• No results found

Testing against alternative hypotheses

In document Essays on the Economics of Taxation (Page 147-151)

Chapter 2 Tax Evasion and Avoidance among U.S Households

2.4 Estimating evasion and avoidance

2.4.7 Testing against alternative hypotheses

Is it possible that the above evidence is also consistent with hypotheses other than

tax evasion and avoidance? Certainly yes. Can we provide additional evidence to support

our hypothesis of tax evasion and avoidance? The answer is also yes.

First, to account for the drop in reported income conditional on consumption when

a family turns from an employee to self-employed, an alternative hypothesis is that the

self-employed consume more with the same true income than employees due to more social

activities. To test the tax evasion hypothesis against this hypothesis, we investigate the

response of the supposed income under-reporting to tax rate change. If the supposed income

Table 2.13. Tax saving (PIT+CIT+payroll tax) from evasion v.s. avoidance (1979 dollars)

Notes: The figures here are based on the estimates in previous tables.

when tax rate is higher, since higher tax rate increases the marginal benefit of evading tax.

Such a test has the same spirit as that did by Gorodnichenko et al. (2009), who show that

the reported consumption-income gap (as a proxy for tax evasion) decreased when the flat

tax reform (a uniform drop in marginal PIT rate for higher income earners) was adopted in

Russia. Without a similar simple tax reform in the U.S., we explore the change in highest

statutory PIT rates to test our hypothesis. The advantage of using statutory highest tax

rate rather than the effective tax rate of each family is in its exogenous property.19 Although

the general families may not reach the highest bracket, the change in statutory highest tax

rate may still well reflect the change in tax rate a family actually faces, since tax rates in all

brackets often increase or decrease together.

To test this we apply the following specification:

lnYit=β0+β1lnCit+β2Selfit+β3Selfit·P ITt+γ·Xit+λt+µi+εit,

19I have also tried to use the effective marginal tax rate (EMTR) instrumented by the predicted EMTR based on income in previous years to test the hypothesis, which renders the opposite sign ofβ3.I guess the endogeneity problem due to the strong positive correlation between income and EMTR is hard to break out even when the traditional instrument is applied. A method to solve this problem might be contributing to an independent paper. Here I thus only show some suggestive evidence.

whereP ITt is the statutory highest personal/corporate income tax rate (in decimal).

We expect to see a negative sign of β3 since higher tax rate would increase the benefit of

evading tax.

Second, to account for the drop in labor income fraction (i.e. the increase in asset

income portion) conditional on reported total income when a family turns from an employee

to self-employed, an alternative hypothesis is that the income structure naturally changes

when a family becomes self-employed. To test the tax avoidance hypothesis against this, we

investigate the response of asset income to the change in PIT-CGT (long-run capital gains tax)

rates gap, conditional on total reported income. If the increase in asset income corresponds

to income-shifting for tax saving purposes, then when the PIT rate increases relative to the

CGT rate, the self-employed families would shift more ordinary income to capital gains. The

reason we look at asset income is that it includes all capital gains income and even in years

that we cannot obtain an estimated capital gains, PSID provides a consistent measure of

asset income. A caveat is that while the (long-term) capital gains face a lower tax rate, the

other parts in asset income face the same marginal tax rates as ordinary income.20 So again

we are proposing a suggestive test here. To do this we apply the following specification:

lnAssetIncit=β0+β1lnYit+β2Selfit+β3Selfit·(P ITtCGTt) +γ·Xit+λt+µi+εit,

where AssetIncit is family asset income, P ITtCGTt is the rates gap between the

PIT and the CGT. We expect to see a positive sign forβ2 and β3.

Table 2.14 shows the results. In column 5 we exclude corporate self-employed because

20An exception is that since 2003 dividends income have faced a lower tax rate. But the post-2003 data only consists a small part in our 1967-2011 sample. By excluding the post-2003 period, our result does not change.

Table 2.14. Responses of evasion and avoidance to tax changes

Notes: The sample in column 1-8 focuses on the first employee-self-employed cycle of families ever turning from employees to self-employed within sample period. Other controls include dummies for age, education, marital status, race, state, head’s hours worked, homeownership, disability, current family scale, and year dummies. Standard errors clustered at family level are in parentheses. Estimates are weighted using the sampling weights of respective surveys. *** p<0.01, ** p<0.05, * p<0.1.

corporations (in particular C corporations) face corporate tax, which may affect their income-

shifting activities in a non-linear way. Doing this does not change much of our results. We

see expected signs in both tests, though the estimates are not precise because the changes in

statutory tax rates are yearly and the highest rate may not well apply to the actual bracket

people face. We thus regard this result as a preliminary evidence to support our tax evasion

and avoidance hypotheses, rather than draw a definitive conclusion. Controlling for family

fixed-effect or not does not change our results.

We make some preliminary interpretation on our estimates. During the 1986 tax reform,

the highest PIT rate decreases from 0.5 in 1986 to 0.28 in 1988, with a transitional rate 0.385

in 1987. Column 2 implies that the under-reported income decreased by 0.077∗(0.5−0.28) = 0.017 = 1.7% due to this tax change. During the 1986 tax reform, the PIT-CGT rates gap

decreased from 0.3 in 1986 to 0 in 1988, with a transitional rate gap 10.5 in 1987. Column

4 implies that this tax reform had the self-employed families reduce income shifting by

0.747∗0.3 = 0.224,i.e. 22.4% of their total income would have been transferred from asset to labor account during this tax reform.

In document Essays on the Economics of Taxation (Page 147-151)