On the Estimation of Payroll Tax Incidence: Reply

Asher makes two points and draws a conclusion from them. The first is that I misinterpreted my OLS results, and that these cannot give an estimate of backward shifting. He is 100 percent correct; I made a mistake (an honest one, I believe). OLS is entirely inapplicable to this problem, because its use inherently means that some (complicated nonlinear) function of the dependent variable is included as an independent variable. The second, related point is that there is a simultaneity between tax rates and wages that stymies any attempt to derive satisfactory estimates of the extent of shifting. There is no doubt that such simultaneity exists. It is also the case that instrumental estimates of the kind presented in Table III of the original paper [2, 1216] take care of this, but they inherently induce a substantial degree of multicollinearity into the estimating equation. The conclusion Asher draws, that one cannot use cross-section data to infer the extent of shifting, is wrong. One can use them, but one must be careful with econometric technique. In the original paper I presented equation (5) [2, 1213]: