Gender tax difference in the U.S. income tax

Abstract

Unmarried women face a significantly lower average federal income tax rate than unmarried men, 6.3% versus 10.9%. Some of the difference arises because women have lower income on average and the tax system is progressive. Using a non-parametric decomposition analysis, we show that tax progressivity accounts for less than 60% of the gender tax rate difference, leaving the rest being explained by gender differences within income classes. This conclusion remains when the decomposition exercise uses an equivalence-scale-adjusted income and considers gender differences in nontaxable income and time use. Regression results reveal that much of the gender tax difference arises because unmarried women are more likely to live with dependents, making them more likely to claim child-related tax benefits relative to unmarried men. Our findings indicate that the current tax system places a high value of raising children beyond the consideration of the families’ larger consumption needs. Because our analysis does not consider the higher risk of economic insecurity facing single parents and the various externalities generated by women’s higher commitment to childcare, future research should focus on determining whether these additional considerations could justify the extent of the gender tax differential.

Appendix

This Appendix first explores how to generalize the decomposition approach from Slemrod (2022) under two income classes to a multiple income class case, and then provides the data used to calculate the 20-group decomposition exercises discussed in the text. The main purpose of decomposition exercise is to understand how much of the gender difference in the average tax rate can be explained by gender difference in income distribution and thereby tax progressivity, and by group tax rate difference within same income classes, horizontal group equity.

As background, we first reproduce the two-class decomposition from Slemrod (2022). In this case, the Group Average Tax Rate Differential (GATRD) is equal to:

$$GATRD = \left( {\frac{{n_{BL} }}{{n_{B} }} - \frac{{n_{AL} }}{{n_{A} }}} \right)AP + \left( {1 - \frac{{n_{AL} }}{{n_{A} }} \times \frac{{n_{BH} }}{{n_{H} }} - \frac{{n_{BL} }}{{n_{B} }} \times \frac{{n_{AH} }}{{n_{H} }}} \right)HGEH + \left( {\frac{{n_{AL} }}{{n_{A} }} \times \frac{{n_{BL} }}{{n_{L} }} + \frac{{n_{BL} }}{{n_{B} }} \times \frac{{n_{AL} }}{{n_{L} }}} \right)HGEL,$$

where the two groups are A or B and two income classes L or H. Denote the number of people in group G and income class Y as ${n}_{GY}$ whereas ${n}_{Y}$ and ${n}_{G}$ are the number of people in income class Y and group G, respectively. The ratio ${(n}_{GY}/{n}_{G})$ is thus the fraction of people of group G in the income class Y. The term $AP$ is average progressivity and $HGE$ is horizontal group equity, defined as

$$AP \equiv t_{H} - t_{L}$$

$$HGEL \equiv t_{AL} - t_{BL}$$

$$HGEH \equiv t_{AH} - t_{BH} ,$$

where ${t}_{Y}$ is the average tax rate for all members in the income class Y whereas ${t}_{GY}$ is the average tax rate for members in income class Y and group G. These equations show that GATRD can be decomposed into the effect of tax progressivity interacted with gender differences in the income distribution and a set of income-class-specific measures of horizontal group differences.

The multiple income class analogue is as follows:

$$\begin{aligned} GATRD & = \mathop \sum \limits_{j = 2}^{N} \left[ {\mathop \sum \limits_{i = 1}^{j - 1} \left( {\frac{{n_{Bi} }}{{n_{B} }} - \frac{{n_{Ai} }}{{n_{A} }}} \right)AP_{j} } \right] + \left[ {1 - { }\left( {\mathop \sum \limits_{i = 1}^{N - 1} \frac{{n_{Ai} }}{{n_{A} }}} \right) \times \frac{{n_{BN} }}{{n_{N} }} - \left( {\mathop \sum \limits_{i = 1}^{N - 1} \frac{{n_{Bi} }}{{n_{B} }}} \right) \times \frac{{n_{AN} }}{{n_{N} }}} \right]{ }HGE_{N} \\ & \quad + { }\mathop \sum \limits_{i = 1}^{N - 1} \left[ {\left( {\frac{{n_{Ai} }}{{n_{A} }} \times \frac{{n_{Bi} }}{{n_{i} }} + \frac{{n_{Bi} }}{{n_{B} }} \times \frac{{n_{Ai} }}{{n_{i} }}} \right)HGE_{i} } \right]. \\ \end{aligned}$$

where i and j are the indexes of N income classes and ${AP}_{j}$ is the difference in the average tax rate between income class j and class j-1. Specifically,

$${AP}_{j}\equiv {t}_{j}-{t}_{j-1}$$

$${HGE}_{i}\equiv {t}_{Ai}-{t}_{Bi}.$$

Plugging ${AP}_{j}$= ${(t}_{j}-{t}_{j-1})$ and ${HGE}_{i}$= (${t}_{Ai}-{t}_{Bi}$) into GATRD yields

$$GATRD = \mathop \sum \limits_{i = 1}^{N} \left[ {\left( {\frac{{n_{Ai} }}{{n_{A} }} - \frac{{n_{Bi} }}{{n_{B} }}} \right)t_{i} } \right] + \mathop \sum \limits_{i = 1}^{N} \left[ {\left( {\frac{{n_{Ai} }}{{n_{A} }} \times \frac{{n_{Bi} }}{{n_{i} }} + \frac{{n_{Bi} }}{{n_{B} }} \times \frac{{n_{Ai} }}{{n_{i} }}} \right)\left( {t_{Ai} - t_{Bi} } \right)} \right].$$

Again, the first term is the effect of tax progressivity interacted with gender differences in the income distribution and the second term is the effect of income-class-specific measures of horizontal group differences.

In the next four tables, we present the information needed to perform the decomposition exercises discussed in the text, for each of four cases listed in Table 3, following the GATRD equation. In each table, columns (1) and (3) display the sizes of ${n}_{Ai}$ and ${n}_{Bi}$, which are the taxpayer counts in Tables 7 and 9 for person-weighted average tax rates and the amounts of taxpayer income in Tables 8 and 10 for income-weighted average tax rates. Note that, in all four tables, each income bin contains roughly the same number of taxpayers, but the weights, $\frac{{n}_{Ai}}{{n}_{A}}$ and$\frac{{n}_{Bi}}{{n}_{B}}$, are the fraction of the total in terms of either taxpayer counts or taxpayer income, depending on the measure of the average tax rate. Columns (2) and (4) show the average tax rates by gender and income group, ${t}_{Ai}$ and${t}_{Bi}$. These are the tax rates depicted in Figs. 1, 2, 3 and 4, corresponding to those from Tables 7, 8, 9 and 10. Column (5) shows the average tax rate for each income class, ${t}_{i}.$ Like the group average tax rates, the tax rates for each income class, including${t}_{Ai}$,${t}_{Bi}$, and${t}_{i}$, are either person-weighted or income-weighted within the income class.

Table 7 Decomposition of the gender difference in average tax rates (person-weighted), AGI bins

Table 8 Decomposition of the gender difference in average tax rates (income-weighted), AGI bins

Table 9 Decomposition of the gender difference in average tax rates (person-weighted), equivalence-scale-adjusted AGI bins

Table 10 Decomposition of the gender difference in average tax rates (income-weighted), equivalence-scale-adjusted AGI bins

With these fractions and tax rates, we compute the magnitudes of the various GATRD components. For each income class, column (6) shows the magnitude of ($\frac{{n}_{Ai}}{{n}_{A}}-\frac{{n}_{Bi}}{{n}_{B}}){t}_{i}$, column (7) shows the within-income tax rate difference or ${HGE}_{i}$, i.e., (${t}_{Ai}-{t}_{Bi}$), and column (8) shows the magnitude of $\left(\frac{{n}_{Ai}}{{n}_{A}}\times \frac{{n}_{Bi}}{{n}_{i}}+\frac{{n}_{Bi}}{{n}_{B}}\times \frac{{n}_{Ai}}{{n}_{i}}\right)*\left({t}_{Ai}-{t}_{Bi}\right)$. The sum of column (6) and column (8) across all income classes corresponds to the share of GATRD explained by tax progressivity interacted with gender income difference and by horizontal group differences, respectively. These shares vary little with the number of income bins specified once the number of bins reaches 20.