Perhaps the most intriguing question related to the taxable-income elasticity is how much of it reflects adjustments in labor supply (reflected either in hours worked or work intensity), and how much of it reflects other types of adjustment? The question, by its nature, is very difficult to answer since researchers are very rarely able to connect data on hourly wages or hours worked to tax data. Also, it is hardly possible to measure the extent of tax evasion (e.g., income underreporting) or tax avoidance (e.g., tax exempt and non-reportable forms of remuneration).
We are able to conduct, however, four tests suggesting that our results are not a result of income shifting. In particular, in this section we show that (1) higher income elasticities are estimated for women, the young and the old, (2) taxpayers who only have wage income exhibit a very similar response to tax changes than taxpayers who have other sources of income, (3) we find no evidence for shifting of wage income into capital income; and (4) no evidence for switching from employer-filed tax forms to individual reporting.
The first of these findings is consistent with previous studies finding that labor supply of women is more sensitive to wage incentives than that of men (see, e.g. the survey of Meghir and Phillips, 2008), while no alternative explanation related to tax evasion or tax avoidance would predict this asymmetry between the sexes. On the contrary, Meghir and Phillips note that with respect to tax avoidance, one should rather expect the opposite asymmetry. Similarly, with respect to tax evasion Semjén et al.
(2009) have found with a survey methodology that men are almost twice as likely to be paid partly or fully in cash, than women, in Hungary.25 A similar argument can be made regarding the age groups:
presumably, older and younger age groups can increase or decrease their work effort (or even working
24The typical taxpayer below the pension contribution ceiling has 2,9 (53/51.5) percentage expected change of METR, and 10,5 (47.5/43) percentage above the ceiling (see the left bottom panel of Figure 2.1, and the bottom panel of Table 2.2). 728 atypical taxpayers were excluded with larger than typical changes in absolute value, and an additional 3 taxpayers with smaller.
25In the survey, 19% of men and 11% of women said that they received such unreported payments in the course of the two years prior to the survey (Semjén et al., 2009, pp. 233−234).
CEUeTDCollection
hours) more easily than those between 30 and 55 years, who are more likely to work full-time in the first place. It is harder, however, to argue that older and younger groups are more prone to tax avoidance or tax evasion.
Result (2) can also be interpreted as indirect evidence against the tax evasion explanation of our results. Arguably high-income taxpayers who only have wage income have less opportunities for tax avoidance than individuals with multiple sources of income (e.g., including contract work). If the elasticity we estimate for this group is similar to that of other groups, it may be an indication that the behavioral response is not simply a result of tax avoidance. This evidence does not definitively settle the question whether the estimated elasticity reflects real labor supply adjustment at the intensive margin, but it provides some evidence against some alternative explanations.
Result (3) and (4) provide direct evidence for that tax evasion is not the main source of the taxable income elasticity. If tax shifting explained much of the estimated elasticity, we should observe a higher increase in capital income of those individuals who are likely to become subject to the extraordinary tax.
Similarly, a larger share of taxpayers who are likely to be affected would switch from employer reported tax forms to individual reporting, this latter presumably providing more possibility for income-shifting.
Contrary to what could be expected based on the income-shifting explanation, there is no indication for these in the data.
2.5.1 Higher income elasticities are estimated for women, the young and the old
To see whether different demographic groups exhibit different behavior, regressions of the main spec-ifications are run for the sexes and age groups separately. Results show that the marginal tax rate influences the taxable income of all subgroups, albeit to a different degree. Table B.4 of the Appendix shows the regression results for women and men separately. The first two columns show the results for women and the last two columns for men. In the regressions reported in column (1) and (3) the average net-of-tax rate was omitted as an explanatory variable. The table shows that estimated coefficients are higher for women (0.29-0.32) than for men (0.21-0.24). For both sexes separately, the coefficient is statistically significant.
The estimated coefficients of the control variables show that the sexes are affected differently by factors controlled for in the estimations. Notably, age affects the taxable-income growth of the sexes in the opposite way: it affects income growth positively for women in the sample, but negatively for men; both effects are highly statistically significant. It is likely that this finding is caused by the fact that many younger women reduce their labor supply when they have young children. The estimated effect of the type-of-locality variables is rarely statistically significant in the sub-samples but is broadly
CEUeTDCollection
in line with overall findings. The coefficient of initial income (tax base in 2005) is negative for men, positive for women, but it is not statistically significant in either case. The income effect seems to be much stronger for women.
In the next step, regressions were run separately for different age groups. Taxpayers were divided into three groups: below 30 years, between 30 to 55 years, and above 55 years (as of 2005). Table B.5 of the appendix shows the results. As above, results from two specifications are reported for all three groups: the full specification is reported in the even columns, while in the odd columns the average net-of-tax rate is omitted as a control variable. The results of the odd-numbered columns are interpreted here, as the inclusion of the average net-of-tax rate makes the estimation of smaller groups unstable (especially column (2) and column (6)). In these two cases the first-stage equation for the average net-of-tax rate is not well specified (or the synthetic average net-of-tax rate is not strong enough as an instrument) as testified by the low values of the Kleibergen-Paap weak identification F-statistics reported in column (2) and column (6). These are the only instances in the analysis where we have a weak instrument problem, caused probably by the small number of observations in these subgroups.
Interpreting the results of the specifications without the income effect, we find that the taxable-income elasticity is estimated to be lowest for taxpayers between 30-55 years of age (a coefficient of 0.17), while it is larger for those under 30 years (0.67) and those above 55 years of age (0.33). The estimated coefficient of the marginal net-of-tax rate is significant at the 5% level (at the 10% level for taxpayers below 30 years). The coefficients for the subgroups indicate that younger and older taxpayers have a higher elasticity than those in-between.
Turning to the control variables we find that the difference in income growth between women and men is affected by age. High-earning women’s disadvantage in income growth is strongest for those under 30 (here the difference is almost 33 percentage points and statistically strongly significant); the disadvantage is above 10 percentage points for those older than 55 years, but here the high variance makes the effect statistically insignificant. The disadvantage of high-earning women between 30 and 55 is about 4 percent and statistically significant.
2.5.2 Similar elasticity is estimated for individuals with wage income only
For this exercise we divided our baseline sample into two groups and repeated the analysis separately for those taxpayers who had only wage income in 2005 (4,239 observations), and the rest (2,656 observations). Additionally, we repeated the analysis for the subsample of taxpayers who had at least some capital income in 2005 (714 observations). The results are reported in Table B.4 of the Appendix.
CEUeTDCollection
The estimated elasticities are very similar in both samples: the estimated coefficient of the marginal net-of-tax-rate is 0.252 in the wage-only, 0.235 in the not-only-wage, and 0.215 in the some-capital-income subsample. The some-capital-income effect is also very similar in the subsamples (-0.841, -0.845, and -0.807, respectively, even though it is not statistically significant in the least numerous capital-income subsample).
Control variables show some differences across subsamples, although not dramatic ones. (Controls that were statistically insignificant in both larger subsamples have been dropped.) The sign of the coefficient of initial income is different in the wage-only and not-only-wage subsamples, but the magni-tude is small in both cases and is statistically insignificant. The interaction of gender and age seems to be significant only in the not-only-wage subsample; here we get the pattern seen in the baseline results, while the interaction terms are insignificant in the wage-only subsample. Missing gender information, on the other hand, is smaller and insignificant in the not-only-wage sample.
In sum, based on the results of the full specification, individuals with wage income only seem to exhibit a similarly sensitive reaction to tax changes as others, contrary to the prediction of the tax-avoidance explanation.
2.5.3 No evidence for income shifting I
If tax shifting explained much of the elasticity estimated in this paper, we should observe a differen-tial increase in capital income of those individuals who are likely, ex-ante, to become subject to the extraordinary tax. In this spirit, we divided our baseline sample to two sub-groups: individuals who, based on the average growth rate of income, are expected to be subject to the extraordinary tax in 2008 (the ‘higher-income group’), and those who are not (‘lower-income group’).
Since three new types of capital income were defined between 2005 and 2008, we applied the 2005 definition also in 2008 to keep the two years comparable. The new types of income are not very significant: combined, they represented about 3% of capital income in our high-income sample. It is thus not surprising that, repeating the same exercise with contemporaneous definitions of capital income, we get the same results.
The lower-income group consists of 3738 taxpayers, while there are 3151 taxpayers in the higher-income group. Six outliers were excluded from the sample: these were cases where an individual received capital income of HUF 100 million (about EUR 400,000) or higher. The income earned by these six individuals was great enough to move the results; the results are robust to any further restriction on the sample. The summary statistics of this comparison are shown in Table 2.4.
Contrary to what could be expected based on the income-shifting explanation, there is no indication
CEUeTDCollection
in the data that the higher-income group increased its capital income to a greater extent than the lower-income group. Indeed, the share of taxpayers that has reported positive capital lower-income grew more in the lower-income group (a growth of 1.3 percentage points compared to 0.1). Average capital income stayed largely flat in both groups, increasing by a mere HUF 17,000 (EUR 68) for the lower-income group as opposed to HUF 10,000 (EUR 36) for the higher-income group. The results stay the same if we compare capital income of both groups as a share of 2005 tax base or as a share of contemporaneous tax base.
Table 2.4: The behavior of capital income in the high-income sample, 2005-2008
Lower group Higher group
Income in 2005 HUF 5–6.12 m. HUF 6.12–8 m.
No. of observationsa 3738 3151
% having capital incomebin 2005 10.0 10.8
% having capital incomebin 2008 11.3 10.9
Average capital income in 2005, HUF thousand 248 164
Average capital income in 2008, HUF thousand 266 174
Average increase of capital income, 2005-2008, HUF thousand +17.4 +10.2 Capital income as a share of taxable income in 2005 (average) 4.49% 2.39%
Cap. inc. in 2008 as a share of 2005 taxable income (average) 4.81% 2.55%
Increase of capital income between 2005 and 2008, as a share of 2005 taxable income (average)
0.31% 0.16%
a Six outliers were removed; these were instances of capital income above HUF 100m. Subsample averages were sensitive to these outliers but not to further restrictions on the data.
b We applied the 2005 legal definition to generate a comparable capital income for 2008. Additional items became taxable as capital income in the years between 2005 and 2008. The inclusion of these items into the definition of 2008 capital income, however, does not change the results.
2.5.4 No evidence for income shifting II
Similarly to the previous exercise, if tax evasion explained much of the estimated elasticity, we should observe that a larger share of taxpayers who are likely to be affected by the extraordinary tax would switch from employer reported tax forms to individual reporting, this latter presumably providing more possibility for income-shifting. In Hungary employees can decide whether they want to submit their income report independently or get it submitted by their employer.26 During the period of 2005 and 2008 the overall share of self-reported tax forms among the universe of taxpayers increased from 55% to 77%.
26The taxpayers have to obligatory self report their income several cases, for instance if the taxpayer has mainly separately taxed income, if the employer declines the request of the taxpayer to forward the tax application form, or if the taxpayer’s main occupation is self-employment, or the taxpayer has no employer at the last day of the tax year, and if he determines his cost deductions based on expenses.
CEUeTDCollection
Table 2.5: The behavior of capital income in the high-income sample, 2005-2008
Lower group Higher group
Income in 2005 HUF 5–6.12 m. HUF 6.12–8 m.
No. of observationsa 3738 3151
% individual reporting in 2005 55% 57%
% individual reporting in 2008 69% 70%
Percentage point increase in individual reporting share 15% 13%
a Six outliers were removed as in the previous subsection; these were instances of capital income above HUF 100m.
The sample is divided up to lower-income and higher-income groups similarly as in the previous subsection where we analysed capital income reporting. The share of self-reporting taxpayers grew more in the lower-income group, compared to those in the higher-income group (a growth of 15 percentage points compared to 13). Contrary to what could be expected based on the tax evasion explanation, there is no indication in the data that the higher-income group increased self-reporting to a greater extent than the lower-income group.
In sum, there is no indication that income-shifting increased more for the group affected by the extraordinary tax.