck, Nathaniel, and Jonathan N. Katz. "What to Do (and Not to Do) with Time-Series Cross-Section Data." American Political Science Review 89 No. 3 (September, 1995): 634-47.

Becsi, Zsolt. "Do State and Local Taxes Affect Relative State Growth?" Economic Review 81 No. 2 (March/April, 1996): 18-36.