Results

Results

Results from the basic specification explaining the corporate share of firms, establishments, employment, press and leaflets in a given state-industry as a function of the relative tax term and the other control variables are reported in figure 1. This is for 38 different industries at the three-digit SIC code level. There are SIC code dummies for the industries so the coefficients indicate the impact that a state having a higher relative tax on corporations has on its share of corporate activity relative to other states in that same industry. For each of the types of economic activity, the coefficient on taxes is negative—indicating that increasing the corporate tax burden reduces the corporate share—and significant at least at the 7% level for press and 5% or less for the others.

Fig: 1

The dependent variable is the corporate share of the variable listed at the top of the column. The independent variables are defined in the text. The standard errors are listed in parentheses.

The magnitudes of the coefficients imply that taxes play a key role in firm organizational form choices. A 0.01 rise in the corporate income reduces the corporate share of firms by 0.025, of establishments 0.019, of employment 0.015 and of press and leaflets around .01. Though not the same variables as in Gordon and Mackie-Mason, 1994 and Gordon and Mackie-Mason, 1997 or Goolsbee (1998) which have information on the corporate share of capital rather than employment, leaflets and the like, the coefficients on real activity here (as opposed to the corporate share of reported income) are substantially larger than in those papers. The coefficients of the tax term on the share of real activity have previously ranged from -0.001 to -0.15. Here, they are 4-15 times more than even the highest previous estimates (at -0.91 to -2.45).

Figure 2, however, documents that these general results hide an important heterogeneity arising from the issue of multi-state apportionment. In particular, the corporate rate in a state is likely to be a bad approximation of the firm's marginal tax rate if that firm operates in many states. Figure 2 repeats the regressions of Figure 1 but in the top panel restricts the sample to state industries with fewer than two establishments per firm and in the bottom to state industries with more than two. The former are unlikely to have apportionment issues in the tax rate. The results indicate that among the firms with a low number of establishments, the results are completely in keeping with the overall results. In the lower panel, though, the results show that measurement error is a problem in the multi-establishment industries. They show no significant responsiveness in any category and the coefficients are much smaller (albeit standard errors large enough that one cannot reject they are the same as those in the upper panel). Fortunately, in the area of retail trade, firms with a low number of establishments are the norm. The average across all state-industries in the sample is only around 1.8 establishments per ...