Review of Austan Goolsbee, 2000 JPE What Happens When You Tax the Rich?

Transcription

1 Review of Austan Goolsbee, 2000 JPE What Happens When You Tax the Rich? Evidence from Executive Compensation Joseph Guse Department of Economics, Washington & Lee University Senior Seminar in Public Finance, Econ 398, Sept 2016 (revised)

2 Goolsbee s Research Question What is short and long-run earning elasticities w.r.t changes in the Marginal Tax Rate (MRT)? More Specifically, how did executives at the 1500 largest companies respond to the 1993 tax bill. 8% point increase in the MTR on 140K to 250K 11% point increase in the MTR on income over 250K

3 Motivation. Feldstein and Feenberg (1996) find large response. Goolsbee thinks their methods are flawed. Why Care? Central Question in Policy Debate. Shape of Laffer Curve. Size of Deadweigt Loss Questions are especially relevant for high income earners Data on high income households is typically poorly suited to answer this question. (e.g. household income surveys, tax returns)

4 Preview of Main Findings Theory. Following Slemrod, Goolsbee lays out three types of responses to changes in MRT. 1 timing shift 2 reporting behavior 3 real behavioral response (actually working less) Empricial Findings Goolsbee finds that timing shifts can account for almost the entire response to the 1993 changes. Minor reporting response: shift income from taxable to non-taxable forms. No evidence of real behavioral response.

5 Contribution and Improvements over Previous Work Previous studies use tax return data. Goolsbee uses Execucomp data. Compared to his data, tax return data does not contain good detail on form of compensation (e.g. salarly, stock options) timing of compensation and in particular the timing of option exercises Focus on executives. Advantage: by focusing on wealthy (and the 1993 act), avoid problem with spurious correlation with tax cuts and general trend of rising inequality Disadvantage: executives are not representative of wealthy tax payers

6 Comparison of Goolsbees Findings with Previous Lit Previous studies in the New Tax Responsiveness literature reported large elasticities - even exceeding 1.0. (Think wrong side of Laffer Curve) Goolsbee results suggest that long-run elasticities are at most.4 and probably closer to 0. Goolsbee reconciles these differences by point out that previous authors were not really measuring long-run and not accounting for timing shifts.

7 Empirical Model Basic Regression Equation log(inc it ) = α i +βlog(1 tax it+1 )+δlog(1 tax it )+X it Γ+e it β. Elasticity of present income w.r.t. future net-of-tax rates. δ. Elasticity of present income w.r.t. pres. net-of-tax rates. β + δ. Non-transitory elasticity. (i.e. long run ) X. is a set of control variables. Timing Shift Hypothesis. An anticipated decrease in the net-of-tax rate (due to announced tax increase) should prompt executives to shift income from the future to the present (e.g. via option exercises). If so, β < 0 and δ > 0

8 Data 5 top executives from 1500 corporations. Compensation reporting is required by securities regulations. Cover years 1991 to 1995 Drop firms whose fiscal year does not end in Decmember. Focus on execs with at least 4 years in data. 40,333 obsversations (executive years) trimmed to 21,299. median income in sample $451K

9 Assumptions on Sample Married Filing Jointly Spouse does not work no AMT no other income sources all options are NQOs (95% on average)

10 Discussion Points what exactly do the coefficient measure in the log-log spec? What is the endgeneity concern on page 368? What does the first difference approach accomplish? Iterpret the coefficients in Tables 3,4,5 Net elasticities tend to fall in the.4 range in most specification. What does this say about DWL concerns? Laffer curve concerns?

11 Log-Log Specifications (Aside) Suppose that log y = α+βlog x Why is β an elasiticty? Transform by taking exp of each side α+β log x y = e y x = β x eα+β log x = β x elog y = βy x β = y x x y Recall Definition of Elasticity of y w.r.t. changes in x. E yx is (roughly) the percentage change in y for a 1% increase in x.

12 Net-of-Tax Rate Elasticities and the Laffer Curve Stylized Model. Tax Revenue (R) is the product of the rate (t), the wage (w), and hours worked (h(ŵ)) which may depend on the net of tax wage rate ŵ w(1 t). R = t w h(ŵ)

13 Net of Tax Rate Elasticities and the Laffer Curve Cont. Laffer Curve. To be on the wrong side of it.. dr dt < 0 w h(ŵ) t w h (ŵ) < 0 h (ŵ) h(ŵ) > 1 wt h (ŵ) h(ŵ) ŵ > ŵ wt h (ŵ) h(ŵ) ŵ > 1 t t Critical tax rates for various elasticities supply critical elasticity tax rate