BRIEF ARTICLE

N a tio n a l Tax J o u rn a l, J u n e 2 0 1 5, 6 8 (2), 2 5 1 - 2 8 0 h ttp ://d x .d o i.o rg /! 0.17310/ntj.2015.2.02

WHITHER THE MARRIAGE TAX?

Ja m es A im a n d J. S e b a s tia n L e g u iz a m o n

We use household data from the Current Population Survey to calculate how the real value ofthe so-called “marriage tax ” or “marriage subsidy ” in the federal individual income tax has changed over the period 1969 to 2009. We examine three issues: the magnitude o f the marriage tax/subsidy and its evolution over time, its effects on the distribution o f income (including the effects o f different demographic characteris­ tics on the magnitudes and trends), and the causal factors in its evolution (e.g., tax changes, demographic changes). We fin d that the tax treatment o f the fam ily has changed significantly over time, from a large average marriage bonus in 1969, to a large marriage tax in much o f the 1990s and early 2000s, to a large marriage subsidy since 2003. We also fin d that the marriage tax varies significantly and sys­ tematically by income level, as well as by the number o f children in the family, the earnings ratios o f the spouses, the race o f the family, and the age o f the household head. Finally, we fin d that changes in income and fam ily composition have influenced the magnitudes and trends o f marriage taxes and subsidies, but that adjustments in the federal income tax code account fo r most o f the observed changes.

Keywords: marriage, individual income tax, marriage tax, taxable unit

JEL Classification Codes: H24, J12, JI6

I. INTRODUCTION

It is well-recognized that the U.S. individual income tax generates implicit penalties or bonuses for marriage (Aim, Dickert-Conlin, and Whittington, 1999). Many couples pay higher taxes when married than their combined tax liabilities as single filers (a “marriage tax” or “marriage penalty”), while many other couples see their joint taxes fall with mar­ riage (a marriage bonus’ or “marriage subsidy”).1 The existence of a marriage tax strikes

1 There are also numerous other federal (and state) programs that treat married couples differently than single individuals, such as those that determine social insurance, veteran, and immigrant benefits. The exact number o f these programs has varied, but is quite substantial. For example, the U.S. General Accounting Office (1996) identified 1,049 federal laws in which marital status is a factor (e.g., welfare programs, veteran benefits, immigrant benefits, social insurance benefits), and 59 provisions in the individual income tax code alone that contributed to a marriage tax/subsidy. The many tax and benefit changes since then have no doubt increased these numbers.

Ja m e s A im : D e p a r tm e n t o f E c o n o m ic s , T u la n e U n iv e rs ity , N e w O rle a n s , LA, USA (ja lm @ tu la n e .e d u )

J. S e b a s tia n L e g u iz a m o n : D e p a r tm e n t o f E c o n o m ic s , W e s te rn K e n tu c k y U n iv e rs ity , B o w lin g G ree n , KY, USA (j.s e b a s tia n .le g u iz a m o n @ w k u .e d u )

252 N a tio n a l T a x J o u r n a l

most people as an unjustifiable feature of the tax and transfer system, and many policy debates over the years have centered on its elimination; the presence of a marriage subsidy does not seem to elicit as heated a discussion. Many o f the tax changes passed in the last decade have tended to reduce the marriage tax, mainly by reducing tax rates, changing standard deductions, adjusting tax brackets, expanding tax preferences, and reforming welfare (Steuerle, 2006).2 Even so, there are still significant income tax consequences from marriage, created by two basic conditions in the individual income tax and its associated programs: basing the tax on household rather than individual income, and imposing the tax at different marginal tax rates at different levels o f income.

Discussion and analysis of the marriage tax, however, seem largely to have disap­ peared from much current policy discussions.3 In particular, the actual current mag­ nitude of the marriage tax and its evolution over time, its effects on the distribution of income, and the causal factors in its evolution (e.g., tax changes and demographic changes) have not been fully quantified or recognized, especially in the last decade when significant tax changes have been implemented. Also, there have been major changes in family structures in the United States since the 1950s, when the “traditional” household was typically a single-earner household with a stay-at-home spouse. Such households still exist, but there are also many other “non-traditional” households. The ways in which all of these households are treated by the tax code may have significant effects on the distribution o f taxes and o f income. However, these effects have not been examined.

In this paper, we use detailed individual data from the Current Population Survey (CPS) to calculate the magnitude of the marriage tax for actual households. We quantify and update changes in the marriage tax over a long period (1969-2009), using a common methodology and a widely used and available data set (the March CPS); in particular, we quantify and update the changes that have occurred since the Bush tax cuts in the early 2000s. We also examine how the marriage tax varies by income level, as well as by various demographic characteristics (e.g., number of children, spousal earnings ratio, race, and age). Finally, we disentangle some of the factors that have generated changes over time in the marriage tax — tax code changes versus family income/demographic changes — by using appropriately deflated (inflated) family and wage information for the year 2010 (or 1970) from the CPS, together with detailed information on the tax code over all years o f our analysis. We are therefore able to calculate the marriage tax holding family structure constant; we also are able to hold the tax code constant for any given year and change family structure over time. These decompositions enable us to determine whether it is primarily tax code changes or family structure changes that have driven changes in the marriage tax over time.

2 For simplicity, we refer in much o f the discussion that follows to the “marriage tax,” rather than to the more accurate “marriage tax/marriage subsidy.”

3 There are of course some exceptions here. Most notably, the Brookings Institution-Urban Institute Tax Policy Center has a useful website that explores various aspects of the marriage penalty, http://www. taxpolicycenter.org/taxtopics/Marriage-Penalties.cfin.

W h i t h e r t h e M a r r ia g e Tax? 253

We find that the 2009 individual income tax is far from marriage neutral, with a large average marriage bonus but significant variation in size across households. Using a longer sample period than most recent studies, we also find that the tax treatment of the family has changed significantly over time: from a large average marriage bonus in 1969, to a large marriage tax in much o f the 1990s and early 2000s, to a large mar­ riage subsidy since 2003. Indeed, the percentage o f families facing a marriage tax has fallen significantly in recent years, while the percentage receiving a subsidy has risen over this same period. Further, we find that the marriage tax varies significantly and systematically by income level, as well as by the number o f children in the family, the earnings ratios o f the spouses, the race o f the family, and the age of the household head. Finally, we find that changes in income and family composition have influenced the magnitudes and trends o f marriage taxes and subsidies, but those adjustments in the federal income tax code account for most o f the observed changes.

II. SOME PRINCIPLES IN TAXING THE FAMILY

Countries have a variety of goals in choosing the structure of the individual income tax.4 One obvious goal is to achieve “equity” in taxation. “Horizontal equity” requires that taxpayers who are equal in all relevant respects pay equal amounts o f taxes (Kaplow, 1989; Musgrave, 1990). The difficulty here lies in defining “equals.” Equals have typically been seen as married couples with equal family income (and identical characteristics), in which case achieving the goal o f Horizontal Equity Across Households requires that mar­ ried couples with equal incomes pay equal taxes.5 A related goal is Marriage Neutrality, which requires that a couple’s combined tax liability remain unchanged with marriage.

Other goals are also commonly invoked in discussions o f taxation and the family. “ Vertical Equity” requires that taxpayers with greater ability pay greater amounts of taxes, and thus relates to the rate structure of the income tax. It is generally felt that a

4 The classic discussion o f the principles o f taxation is by Musgrave (1959). A more recent discussion is in Kaplow (2008).

5 While the notion o f Horizontal Equity Across Households may seem straightforward, there are in fact some significant difficulties in application. One issue relates to the impact o f home production, in which households o f similar size have different levels o f utility and broadly defined leisure due to home produc­ tion (Laurin and Kesselman, 2011). For example, consider a one-earner couple versus a two-earner couple. Even if these two couples have equal money incomes, their effective incomes are unlikely to be equal. The one-earner couple has a substantial advantage o f an at-home spouse who can produce many valued goods and services. These additional home-produced goods and services include such items as daycare, cooking, laundry, cleaning, repairs, and the like, all o f which give a significant gain over the out-of-pocket expenses that the two-eamer couple would incur to provide similar goods and services. The one-earner couple likely saves also on commuting and other work-related expenses that are incurred by the two-eamer couple. A different issue relates to scaling income to adjust for different family sizes, or household equivalence scales (Aim, Whittington, and Fletcher, 2002). In short, horizontal equity based only on a simple assessment o f equal money incomes between one-earner and two-eamer couples is too simplistic, and its weight rela­ tive to other possible goals in any assessment o f the appropriate tax treatment o f families requires careful consideration.

254 N a tio n a l T a x J o u r n a l

progressive rate structure is best able to achieve vertical equity, which is often referred to as the Progressivity goal o f taxation (Samuelson, 1947; Kiefer, 1984; Young, 1994).

It is now well-known that no individual income tax can achieve the simultaneous goals of Horizontal Equity Across Households, Marriage Neutrality, and Progressivity (Rosen, 1987). Choosing the features of the individual income tax therefore requires that countries must make choices in their pursuit o f these (and still other) worthwhile goals.6

Indeed, countries around the world have made very different choices (Aim and Melnik, 2005). In the United States the income tax treatment of the family has varied since the establishment of the tax in 1913. In its early years, the basic unit o f taxation was the individual, so that each individual was taxed on the basis o f his or her income independently o f marital status. Because tax liability did not change much with mar­ riage, the federal income tax largely achieved Marriage Neutrality at the same time as marginal tax rates that increased with income generated Progressivity. However, the Revenue Act of 1948 changed the unit o f taxation from the individual to the family via the adoption o f income splitting for married couples, partly because taxpayers in community property states were already implementing joint filing on their own. This change meant that couples with equal incomes paid equal taxes; that is, the income tax satisfied the goal o f Horizontal Equity Across Households. However, because o f the progressive tax rates in the income tax, the change also meant that a couple’s joint tax liability could fall when they married, so that the income tax was no longer character­ ized by Marriage Neutrality but was effectively subsidizing marriage via a marriage subsidy or bonus, with the size of the subsidy increasing with the size o f the difference in spouses’ earnings.

This marriage bonus grew over the next two decades, and by the late 1960s it was possible for a single person’s income tax burden to be as much as 40 percent greater than that o f married couple with identical earnings (Rosen, 1987). Public pressure to remedy this disparity led to the adoption o f provisions in the Tax Reform Act of 1969 that established a new and separate tax schedule for single individuals that insured that single persons would incur a maximum tax liability of 120 percent o f a married couple with equal income. However, a side effect of the 1969 legislation was the creation, for the first time, of a widespread and significant marriage tax or penalty for many married couples, especially for couples where both spouses had similar incomes — even though a potential marriage bonus still existed for couples with sufficiently large differences in their earnings.

6 For example, Efficiency (or minimizing the distorting effects o f taxation) is an essential goal o f taxation, a goal that is related to reducing the effects o f taxation on marriage (Marriage Neutrality). Also, there are other dimensions o f equity. It may be argued that equals can be defined as any kind o f “household” with equal income, such as a married couple but also a single individual, an unmarried cohabitating couple, a same-sex couple, an extended family, or even a group of unrelated individuals living together. In this view, an additional aspect o f Horizontal Equity Across Households would require that all o f these households pay equal taxes, an additional and separate goal that Berliant and Rothstein (2003) call Equal Payments by Singles and Couples. However, just as with Horizontal Equity Across Households, this goal ignores the issues o f home production and household equivalence scales, as discussed earlier.

W h i t h e r t h e M a r r ia g e Tax? 255

Since then, various legislative changes have markedly affected the potential for marriage taxes and subsidies, as well as their magnitudes, as demonstrated by Rosen (1987), Feenberg and Rosen (1995), U.S. General Accounting Office (1996), Aim and Whittington (1996), Congressional Budget Office (1997), Bull et al. (1998), Dickert- Conlin and Houser (2002), Eissa and Hoynes (2000), Holtzblatt and Rebelein (2000), Aim, Whittington, and Fletcher (2002), Carasso and Steuerle (2005), Gillette, Holtz­ blatt, and Lin (2005, 2006), Steuerle (2006), Feucht, Smith, and Strawser (2009), and Lin and Tong (2012), among others. However, despite these many studies, we believe there are many aspects o f the marriage tax that merit new analyses. The next sections present our methods and then our results o f these analyses.

III. DATA AND METHODS

To estimate the marriage tax, it is necessary to compare the tax liabilities for mar­ ried couples to their combined taxes assuming that they are single. If the couple pays more in taxes as a married couple than their combined taxes as singles, then they face a marriage tax; if they pay less as a married couple, then the negative difference implies a marriage subsidy.

Data used to determine family characteristics, income, and tax deductions (but not taxes paid) were obtained from the March Current Population Survey (CPS) through the Minnesota Population Center (IPUMS) (Ruggles et al., 2010). The CPS compiles information for a representative sample o f working individuals, recording personal and family characteristics such as marital status, spouse’s working status, total family income, and total number of children in the household. We gather information on wage, business, and farm income for all married individuals in the sample.7 We assume that they claim all possible deductions and credits relating to income and family size, largely in order to measure the marriage tax/subsidy in a consistent way over time.

Our sample includes an average of 50,000 observations per year from 1969 to 2009 (approximately 25,000 couples).8 An important issue in using survey data over such a long period o f time is the possibility for changes in the questionnaire and/or meth­ odology. This is the case for income, including business income. Between 1969 and 1979, questions about income were slightly different than in subsequent years. As one example, after 1979 the CPS redefined “wages” to include overtime pay, tips, bonuses,

7 The CPS reports business income (and also income from sources other than wages and salaries) at the individual level, so that we are able to allocate business income to each individual in a household; that is, business income as reported in the CPS is allocated to individuals, not to households, and we use these individual allocations in calculating individual income.

8 Income and employment data in the March Current Population Survey refer to the preceding year. To calculate a person’s tax liabilities in a given year, we use the following year CPS sample. For example, we use the 1970 CPS to calculate income tax liabilities in 1969, and the 2010 CPS to calculate tax li­ abilities in 2009. As such, even though our data come from the 1970 to 2010 March CPS, the tax analysis captures changes in the marriage tax/subsidy from tax years 1969 to 2009. For consistency, we use tax years throughout the text.

256 N a tio n a l T a x J o u r n a l

and commissions. To the extent that these types of income are concentrated in, say, lower income classes, it is possible that the income reported by these groups will vary overtime due only to this change in definitions. Also, after 1994 the CPS eliminated the underreporting o f labor force participation o f women who were working part-time. The question is whether these (and any other) changes introduce a systematic bias. Such bias is possible. However, we do not see a significant change in reported incomes following the 1979 (versus 1980) or the 1994 (versus 1995) changes, so we do not believe that such bias is important. Also, we have compared our tabulations of income and taxes (both total and per return) with tabulations from the Statistics of Income (SOI) from the Internal Revenue Service for several recent years, and our CPS tabulations are quite comparable to (if slightly lower than) the SOI numbers. The SOI data are derived from a sample of tax returns (not from the entire population), just as are our CPS data.

We input our individual level data into the National Bureau of Economic Research (NBER) TAXSIM model to simulate the estimated tax liability for each household in our CPS subsample o f married individuals, using the relevant year’s income (and other) information and the relevant year’s tax code.

We calculate the marriage tax in three steps. First, we calculate the married couple tax liability. Second, we calculate the tax liabilities o f each person in the household as if each was single, and we then add these two individual tax liabilities to get their combined tax liabilities as singles. In households with only one earner, individuals file as head of household if there are eligible dependents. Couples without any eligible dependents are assumed to file as single individuals when calculating their separate liabilities. Third, the difference in tax liability for the married couple when filing jointly versus the combined tax liabilities when filing as a single person (or head o f household) constitutes each couple’s marriage tax (positive) or subsidy (negative). To generate aggregate estimates for each year, we adjust the estimates using the household weights in the CPS.

In all cases, following Aim and Whittington (1996), we assume that tax liabilities are minimized, whether filing jointly or separately. For example, we assume that individuals itemize their deductions when doing so reduces their liabilities. Also, total deductions available to a married couple must be split for the purposes o f estimating tax liabilities as two singles, and we follow standard practice in assuming that any available deductions go to the higher earner when filing as single while equal earners equally split deductions. Similarly, we assume that child tax credits and deductions are taken by the higher earner or, in the case o f equal earners, we assume that they go to the mother. Additionally, to keep the analysis tractable, we also assume that there are no labor supply responses due to separation or divorce and that there are no marriage, divorce, cohabitation, or timing responses to the marriage tax/subsidy; these assump­ tions imply that each individual’s income remains constant when filing separate tax returns. These are common assumptions in the calculation of the marriage tax, although we acknowledge the existence o f considerable evidence that such behavioral responses exist even if they are relatively small.9

9 For example, see Clarke and Strauss (1996), Aim and Whittington (1999), Dickert-Conlin (1999), and Chade and Ventura (2005). Burstein (2007) reviews much o f this literature and concludes that these responses are generally present but small.

W h ith e r th e M arriage Tax? 257

Note that the Bush tax cuts (or the Economic Growth and Tax Relief Reconciliation Act o f 2001 and the Jobs and Growth Tax Relief Reconciliation Act o f 2003) had sunset provisions that were scheduled to automatically expire at the end o f 2010, but that these provisions were extended for an additional two years in the Tax Relief, Unemployment Insurance Reauthorization, and Job Creation Act of 2010, and many o f these tax cuts were made permanent in the American Taxpayer Relief Act o f 2012. We conduct our analysis using the tax codes relevant for each particular year. If these measures are extended or permanently adopted, it would be reasonable to view 2010 estimates as a valid short-term forecast o f marriage tax estimates; if they are not, the tax code relating to the marriage tax would more closely resemble the 2000 code.

During the 1969-2009 period, there were significant changes in family composition and income, both o f which are strong determinants o f the size o f the marriage tax ( Aim and Whittington, 1996; Eissa and Hoynes, 2000). The percentage o f one-earner mar­ ried couple households decreased significantly over this period, from approximately 51 percent in 1969 to only 33 percent in 2009, with most o f this change occurring by the early 1990s. Average earned income over time (in 2010 dollars) also changed con­ siderably over this period (Figure 1). Although total family income in households with two earners has always been significantly higher than that in households in which only

Figure 1 Real Average Household Earnings Among Married Couples, 1969-2009

O O

--------- — O n e -E a rn e r Fam ily ---------------- T w o -E a rn e r Fam ily

N o te : C a lc u la tio n s a re f o r m a r r ie d in d iv id u a ls fo r a ll ye a rs f r o m t h e M a rc h C u r re n t P o p u la tio n S urvey.

258 N ational Tax Journal

one person works, this gap has widened over time. Figure 1 also shows that the fam ­ ily income o f households with a single earner has remained fairly constant during the past 40 years. Total family income for this group decreased throughout the 1980s, rose during the 1990s, and by 2009 dropped to almost the 1969 level. In contrast, average total family income for tw o-earner couples has consistently increased throughout our sample period. These two phenom ena are partly explained by increasing numbers o f one-earner families becom ing tw o-earner ones.10

IV. RESULTS

Recall that that we are interested in three issues: the magnitude o f the m arriage tax/ subsidy and its evolution over time, its effects on the distribution o f income (including the effects o f different dem ographics characteristics like the num ber o f children, race, and age on the average m agnitudes over time), and the causal factors in the evolution o f the marriage tax (e.g., tax changes versus dem ographic changes). Each o f these issues is exam ined in turn.

A. M a g n itu d e s a n d T ren ds

We first consider the m agnitudes and the trends in estim ated average tax liabilities using year-specific data and tax codes.11 Table 1 presents the average differences between filing separately and filing jointly, w here all values are expressed in 2010 dollars and marriage taxes and revenue gains are shown as positive values while marriage subsidies and revenue losses are shown as negative values. Averaged across all couples, we find that m arried couples have faced a marriage subsidy for most o f the earlier years o f our analysis, 1969 through 1986. Following the Tax Reform A ct o f 1986, the average couple

10 We also estimated the marriage tax/subsidy under several different assumptions: that all individuals take the relevant standard deduction (rather than itemized deductions), that income is adjusted by a wage index (rather than by a price index), and that the children are allocated to the mother (rather than to the higher earner). With the exception of the allocation o f children, the results for these alternative scenarios are quite similar (indeed, almost indistinguishable) from the results that we report. For example, the assumption on deductions does not affect our results, given the lack of information on other sources o f income (e.g., capital gains, mortgage payments, and dividends); similarly, the assumption on indexing also does not affect our results (especially in recent years) given the high correlation between the indices. The assumption on who claims the children (and other tax preferences) has a more significant impact. Allocating children to the mother/wife generally yields a lower average marriage tax than allocating them to the higher earner, as demonstrated by Aim and Whittington (1996) and Eissa and Hoynes (2000). Although formal and reliable statistics on child custody in the United States are hard to find and vary from source to source, it has been suggested that the number of fathers obtaining custody has been increasing over time. Since our objective is to calculate the cost of marriage, and not the cost o f divorce, and also to compare our results with previous analyses, we largely follow Rosen (1987), Feenberg and Rosen (1995), and Eissa and Floynes (2000) by calculating the marriage tax/subsidy under the assumption that couples minimize their tax liabilities when separated.

11 We have also made estimates based on the median marriage tax, rather than the average. Although some o f the magnitudes are slightly different, the patterns observed are very much the same. All results are available upon request.

W h ith e r th e M arriage Tax? 259

W ei

gh te

d E

st im

at ed

E

st im

at ed

N o.

o f

W ei

gh te

d N

et R

ev en

ue

Pe rc

en t

W ei

gh te

d R

ev en

ue

P er

ce nt

W

ei gh

te d

R ev

en ue

H ou

se ho

ld s

A ve

ra ge

S

ta nd

ar d

G ai

n/ L

os s

o f T

ot al

A

ve ra

ge

S ta

nd ar

d L

os s

o f T

ot al

A

ve ra

ge

S ta

nd ar

d G

ai n

260 N a tio n a l T a x J o u r n a l

* . o .to

vO NO OO vO r " - on no vo t o m ■O C-l <N ( N <N

vO vO r--

O' O' ON ON

vo no r-

o n ( N i n ( N ro cn roNO VO NO VO

<N <N <N <N <N (N

NO NO NO NO

00 00 00 ON ON ON

— <N

ON Ov <N (N

(N <N (N (N

NO — NO —

On On On On On OnOn O' On On On On

W hither the Marriage Tax? 261

262 N a tio n a l T a x J o u r n a l

faced a marriage tax of as much as $617 in 1999 and 2000. Since 2003 and the passage o f the Economic Growth and Tax Relief Reconciliation Act o f 2001 (EGTRRA) and the Jobs and Growth Tax Relief Reconciliation Act of 2003 (JGTRRA), couples have on average benefited from filing jointly, as the average marriage subsidy was $423 in 2002, increasing to $671 in 2009.

However, there is significant variation across married couples in any given year, with some couples receiving a bonus and others facing a tax. Table 1 provides the percentage of couples that has faced a marriage bonus and tax throughout the years; additionally, we provide for each year the average tax for those paying a tax and the average subsidy for those receiving a subsidy, along with the corresponding standard deviation. The percentage of married couples benefiting from a marriage subsidy generally decreased from 1970 (64.8 percent) throughout the next three decades, falling to roughly one- third of all households by the year 2000 and then increasing after the 2001 and 2003 tax cuts to nearly one-half o f all households by 2009. Between 2003 and 2009, this benefit oscillated around $2,100. On the other hand, the size o f the average marriage tax has fluctuated around $1,500 per couple since the early 1980s, with the percentage of couples facing a tax rising in the 1990s and then sharply declining after 2002 to under 40 percent. The total percent of couples with a marriage tax or subsidy does not equal 100, since some couples have experienced marriage neutrality.

Table 1 also shows the overall tax revenue consequences o f the estimated weighted averages. In 1969, when the number o f couples receiving a marriage subsidy was more than double the number o f couples being taxed, marriage non-neutralities implied approximately $35.6 billion in forgone revenue (in 2010 dollars). Although the non­ neutralities generated net revenue gains in the late 1990s and early 2000s (or $28.6 billion in 1997 and $29.8 billion in 2000), by 2009 the net revenue impact was $32.6 billion o f total forgone revenue. The additional gains in tax revenue have come at the expense o f marriage penalties for many couples.

Figure 2 shows the weighted average marriage tax, differentiating between one-earner and two-earner couples. The vertical lines highlight the major tax reforms during the sample period. As expected, one-earner couples have benefited from filing jointly, while two-earner families have typically faced a marriage tax, even though the Bush tax cuts o f 2001 and 2003 have decreased their average penalty to nearly zero.

Our results are generally comparable to those of earlier studies, at least for the periods in which our study overlaps the earlier studies. Although any estimated magnitudes depend on the various assumptions o f the analysis (e.g., the sample, the period of coverage, the year of indexation, and allowed behavioral changes), our estimates are comparable to those found in the literature for shorter time periods. For instance, we find a similar time trend to that observed by Aim and Whittington (1996) who used PSID data. Their estimates indicated that, as expected, married couples experienced an average marriage subsidy prior to the 1969 tax changes. Then, with the exception of 1977 and 1978, they found that on average married couples were subject to mar­ riage penalties throughout the 1980s and the early 1990s. Their estimated marriage tax when children are assumed to be claimed by the mother follows the same trend as our

W h i t h e r t h e M a r r ia g e Tax? 263

Figure 2

Average Marriage Tax/Subsidy, 1969-2009

Year

----------------- A v e ra g e ........................ O n e - E a r n e r F a m ily ----------------- T w o - E a r n e r F a m ily

N o te : C a lc u la t io n s a r e f o r m a r r ie d i n d iv id u a ls f o r e v e r y y e a r f r o m t h e M a r c h C u r r e n t P o p u la t io n S u rv e y .

results, although marriage taxes are not observed until the late 1980s. Our estimations follow similar trends and fall somewhere between their two estimates, one o f which assumes that the mother claims the dependent children and the other o f which assigns the dependent children to the higher earner.

Similarly, Eissa and Hoynes (2000) found that married households faced on average a marriage subsidy throughout the 1980s and 1990s. As with our calculations, they also found that not all couples are subject to a subsidy. For example, in 1997 (the last year o f their data), they estimated that 55 percent of married couples faced a marriage tax ($1,300 in 1997 dollars) and that 34 percent faced a marriage subsidy ($2,200 in 1997 dollars). As shown in Table 1, our estimates indicate very similar percentages of couples facing marriage taxes and subsidies. Our estimated levels o f the marriage tax/ subsidy are also quite similar when the estimates are adjusted to a similar indexed year.

Turning to the results of Feenberg and Rosen (1995), their analysis focused on the changes to the marriage tax after the Budget Reconciliation Act of 1993, and not on time trends. Although they used a different dataset, their estimates for 1994 also found an overall average marriage tax. Using 1994 dollars, they found an average overall

264 N ational Tax Journal

marriage tax o f $124; adjusted to 2010 dollars, their average overall marriage tax is $183. Our estimates use 2010 dollars, and suggest a higher average overall marriage tax of $478. When the estimates are disaggregated for those couples with a marriage tax and those with a marriage subsidy, our estimates become somewhat more comparable to theirs. Feenberg and Rosen (1995) estimated that 52 percent of couples face a mar­ riage tax in 1994, while 38 percent face a marriage subsidy; our comparable estimates are 63.0 percent and 34.1 percent. They estimated an average tax o f $1,244 (for those with a tax) and an average subsidy o f $1,399 (for those with a subsidy); after adjust­ ing to 2010 dollars, their marriage tax is $1,830, and their subsidy is $2,050. These estimates are slightly higher than our estimates o f $1,578 and $1,514 for the tax and subsidy, respectively.

B. Distributional and Compositional Effects12

The marriage tax/subsidy clearly burdens some people more than others, as shown especially in the analyses by Aim and Whittington (1996) and Eissa and Hoynes (2000). Our earlier discussion indicated that couples with one earner generally benefit from marriage, while couples with two earners are often penalized.

Similarly, other characteristics influence who benefits and who loses. One factor is household income. We divide each sample year by income quartile based on the fam­ ily’s total income, and we then track the marriage tax/subsidy for each group through time. Figure 3 displays the average marriage tax/subsidy by income group (or quartile). Also, Table 2A shows the changes in the marriage tax by income quartile broken down by sub-periods that highlight the major income tax reforms; for now we focus on the column denoted “Actual,” which shows the actual changes in the average marriage tax by income quartile broken down by sub-periods.13 Averaged across families in each quartile, we estimate an average marriage penalty for those in the highest quartile for much o f the period. Although this does not mean that all families in this income group faced a penalty throughout the period analyzed, we estimate that during the years in which the penalty is more evident approximately 50 to 70 percent o f families in this group experienced a marriage tax. In addition, any decrease in the average marriage tax was most prominent for those couples with incomes above the median. In fact, while the average marriage tax has in recent years become an average marriage subsidy for most groups, the Bush tax cuts were not sufficient to eliminate the marriage tax for many families in the lowest quartile, yielding an average marriage penalty for this group during the last years o f the sample period. This last result is mainly driven by

12 We are grateful to several anonymous referees for suggesting these distributional analyses. 13 Quartiles are estimated using each year’s income distribution, and then adjusted for inflation so that the

ranges of income are not constant throughout the years in the sample period. For example, the second quartile ranges between $39,482 and $59,416 in 1969, and between $39,997 and $71,151 in 2009. Simi­ larly, the average income in the third quartile in 1969 is $81,875 and $111,809 in 2009, and the maximum income per couple in 1969 is $713,000 and $1.1 million in 2009.

W h i t h e r t h e M a r r i a g e T a x ? 2 6 5

Figure 3

Average Marriage Tax/Subsidy by Income Quartiles

----------------- L o w e s t Q u a rtile -------------------- S e c o n d Q u a rtile

----------------- T h ird Q u a rtile -----------— - H ig h e s t Q u a rtile

N o t e : C a lc u la t io n s a r e f o r m a r r ie d in d i v i d u a l s f o r e v e r y y e a r f r o m t h e M a r c h C u r r e n t P o p u la t io n S u rv e y .

the increase in the percentage o f two-earner couples in this income group. Indeed, the percentage o f families experiencing a marriage penalty in this group increased from 15 percent in the early to mid-1970s to 47 percent in the mid-2000s.

The presence o f dependent children affects the amount o f taxes paid by a parent. Tax deductions are available to individuals and couples with one or more children, and these may vary depending on one’s marital status, since single individuals are allowed to file as heads o f households. As such, children can be a determinant o f the magnitude of the marriage tax/subsidy. Although not shown, we find that the marriage tax increased for most households in the 1980s and 1990s (except for couples without dependents), with this upward trend reversed in the 2000s.14 Also, the average marriage tax/subsidy is consistently lower for families with a higher number o f dependent children because the presence of more children implies more deductions. However, in contrast, those

For brevity, we do not report detailed graphs and tables about these trends. All results are available upon request.

266 N ational Tax Journal

Table 2A

Change in Average MarriageTax/Subsidy by Income Quartile (2010-$ )

A c tu al D e m o g ra p h ic s T ax L aw

O verall

(1) 1972-2009 -$ 6 1 1 .3 4 $684.12 -$ 1 ,2 9 5 .4 6

(0.0099) (0.0106) (0.0101)

(2) 1997-2004 -$ 1,041.35 $385.21 -$ 1 ,4 2 6 .5 6 (0.0100) (0.0098) (0.0098)

(3) 1997-2009 -$ 1,283.65 -$ 1 3 0 .0 3 -$ 1 ,1 5 3 .6 3 (0.0102) (0.0103) (0.0101)

Income Q uartiles I s' Quartile (Given Year)

(1) 1972-2009 $128.68 $273.81 -$ 1 4 5 .1 3 (0.0119) (0.0151) (0.0126)

(2) 1997-2004 -$ 3 7 1 .2 0 $3.70 -$ 3 7 4 .8 9 (0.0141) (0.0140) (0.0142)

(3) 1997-2009 -$ 6 3 3 .3 3 $55.30 -$ 6 8 8 .6 3 (0.0144) (0.0147) (0.0147)

2'“‘ Quartile (Given Year)

(1) 1972-2009 -$ 3 5 4 .2 0 $365.31 -$719.51

(0.0140) (0.0157) (0.0137)

(2) 1997-2004 -$ 5 3 6 .6 6 $101.48 -$ 6 3 8 .1 4

(0.0150) (0.0147) (0.0148)

(3) 1997-2009 -$ 645.05 -$ 1 9 6 .9 4 -$ 4 4 8 .1 2 (0.0155) (0.0161) (0.0157)

3"1 Quartile (Given Year)

(1) 1972-2009 -$1,339.01 -$ 1 .0 8 -$ 1 ,3 3 7 .9 2 (0.0169) (0.0189) (0.0162)

(2) 1997-2004 -$ 1 ,1 4 8 .1 2 $495.27 -$ 1 ,6 4 3 .3 9 (0.0180) (0.0173) (0.0174)

(3) 1997-2009 -$ 1 ,4 8 6 .4 8 -$ 2 4 7 .2 3 -$ 1 ,2 3 9 .2 4 (0.0188) (0.0192) (0.0185)

4"1 Quartile (Given Year)

(1) 1972-2009 -$ 8 8 8 .8 4 $2,093.86 -$2,982.71 (0.0262) (0.0262) (0.0258)

(2) 1997-2004 -$2,134.31 $924.08 -$ 3 ,0 5 8 .3 8

(0.0254) (0.0244) (0.0244)

(3) 1997-2009 -$2,384.01 -$ 1 3 9 .5 2 -$ 2 ,2 4 4 .4 9 (0.0259) (0.0259) (0.0253)

Note: Authors’ calculations using the March Current Population Survey (CPS) and tax estimations from NBER’s TAXS1M. Heteroscedastic standard errors are in parentheses.

W h ith e r th e M arriageTax? 267

families without any dependents have consistently received a marriage subsidy, perhaps because many couples without dependents are either one-earner couples or two-earner couples with very different income levels. Looking at family composition through time, increases in the percentage o f families without dependent children may have helped offset increases in the marriage tax over time. Although the share o f couples with three or more dependent children decreased early in our sample period, the total percentage remained relatively low after 1980, with little influence on the overall average trend.

It is also of interest to look at how the marriage tax varies across other demographic characteristics. Following Eissa and Hoynes (2000), we calculate the marriage tax by comparing the incomes of the first and second earner (Figure 4 and Table 2B). As expected, as the two incomes become closer, the couples are more likely to face a mar­ riage tax. This trend is persistent through time, despite the many changes to the tax code over the years.

Although race and age are not factors considered in the filing o f taxes, these char­ acteristics are often correlated with income. Additional (unreported) estimations show that on average “Black” households have consistently faced a higher marriage tax than “White” households; “Other” minorities also have generally experienced a higher

Figure 4

Average Marriage Tax/Subsidy by Earnings Ratio o f Secondary Earner (Relative to Head of Household)

Year

------------ 0< and < 25% ------------- 25% <= and < 50% ............... 50% <= and < 75%

— - 75% <= and < 1 0 0 % ------------- 100% +

Note: C alculations are fo r m arried in d iv id u a ls fo r every year from th e March C urrent Population Survey.

268 N ational Tax Journal

Table 2B

Change in Average M arriage Tax/Subsidy by Earnings Ratio o f Second Earner Relative to Head o f Households (2010-$)

A c tu al D e m o g ra p h ic s T ax L aw

G reater than 0% a nd Less than 25%

(1) 1972-2009 -$ 1 ,4 6 3 .8 3 -$ 2 2 4 .7 5 -$ 1 ,2 3 9 .0 8

(0.0215) (0.0226) (0.0193)

(2) 1997-2004 -$ 1 ,2 2 5 .9 0 $767.09 -$ 1 ,9 9 3 .0 0

(0.0236) (0.0233) (0.0230)

(3) 1997-2009 -$ 1 ,6 6 3 .1 4 -$305.51 -$ 1 ,3 5 7 .6 2

(0.0239) (0.0240) (0.0234) B etw een 25% a n d L e ss than 50%

(1) 1972-2009 -$ 1 ,2 0 6 .1 4 $186.54 -$ 1 ,3 9 2 .6 8 (0.0209) (0.0196) (0.0206)

(2) 1997-2004 -$ 1 ,0 2 3 .7 9 $708.59 -$ 1 ,7 3 2 .3 8

(0.0212) (0.0190) (0.0202) (3) 1997-2009 -$ 1 ,1 2 1 .1 8 $48.15 -$ 1 ,1 6 9 .3 2

(0.0211) (0.0197) (0.0209)

B etw een 50% a nd L ess than 75%

(1) 1972-2009 -$ 7 6 5 .0 8 $378.78 -$ 1 ,1 4 3 .8 6

(0.0240) (0.0215) (0.0248)

(2) 1997-2004 -$ 1 ,0 1 7 .1 7 $646.46 -$ 1 ,6 6 3 .6 3

(0.0194) (0.0169) (0.0174)

(3) 1997-2009 -$ 9 4 7 .0 9 $138.43 -$ 1 ,0 8 5 .5 2 (0.0202) (0.0191) (0.0187)

B etw een 75% a n d Less than 100%

(1) 1972-2009 -$ 6 4 1 .6 6 $351.25 -$992.91

(0.0280) (0.0250) (0.0304)

(2) 1997-2004 -$ 1 ,1 4 1 .2 9 $571.68 -$ 1 ,7 1 2 .9 7

(0.0233) (0.0207) (0.0220) (3) 1997-2009 -$ 1 ,1 6 0 .9 5 -$ 2 .2 3 -$1,158.71

(0.0232) (0.0220) (0.0233) 100% or M ore

(1) 1972-2009 -$ 5 8 1 .7 7 $531.53 -$ 1,113.30

(0.0922) (0.0745) (0.1026)

(2) 1997-2004 -$ 7 2 6 .8 6 $615.30 -$ 1 ,3 4 2 .1 6

(0.0559) (0.0515) (0.0537)

(3) 1997-2009 -$ 6 7 6 .1 9 $293.83 -$97 0 .0 1 (0.0586) (0.0575) (0.0568)

Note: Authors’ calculations using the March Current Population Survey (CPS) and tax estimations from NBER’s TAXSIM. Heteroscedastic standard errors are in parentheses.

W h i t h e r t h e M a r r i a g e T a x ? 2 6 9

marriage tax than whites.15 As noted above, this result is partly because the average incomes o f black and other households are lower than whites and also because lower income households are more likely to have both spouses working and earning comparable incomes. As for age, younger couples have consistently faced a marriage tax. Younger couples are likely to have both members working, and two-earner couples (with similar incomes) typically pay a marriage tax.

C. D e c o m p o s in g t h e C a u s e s o f C h a n g e s in t h e M a r r i a g e T a x /S u b s id y

Changes to the marriage tax are due either to changes in the tax code or to changes in family structure. In order to capture the influence o f changes in family composition and income, we calculate the tax liabilities o f each couple using the 1972 tax code and the 2009 tax code.16 Figure 5 shows how the marriage tax faced by married families

F ig u r e 5

A ve ra g e M a rria g e T a x/S u b sid y u sin g 1972 Tax C ode fo r All Years, 1 9 6 9 -2 0 0 9

O O

S? ^ o o CNĴ >* o 2 o _Q - =J ^ C/)

CD \ -

(D CDTO 0 _

*

s ' «*■ —*

s ' ~ ~

----- S .

t ° CD

CD CD

2 q3 o > o

i

1970 1980 1990 2000 2010

Year

----- ------ - Average O ne-Earner Family --------------- Tw o-Earner Family

Note: C alculations are fo r m arried indiv id u a ls fo r every year fro m th e M arch C urrent P opulation Survey.

15 Again, these results are not reported. All results are available upon request. 16 In 1969, prior to the full implementation o f the changes to the federal tax code, the average couple enjoyed

a marriage tax/subsidy similar to that in 2009. In order to capture changes to the marriage tax, we allow for all 1969 changes in the tax code to take full effect, and we start this part o f the analysis in 1972 instead o f in 1969.

270 N ational Tax Journal

would have changed over time had the tax code remained unchanged over this entire period, using the 1972 tax code.17 Increases in family (real) income, coupled with decreases in the number of children per couple, have partially increased the magnitude o f the marriage tax (or reduced the marriage subsidy) for two-income families, as well as increased the magnitude o f marriage subsidies for one-earner couples. The results indicate that changes in demographics have clearly influenced the changes in the mag­ nitudes o f marriage taxes and subsidies.

To isolate the influence of changes in the tax code rather than changes in family composition and income, we estimate the marriage tax for every year using individu­ als from the 1973 CPS March data.18 Since each year’s tax brackets may be adjusted statutorily for increases and/or decreases in income, we adjust each family’s incomes in 1972 to account for inflation. We then calculate the marriage tax/subsidy for each family and inflate it to 2010 dollars. This procedure allows us to disentangle the influence of changes in tax laws versus changes in demographic characteristics.

Figure 6 shows the average marriage tax/subsidy faced by each family type through time, holding family composition and income constant with the 1972 CPS. As expected, one-earner families have experienced tax benefits from filing as married individuals instead o f singles. The marriage tax/subsidy for one-earner couples was the lowest in the early 1970s, following the 1969 changes; it increased between the mid-1970s and mid-1980s, it decreased again through the 1990s, and the tax cuts implemented in 2001 and 2003 increased the marriage subsidy to levels above those observed in the 1980s.19

As Figure 6 shows, the increases in the tax benefits for one-earner couples during the 1970s and 1980s contrast with the increases in the tax penalties faced by two-earner couples. The trends in the figure indicate that the most significant changes in the mar­ riage tax/subsidy stemming from changes in the tax codes for two-earner couples began with the 1969 reform, then continued through the early 1980s, and appeared again in the early 2000s. If individuals in two-earner couples had kept their 1972 income levels (adjusted for inflation), then their marriage taxes would have nearly halved in the mid- 1980s. Aggregating all married couples in the 1972 sample, changes in the 1970s and 1980s suggest an overall average marriage subsidy. If two-earner couples had enjoyed wages similar to those in 2009 (adjusted for inflation) over the last 40 years, they would have paid a marriage tax that peaked in 1981, remained somewhat steady through the mid-1990s, and then rose slowly until the Bush tax cuts resulted in average marriage subsidies. With average incomes now being largely stagnant, these couples may well start paying a marriage tax again, especially if tax rates increase. Figure 6 demonstrates that

17 We conducted a similar analysis using the 2009 tax code. These results display similar trends. For example, in cases for which two-earner couples experience a marriage subsidy, as is the case for many couples in 2009, increases in income decreased the potential gain from being able to file as a married couple. All results are available upon request.

18 Again, recall that income and employment data for these two CPS samples refer to the preceding year. 19 Once again, we repeated the exercise using the 2010 March CPS. Although we observe some differences in

magnitudes due to differences in average real incomes between 1972 and 2009, the results are qualitatively the same. All results are available upon request.

W h ith e r th e M a rria g e T a x ? 271

Figure 6

A v e ra g e M a r ria g e T a x /S u b s id y u s in g t h e 1 9 7 2 CPS fo r A ll Tax Years, 1 9 6 9 - 2 0 0 9

1970 1980 1990 2000 2010 Y ear

-------------- Average .................... O ne-Earner Family -------------- Tw o-Earner Family

Note: C alculations are fo r m arried in d iv id u a ls fo r every year fro m th e March C urrent Population Survey.

some o f the most notable changes to the marriage tax/subsidy are preceded by changes to the tax law (vertical lines). This is opposite to the patterns tracking the changes due to demographic changes, where increases and decreases tend to be smoother and no stark changes are observed.

In order to better quantify the effects o f changes in demographics versus changes in taxes, we follow Eissa and Hoynes (2000) and decompose the changes in the marriage tax/subsidy after each major tax reform, including the 2001 and 2003 tax cuts. As they show, the increase or decrease in the marriage tax/subsidy between two given years (t and t - 1) is given by:

(1) E{T\Xt ,L ,) - E { T \ X t_v Lt_x)

where£'(7’|2f, Y() is the mean tax or subsidy given the tax law (L ) and the demographic characteristics (X{) at time t. The first term in parentheses on the right-hand side repre­ sents the difference in the mean tax due to demographic changes (i.e., holding the tax

272 N a tio n a l T a x J o u r n a l

law constant), and the second term represents the mean tax difference due to tax changes (i.e., holding the demographic characteristics constant). We repeat this exercise several times for several groups (e.g., those grouped by income, children, earnings ratio, race, and age as in the discussion above), using the major tax reforms (1969, 1972, 1982, 1987, 1994, and 2004) as breaking points. We focus on the period not covered in Eissa and Hoynes (2000), highlighting the Bush Tax Cuts and thus the 2004 breaking point.

Tables 2A and 2B show the results for all income and earnings ratios group break­ downs.20 The “Demographics” column shows the impact on the average marriage tax/ subsidy holding the tax law constant, the “Tax Law” column shows the impact holding the demographic characteristics constant, and the “Actual” column shows the combined impact o f demographics and taxes. In order to capture the change over the entire sample period, one would look at changes between 1969 and 2009. However, we emphasize the analysis using the 1972-2009 period given the phasing in o f the Tax Reform Act o f 1969.

Consider the decomposition by income and by income quartile (Table 2A). For all income groups, there has been a decrease over time in the average marriage tax/subsidy, as demonstrated by the change o f —$611 between 1972 and 2009 (Table 2 A, top panel, row 1). This decrease over time is largely due to changes in the tax law (-$1,296), and not to demographic changes ($684), particularly the tax cuts during the early years of the Bush administration. Even so, changes to the composition and income of the fam­ ily, including the increase in two-earner families, the increase in real income, and the decline in the number of children, have reduced the overall gains by about one-half.

Rows 2 and 3 o f the top panel of Table 2 A further decompose the changes that occurred after the major tax changes o f 2001 and 2003 for all income groups. For example, row 2 captures the changes to the marriage tax/subsidy before and after the tax cuts. Although the legislative change alone decreased the marriage tax by - $ 1,426 (“Tax Law”), changes in “Demographics” offset this impact ($385), so that the total impact was to increase the marriage tax by -$1,041 between 1997 and 2004. The other panels in Table 2A repeat this decomposition by income quartile. Tax law changes have always worked to decrease the average marriage tax, while demographic changes have worked in the opposite direction, until the most recent 1997-2009 period.

The remaining panels o f Table 2 A repeat this decomposition by income quartile. Tax law changes have largely dominated demographic changes for the entire 1972-2009 period for all income quartiles other than the 1st quartile, where demographic changes have increased the average marriage tax ($274) and tax changes have lowered the mar­ riage tax (-$145). Overall, tax law changes have tended to decrease the marriage tax/ subsidy. The impacts by specific sub-period vary considerably.

In a similar manner, Table 2B makes these decompositions by the earnings ratio. Although specific patterns vary considerably across demographic and income groups, demographic changes have primarily increased the average marriage tax, while tax changes have primarily decreased the tax. For example, in estimates for whites not shown here, we find that tax law changes over the 1972-2009 period have reduced the

20 Breakdowns with respect to race, number of children, and age are available upon request.

W h i t h e r t h e M a r r ia g e Tax? 273

marriage tax by -$1,350, demographic changes have increased the marriage tax by $732, and the combined impact is a change o f - $ 6 18 (top panel, row 2). The comparable numbers for blacks (others) are -$618, $37, and -$581 (-$1,286, $375, and -$911). Again, tax changes have tended to dominate demographic changes, with the result that the average marriage tax/subsidy has declined over the entire period. Again, the impacts by specific sub-period are much more variable.

For other groups (children, earnings ratio, age), the negative effects o f tax changes also largely outweigh the positive impacts o f demographic changes for the entire 1972-2009 period, although sub-periods often exhibit different patterns.21

V. CONCLUSIONS

Our results show that both changes in the federal tax code and changes in the family composition in the United States have resulted in changes in the way some couples gain and lose when having to file taxes jointly. Consistent with the results obtained by Aim and Whittington (1996) and by Eissa and Hoynes (2000), both of whom used a much shorter sample period, we find that changes in the tax code account for the great part o f the changes in the marriage tax/subsidy. Judging from the results obtained using the last year o f data (2009), and the fact that the Bush tax cuts were recently extended, we estimate that married couples on average benefit from the current system, even though some couples still lose from filing together.

Indeed, our estimates suggest that the federal government is losing significant amounts o f revenues from the current practice o f taxing the family, relative to taxing on an indi­ vidual basis independent o f marital status. As indicated in Table 1, in 2009 the federal government lost an estimated $32.6 billion in revenues from the existing marriage tax/ subsidy.22 A system where the individual is the unit o f taxation (under which everyone files as a “single”) could generate significant amounts of additional revenues.

There is also evidence that these current practices treat couples and singles (and other households) very differently, punishing some couples and singles while subsi-

21 We are also able to examine regional patterns. Differences in income and household specialization patterns across states can generate substantial regional differences in the relative tax burden o f married relative to unmarried couples. For example, in 2009 states with higher family incomes have smaller marriage subsidies for dual-income families, but larger subsidies for single-income families. States like California and New York have some o f the smallest subsidies for two-earner couples, and some o f the highest for one-earner couples. In contrast, in 1977, or one o f the years with high marriage penalties, dual-earner families in these states suffered contrastingly high penalties with marriage, compared to those in poor states like North Dakota and South Dakota. As expected, one-earner families in richer states enjoy a relatively higher benefit from being able to file as married, compared to those in relatively poor states. For brevity, these results are not shown, but are available upon request.

22 Other scenarios can generate even larger impacts. For example, suppose that the government returned to the pre-1948 system in which the individual becomes the unit o f taxation (and in which the head o f household filing status is also eliminated). Using the 2012 CPS for married and unmarried taxpayers and the NBER TAXS1M software and applying tax rates in the American Taxpayer Relief Act o f 2012, we estimate that the individual income tax system would generate about $140 billion in additional annual tax revenues.

274 N a tio n a l T ax J o u r n a l

dizing others. Using the 2012 CPS and the NBER TAXSIM tax calculating software, we compare how equitably individuals are treated by the current federal tax system by calculating the percent of annual wages taxed by the federal government for individuals with different income levels. Table 3 shows tax liabilities as a percent o f annual wages for individuals who qualify as either “single,” “head of household,” or “married.” For singles, heads of households, and married individuals with a single earner, we simply divide the total tax liability found by TAXSIM by total annual earned income; for mar­ ried couples with two earners, we assume that the total tax liabilities are split between the two individuals. As such, the percent of taxed wages for the individual is calculated as half o f the ratio of the total liabilities to combined income. This procedure allows us to compare the percent of annual wages that is taxed for individuals with similar income but different filing status.

As shown in Table 3, the average share o f wages taxed by the federal government for individuals with incomes between $30,000 and $33,000 and filing as single is 9.33 percent (with a 5.62 percent standard deviation). In contrast, a similar individual fil­ ing as a head of household (e.g., due to having dependents) is only taxed 1.15 percent o f wages. While this burden is higher for a similar individual filing jointly, it is 3.4 percentage points lower than one filing as single. Similar differences remain among individuals with higher annual wages.

The fundamental issue here is: what do we as a society want the individual income tax to achieve in the choice o f the “unit” o f taxation (Bittker, 1975; Bakija and Steuerle, 1991; Steuerle, 1999; Kleven, Kreiner, and Saez, 2009)?23 There is today an enor­ mous, and increasing, diversity o f family structures in many countries. In the 1950s, the “traditional” family was typically a single-earner household with a stay-at-home spouse. Now, many individuals choose to live alone, two-earner families are the norm, non-marital cohabitation among opposite and same-sex couples is common, extended families are increasing in numbers, and there are widespread instances o f unrelated individuals living together. These newer types o f households are, by many definitions, “families.” However, they are treated very differently, and often much less favorably, than the traditional households once envisioned as the norm by the tax codes in many countries.

If concern with the marriage tax/subsidy is an overriding issue, it is certainly pos­ sible to move the individual income tax toward Marriage Neutrality. As noted above, one obvious method here is to make the individual the unit o f taxation, a change that would also increase tax collections. However, it is possible to move closer to a mar­ riage neutral tax system even while retaining the family as the unit o f taxation, by such piecemeal reform options as increasing the standard deduction for married couples, expanding the tax brackets facing married couples, re-introducing a secondary earner deduction, expanding the phase-out range o f transfer programs, flattening overall rate structures, or allowing optional individual filing. More fundamental reform options

23 See Bittker (1975), Bakija and Steuerle (1991), Steuerle (1999), and Kleven, Kreiner, and Saez (2009) for discussions and analyses o f the appropriate choice o f the unit o f taxation.

W h ith e r th e M arriage Tax? 275

276 N a tio n a l T a x J o u r n a l

include eliminating progressivity (for example, a flat rate tax) or even eliminating the individual income tax and replacing it with a national sales tax or a value-added tax.

Apart from the last option, these reforms would only reduce the marriage penalties (and bonuses) under the individual income tax, and there would still be marriage penalties and bonuses throughout other parts o f the tax and transfer system. Moreover, suppose the individual is made the unit o f taxation, as is the case in many OECD countries. This choice is not without its own efficiency and equity problems. An important justification for the use o f the family as the unit o f taxation is the notion that families with equal family income should pay equal taxes. There is no question that making the individual the unit of taxation would violate this goal o f Horizontal Equity Across Households, even with its obvious limitations. There are also significant administrative and compli­ ance issues under an income tax system based on the individual as the unit of taxation. How are itemized deductions split between partners? How is unearned (or capital) income split between partners? Who claims the tax benefits from children? How do the tax enforcement agencies verify the legitimacy of these declarations? What are the compliance costs o f individual filing? Many other issues arise, and the ways in which these issues are resolved vary greatly across countries.

It may well be that the importance of the traditional family unit still justifies its largely favorable tax treatment. This is clearly the avenue that the United States has chosen, and it seems unlikely that this choice will change any time soon. However, it may also be time to recognize that a diverse society can no longer treat one family structure so differently from the others. Making the individual the unit o f taxation would eliminate the marriage tax/subsidy (and the singles tax), and would also re-establish the principle of horizontal equity, broadly defined to apply to individuals and not to households. Many countries have in fact chosen to make the individual the unit of taxation.

There are no easy choices for taxing the family, and it is inevitable that the goals o f taxation are often conflicting. Taxing the family requires facing these difficult trade-offs directly and with knowledge o f how different ways o f taxation generate different results. We hope that our paper provides some useful information on these trade-offs.

ACKNOWLEDGMENTS

We are grateful to Wesley Howden, Rachel Donga, and Tamara Walker for the assistance in assembling and analyzing the data set, as part o f the Tulane Research Experience for Undergraduates in Applied Microeconomics and Program Evaluation (Treu-Ampe), funded by the National Science Foundation and supervised by Professors Keith Finlay and Jonathan Pritchett. Daniel Feenberg provided invaluable help with the NBER TAXSIM calculations. We also thank three anonymous referees, George Zodrow, and William Gentry for many helpful comments and suggestions.

W h ith e r th e M arriageTax? 277

DISCLOSURES

The authors have no financial arrangements that might give rise to conflicts o f interest with respect to the research reported in this paper.

REFERENCES

Aim, James, Stacy Dickert-Conlin, and Leslie A. Whittington, 1999. “Policy Watch: The Mar­ riage Penalty.” Journal o f Economic Perspectives 13 (3), 193-204.

Aim, James, and Mikhail I. Melnik, 2005. “Taxing the ‘Family’ in the Individual Income Tax — An International Perspective.” Public Finance and Management 5 (1), 67-109.

Aim, James, and Leslie A. Whittington, 1996. “The Rise and Fall and R ise...o f the Marriage Tax.” National Tax Journal 49 (4), 571-589.

Aim, James, and Leslie A. Whittington, 1999. “For Love or Money? The Impact o f Income Taxes on Marriage.” Economica 66 (4), 297-316.

Aim, James, Leslie A. Whittington, and Jason Fletcher, 2002. “Is There a ‘Singles Tax’? The Relative Income Tax Treatment o f Single Households.” Public Budgeting & Finance 22 (2), 69-86.

Bakija, Jon, and C. Eugene Steuerle, 1991. “Individual Income Taxation since 1948.” National Tax Journal 44 (4), 451-475.

Berliant, Marcus, and Paul Rothstein, 2003. “Possibility, Impossibility, and History in the Origins o f the Marriage Tax.” National Tax Journal 56 (2), 303-317.

Bittker, Boris I., 1975. “Federal Income Taxation and the Family.” Stanford Law Review 27 (4), 1388-1463.

Bull, Nicholas, Janet Holtzblatt, James R. Nunns, and Robert Rebelein, 1998. “Defining and Measuring Penalties and Bonuses.” OTA Working Paper 82. U.S. Department o f the Treasury, Office o f Tax Analysis,Washington, DC.

Burstein, Nancy R., 2007. “Economic Influences on Marriage and Divorce.” Journal o f Policy Analysis and Management 26 (2), 387-429.

Carasso, Adam, and C. Eugene Steuerle, 2005. “The Hefty Penalty on Marriage Facing Many Households with Children.” The Future o f Children 15 (2), 157-175.

Chade, Hector, and Gustavo Ventura, 2005. “Income Taxation and Marital Decisions.” Review o f Economic Dynamics 8 (3), 565-599.

Clarke, George R. G., and Robert P. Strauss, 1996. “Effects o f the Federal Individual Income Tax on Marriage and Divorce Decisions.” RCER Paper No. 43. University o f Rochester, Center for Economic Research, Rochester, NY.

278 N a tio n a l T ax J o u r n a l

Congressional Budget Office, 1997. For Better or For Worse: Marriage and the Federal Income Tax. Congressional Budget Office, Washington, DC.

Dickert-Conlin, Stacy, 1999. “Taxes and Transfers: Their Effects on the Decision to End a Mar­ riage.” Journal o f Public Economics 73 (2), 217-240.

Dickert-Conlin, Stacy, and Scott Houser, 2002. “ EITC and Marriage.” National Tax Journal 55 (1), 25-40.

Eissa, Nada, and Hilary Williamson Hoynes, 2000. “Explaining the Fall and Rise in the Tax Cost o f Marriage: The Effect o f Tax Laws and Demographic Trends, 1984-97.” National Tax Journal 53 (3, Part 2), 683-711.

Feenberg, Daniel R., and Harvey S. Rosen, 1995. “Recent Developments in the Marriage Tax.” National Tax Journal 4 8 (1 ), 91-101.

Feucht, Frederick J., Murphy Smith, and Robert Strawser, 2009. "The Negative Effect o f the Marriage Penalty Tax on American Society.” Academy o f Accounting and Financial Studies Journal 13 (1), 103-125.

Gillette, Robert, Janet Holtzblatt, and Emily Y. Lin, 2005. “ Marriage Penalties and Bonuses: A Longer Term Perspective.” Proceedings o f the Annual Conference on Taxation o f the National Tax Association, 468^478. National Tax Association, Washington, DC.

Gillette, Robert, Janet Holtzblatt, and Emily Y. Lin, 2006. “What Gives Rise to Changes in Mar­ riage Penalties and Bonuses? A Panel Model Approach.” Proceedings o f the Annual Conference on Taxation o f the National Tax Association, 452^463. National Tax Association, Washington, DC.

Holtzblatt, Janet, and Robert Rebelein, 2000. “ Measuring the Effect o f the EITC on Marriage Penalties and Bonuses.” National Tax Journal 53 (4), 1107-1133.

Kaplow, Louis, 1989. “Horizontal Equity: Measures in Search o f a Principle.” National Tax Journal 42 (2), 139-154.

Kaplow, Louis, 2008. The Theory o f Taxation and Public Economics. Princeton University Press, Princeton, NJ.

Kiefer. Donald W., 1984. "Distributional Tax Progressivity Indexes.” National Tax Journal 37 (4), 494-513.

Kleven, Henrik Jacobsen, Claus Thustrup Kreiner, and Emmanuel Saez, 2009. “The Optimal Income Taxation o f Couples.” Econometrica 77 (2), 537-560.

Laurin, Alexandre, and Jonathan Rhys Kesselman, 2011. “Income Splitting for Two-Parent Families: Who Gains, Who Doesn’t, and at What Cost?” C.D. Howe Commentary Working Paper No. 335. C.D. Howe Institute, Toronto, Canada.

Lin, Emily Y., and Patricia K. Tong, 2012. “Marriage and Taxes: What Can We Learn from Tax Returns Filed by Cohabiting Couples?” National Tax Journal 65 (4), 807-826.

W h ith e r th e M arriage Tax? 279

Musgrave, Richard A., 1959. Theory o f Pubic Finance: A Study in Public Economy. McGraw-Hill Book Company, New York, NY.

Musgrave, Richard A., 1990. “Horizontal Equity, Once More.” National Tax Journal 43 (2), 113-122.

Rosen, Harvey S., 1987. “The Marriage Tax Is Down But Not Out.” National Tax Journal 40 (4), 567-576.

Ruggles, Steven, J. Trent Alexander, Katie Genadek, Ronald Goeken, Matthew B. Schroeder, and Matthew Sobek, 2010. Integrated Public Use Microdata Series: Version 5.0. University o f Minnesota, Minneapolis, MN.

Samuelson, Paul A., 1947. Foundations o f Economic Analysis. Harvard University Press, Cam­ bridge, MA.

Steuerle, C. Eugene, 1999. “Valuing Marital Commitment: The Radical Restructuring o f our Tax and Transfer Systems.” Responsive Community 2 (9), 35—45.

Steuerle, C. Eugene, 2006. “The Widespread Prevalence o f Marriage Penalties.” Urban Institute/ Brookings Institution Tax Policy Center, Washington, DC.

U.S. General Accounting Office, 1996. Income Tax Treatment o f Married and Single Individuals (GAO/GGD-96-175). USGAO, Washington, DC, http://www.gao.gov/assets/230/223096.pdf.

Young, H. Peyton, 1994. Equity: In Theoiy and Practice. Princeton University Press, Princeton, NJ.

Copyright of National Tax Journal is the property of National Tax Association and its content may not be copied or emailed to multiple sites or posted to a listserv without the copyright holder's express written permission. However, users may print, download, or email articles for individual use.