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

(-8.37) (-9.72) (-9.96) Number of observations 288 288 288 [R.sup.2] 0.938 0.933 0.935 SIC 624.70 572.52 573.91 AICc 675.36 647.78 645.00 (1) Summary statistics for each of the variables is reported in the Appendix. Note: The regression equation follows the general specification of equation [6] in the text. "D" and "L" stand for differenced and level forms of the variables. In addition to the variables listed above, the model includes the variables DLNK, DLNL, DLNN, and state and time fixed effects. t-statistics are listed in parenthesis below each estimated coefficient and are calculated using White heteroscedasticity-robust standard errors. TABLE 4 ROBUSTNESS CHECK USING ALTERNATIVE ESTIMATION PROCEDURES Procedure Tax Burden (D) Tax Burden (L) OLS -0.5470 -0.6905

with robust VCE for (-3.59) (-3.20)

heteroscedasticity

with panel-corrected standard (-2.88) (-3.35)

errors

with robust VCE for (-5.43) (-6.57)

heteroscedasticity and

cross-sectional correlation FGLS (weighted on groupwise -0.5086 -0.6494

heteroscedasticity)

with robust VCE for (-3.83) (-3.89)

heteroscedasticity

with panel-corrected standard (-3.52) (-4.43)

errors

with robust VCE for (-3.54) (-8.00)

heteroscedasticity and

cross-sectional correlation DPD (difference) -0.6834 -0.8547

Arellano-Bond one-step procedure (-3.65) (-3.67) DPD (difference) -0.5341 -0.7684

Arellano-Bond two-step procedure (-1.74) (-2.16) Note: Coefficient estimates are boldface; t-statistics are reported in parentheses. Each estimation procedure estimates the same variable specification as column 2 of Table 3. The first set of OLS results repeats those results for comparisons sake. The respective estimation procedures are described in greater detail in the fourth section. TABLE 5 ROBUSTNESS CHECK USING ALTERNATIVE CUTS OF THE DATA

(1) (2) (3)

5-Year Data 5-Year Data 5-Year Data

1970-1974, 1970-1975, 1971-1975,

..., ..., ..., Variable Name D/L 1995-1999 1995-2000 1996-2000 Tax burden D -0.5086 -0.5616 -0.4545

(-3.54) (-3.91) (-3.47)

L -0.6494 -0.4179 -0.3096

(-8.00) (-3.00) (-2.01) Education D 1.6935 1.4690 1.7034

(4.81) (5.33) (6.12)

L 1.0819 1.3104 1.0290

(5.69) (5.94) (5.03) Working population D 1.7580 0.8836 0.5753

(4.90) (4.06) (1.97)

L 1.2660 1.3152 1.0980

(13.50) (13.92) (10.50) Nonwhite D 1.3004 0.9542 0.5168

(4.84) (4.08) (3.06) Female D -3.2024 -1.1457 -0.8878

(-4.38) (-1.51) (-2.12) Population L 4.1887 5.5797 4.6138

(6.63) (3.75) (2.88) Agriculture D 0.5061 0.5657 0.3047

(5.66) (5.44) (2.24)

L 0.3592 0.4739 0.5536

(5.79) (6.47) (6.00) Mining D -0.4663 0.0357 0.0780

(-2.48) (0.23) (0.57) Union D 0.1045 0.0776 0.0843

(3.07) (1.73) (4.96) Diversity L -0.2312 -0.2516 0.0764

(-1.86) (-1.81) (0.94) LNY_1 -39.042 113.288 -34.17

(-12.67) (-12.85) (-11.61) Number of observations 288 288 288 Note: Each of the three sets of regression results employs FGLS (weighting on groupwise heteroscedasticity) with robust VCE for heteroscedasticity and cross-sectional correlation. t-statistics are reported in parentheses below the respective coefficient estimates. The first column reproduces previous results for comparison's sake (cf. third row from bottom in Table 4). The next two columns show the effects of using different cuts of the data. TABLE 6 ROBUSTNESS CHECK ACROSS TIME PERIODS

Time-Specific Coefficients

Tax Burden (D) Tax Burden (L) Time Period Coefficient t-statistic Coefficient t-statistic 1970-1974 -1.1551 -9.83 -0.3062 -5.18 1975-1979 -0.7518 -4.58 -0.6710 -7.32 1980-1984 -0.0615 -0.24 -0.6455 -7.18 1985-1989 -0.1642 -0.71 -0.9044 -6.61 1990-1994 -0.6450 -2.97 -1.0086 -8.20 1995-1999 -1.1014 -5.17 -0.5286 -4.98 Note: Regression results are derived from a regression equation having the variable specification of column 2 of Table 3, supplemented with the respective time-interaction dummy variables (cf. the text for further details). The estimation procedure is FGLS (weighting on groupwise heteroscedasticity) with robust VCE for heteroscedasticity and cross-sectional correlation. TABLE 7 ROBUSTNESS CHECK ACROSS BEA REGIONS AND STATES

Tax Burden (D)

Coefficient t-statistic Region-Specific Coefficients:

Great Lakes -0.2226 -0.36

Mid-Atlantic -0.6228 -1.97

New England -0.5956 -1.58

Plains -0.7237 -2.15

Rocky Mountain -0.3329 -2.22

South -0.5232 -2.29

Southwest -1.2132 -1.68

West -0.7435 -2.49 State-Specific Coefficients:

Total number of coefficients/Number 48/35

negative

(Percent negative) (72.9)

Total number of significant 15/13

coefficients/Number negative

(Percent negative) (86.7)

Tax Burden (L)

Coefficient t-statistic Region-Specific Coefficients:

Great Lakes -0.5268 -1.79

Mid-Atlantic -0.4599 -1.90

New England -0.7789 -2.13

Plains -0.2635 -2.16

Rocky Mountain -1.1408 -3.39

South -0.4521 -1.28

Southwest 0.4150 0.87

West -0.0570 -0.10 State-Specific Coefficients:

Total number of coefficients/Number 48/31

negative

(Percent negative) (64.5)

Total number of significant 13/9

coefficients/Number negative

(Percent negative) (69.2) Note: Regression results are derived from a regression equation having the variable specification of column 2 of Table 3, supplemented with the respective interaction dummy variables (cf. the text for further details). The estimation procedure is FGLS (weighting on groupwise heteroscedasticity) with robust VCE for heteroscedasticity and cross-sectional correlation. For the last row, statistical significance is defined at the 10 percent level. TABLE 8 ROBUSTNESS ACROSS ALTERNATIVE SPECIFICATIONS OF GOVERNMENT FINANCES

(3)

Taxes +

(2) Non-Tax (4)

Taxes + Revenues + Taxes +

(1) Non-Tax Welfare Welfare

Only Taxes Revenues Expenditures Expenditures Tax Burden (D) -0.5086 -0.4020 -0.4654 -0.5622

(-3.54) (-2.82) (-2.58) (-3.45) Tax Burden (L) -0.6494 -0.4866 -0.5913 -0.7277