The influence of taxes on employment and population growth: evidence from the Washington, D. C. metropolitan area.

(1) (2) (3) District of Columbia -0.066 -0.067 -0.225

(0.011) (0.055) (0.154)

[0.000] [0.220] [0.146] Charles County, MD -0.017 0.003 0.032

(0.011) (0.024) (0.039)

[0.141] [0.895] [0.408] Montgomery County, MD -0.035 -0.016 -0.130

(0.011) (0.021) (0.074)

[0.003] [0.443] [0.082] Prince George's County, MD -0.040 -0.025 -0.056

(0.011) (0.020) (0.068)

[0.000] [0.222] [0.411] Alexandria City, VA -0.050 -0.017 -0.083

(0.011) (0.010) (0.034)

[0.000] [0.097] [0.017] Arlington County, VA -0.045 -0.004 -0.082

(0.011) (0.012) (0.038)

[0.000] [0.711] [0.033] Fairfax County, VA -0.007 0.011 -0.063

(0.011) (0.009) (0.046)

[0.516] [0.199] [0.179] Loudoun County, VA -0.006 0.036 0.013

(0.011) (0.011) (0.039)

[0.583] [0.001] [0.743] Sales tax rate -- -2.029 -2.078

(1.104) (1.172)

[0.047] [0.079] Commercial property tax rate -- 3.410 1.584

(1.459) (1.662)

[0.021] [0.342] Corporate income tax rate -- 0.691 2.757

(1.278) (2.951)

[0.590] [0.352] Personal property tax rate -- -1.586 -2.441

(0.652) (0.672)

[0.016] [0.000] Log (unemployment -- -0.018 -0.021 insurance cost) (0.013) (0.013)

[0.160] [0.113] Log (per capita income) -- -- 0.141

(0.073)

[0.054] Log (total crime index) -- -- 0.0080

(0.025)

[0.750] Log (per capita non-AFDC -- -- 0.029 expenditures) (0.0073)

[0.000] Constant 0.074 0.154 -1.282

(0.008) (0.089 (0.748)

[0.000] [0.087] [0.089] Observations 225 171 161 R squared 0.23 0.706 0.742

Notes: All explanatory variables are lagged one year. Tax rates and the growth rate are in decimal. Prince William County, VA is the omitted county. Regression in columns (2) and (3) include year dummies. Standard errors are in parentheses and p values are in brackets.

The most complete model is presented in column (3) of Table 5. Higher rates of two business taxes, the personal property tax and the sales tax, are associated with much lower employment growth in the following year. Our estimate indicates that a 1 percentage point higher tax rate on personal property reduces annual employment growth by 2.44 percentage points (with a p value of 0.00). This is an economically large effect and is highly statistically significant. At the means of the sample, this implies an elasticity of employment growth with respect to the business personal property tax of -2.12.

The sales tax also has a similarly sized negative effect on employment growth and is statistically significant at our more generous significance levels (p value of 0.08). We estimate that a 1 percentage point increase in the sales tax rate reduces the annual growth rate in employment by 2.08 percentage points. This implies an elasticity of -2.17 of employment growth with respect to the sales tax at the means of the sample.

The remaining two taxes--the corporate income tax and the commercial property tax--have the wrong sign and are statistically insignificant. Both variables have measurement problems that may account for this result. Because we do control for school quality, if high property taxes are correlated with high spending on schools, this coefficient may be picking up an effect of good schools on business location decisions.(9) Its difficult to conclude much of anything since the p value is 0.34.

The corporate income tax coefficient is positive, which is not the sign we expected, but because there is little variation in this variable, has a wide confidence interval. The corporate tax rate varies only across states, and Maryland and Virginia did not change their rates over this time period. The only variation left to relate to influence employment is the variation over time in the District, and this does not appear to influence employment growth. Statistically, our estimate is not different from zero.

Higher unemployment insurance costs appear to have a negative effect on employment growth, and the estimated coefficient is close to statistical significance with a p value of 0.11. A ten percent increase in the average cost of unemployment insurance is estimated to reduce annual employment growth by 0.21 percentage points.

While public expenditures do not appear to play a role in residential choice, they are an important influence on the location of employment. We estimate that a ten percent increase in nonwelfare public spending increases annual employment growth by 0.29 percentage points. This effect is highly statistically significant (with a p value of 0.00) and implies an elasticity at the means of the sample of 0.66. Recall, these are increases in public services apart from AFDC payments--public works and safety, for example.

As in the population growth regressions, employment growth is higher in jurisdictions with higher levels of per capita income. A ten percent increase in per capita income is associated with an increase in private employment growth of 1.41 percentage points (with a p value of 0.05). Crime does not appear to influence employment growth in these regressions. The coefficient on the per capita crime rate (0.008) is essentially zero and has a p value of 0.75.

When we drop the jurisdiction fixed effects for the model estimated in column (3), the coefficients are estimated more precisely because we are not conditioning on a fixed effect. The sales and personal property tax coefficients become more negative (in the decimal places) and more statistically significant. The expenditure coefficient is smaller but still highly statistically significant. The commercial property tax results are qualitatively unchanged.

There are two notable differences. When estimated without fixed effects, the crime rate and the corporate income tax rate coefficients become negative and, in the case of the crime rate, highly statistically significant (the coefficient on the lagged crime rate is -0.011 with p value 0.00, and the coefficient on the corporate income tax is -0.803 with p value 0.13). We know the corporate income tax rate changes over this time period only for the District, so the differences between the jurisdictions are permanent and therefore absorbed in the fixed effect. A similar argument may be made for the crime rate. Thus, while fixed effects do eliminate the statistical significance of some variables, the primary findings remain strong.

To summarize, after controlling for aggregate time and jurisdiction effects, two taxes on business--the personal property tax and the sales tax--have economically large negative effects on the annual growth rate of private employment. The coefficients imply elasticities for these taxes over two in absolute value. Higher unemployment insurance costs exert a marginally statistically significant negative influence. Higher levels of public services aside from welfare payments are associated with greater employment growth, with an estimated elasticity at the mean values of 0.66. Employment growth is greater in jurisdictions with higher per capita income, but appears to be unaffected by crime levels.

Limitations of Our Econometric Analysis

Our conclusions regarding influences of policy variables on economic development in the District and surrounding area must be qualified by the limitations of our data. We have used publicly available tax information, but we have not controlled for the various incentives that jurisdictions offer to business, either as a matter of course or through individual negotiations. Further, in addition to crime, primary concerns in the District area include quality of education services and quality of infrastructure. We attempted to examine these influences, but limitations in school data reduced the sample size in our regressions to the point that we considered the results to be unrepresentative of the area and we could find no reliable measures of public infrastructure quality to use.

CONCLUSIONS