Our analysis provides four main contributions. First, this paper documents maritalstatus misreporting as a novel source of tax noncompliance related to marriage penalties. Second, linked marriage record and tax data show how actual (not reported) marriage timing varies with marriage penalties, making these data better suited to analyzing the impacts of marriage penalties than prior studies that relied on self-reported marital status. Administrative tax data also provide more credible marriage penalty estimates than studies using survey data—where estimated EITCs can be understated by one-third (Meyer et al. 2022). Third, while state income taxes were ignored in prior studies, we consider total (federal and state) marriage penalties. State taxes reinforce our main result, widening the average marriage penalty gap between compliant and misreporting newlyweds from about $1,000 to $1,200. Finally, we find cross-sectional and panel-based evidence of strategic misreporting. Couples who misreport and then become compliant, tend to do so immediately upon no longer facing a marriage penalty. That is, strategic behavioral responses are seen not only in the year of marriage—where newlywed misreporting rates increase with marriage penalties—but also in later years, when misreporting couples start correctly reporting their marriage.

This study has some limitations. First, we do not observe divorces, hence our analysis focuses on marriage penalties that discourage reporting new marriages. Second, our data-linking approach requires strict name matches for both spouses. This should avoid false matches but results in over half of newlyweds not being linked to the tax data—although these omissions should result from missing middle names and largely random differences in how names are spelled. A less restrictive approach capturing twice as many matches shows higher misreporting rates (likely from false matches) but a nearly identical pattern as Figure 1. Finally, the marriage records come from only one state: Minnesota. This could limit generalizability, but the effects of state income taxes are relatively modest.

II. Policy Background and Prior Literature

A marriage penalty occurs when a couple owes more income tax when filing a return as married, while a marriage bonus occurs when they owe less. A couple married on December 31 must file as married that tax year. While some unmarried couples file as married, in this study, we only consider marital-status misreporting as a married person filing as single or head-of-household instead of married filing jointly or married filing separately.²

Marriage penalties often result from a couple having similar incomes, especially if eligible for the EITC. Consider a couple with three EITC-eligible children in 2022 where each spouse earns $20,000. Filing jointly, the couple's federal EITC would be about $4,000. Had the couple misreported their marital status by filing two separate tax returns, each with a head-of-household filing status, the couple's EITCs would instead total about $10,000: over $6,000 from the spouse claiming two children and nearly $4,000 from the spouse claiming one child (see Figure 2B). This dual-earner couple would receive about $6,000 less in EITCs by filing jointly and, after accounting for the refundable child tax credit phase-in threshold, would face a marriage penalty of about $5,500. This penalty is nearly 15% of the couple's income.

Marriage bonuses, in contrast, often result from a couple having unequal incomes. Consider a couple in 2022 composed of one spouse with no income and the other spouse earning $100,000. Filing jointly, the couple has $74,100 of taxable income after deducting their standard deduction and owes federal income taxes of about $8,500. Had the married couple misreported by filing two single returns, the single-earner spouse would have $87,050 of taxable income after deducting the smaller single standard deduction and would owe about $14,800. The other spouse with no income would owe no income tax. This single-earner couple owes $6,300 less in tax by filing jointly and therefore has a marriage bonus of that amount.

The tax system is not marriage neutral—progressive taxation combined with joint filing creates inherent marriage penalties and bonuses. Marriage neutrality would require less tax progressivity (and no credit phase-outs) or eliminating joint filing that combines spousal incomes.³ Given these tradeoffs, the federal tax system has alternated between higher and lower marriage penalties and bonuses over time (Beebe 2019). Initially, only taxpayers in community property states—where assets acquired by either spouse during the marriage are considered jointly owned—could split income across two federal tax returns, resulting in marriage bonuses in those states because all filers were subject to only one set of tax brackets. In 1948, legislation extended these marriage bonuses to all federal taxpayers by creating a married-filing-jointly status with double-sized tax brackets (Joint Committee on Taxation 2001). Consistent with changes in tax policy, references in historical print sources to the "marriage penalty" increased in 1945–1947, 1975–1985, and 1991–2001 (see online appendix). After each increase, there were notable declines coinciding with reforms that mitigated marriage penalties: the 1948 universal incomesplitting legislation, the Tax Reform Act of 1986, and the 2001 and 2003 tax changes.⁴

A. Prior Literature on Marriage Penalties

1. Marriage penalty estimation issues. Strong assumptions are needed for standard marriage penalty estimates because of unobserved spousal splits for jointly reported items (Bull et al. 1999; Holtzblatt and Rebelein 2000). This results in uncertainty for marriage penalty estimates, especially when dividing dependents, capital income, and deductions between spouses. This issue, however, does not apply to our marriage penalty estimates for misreporters because we combine the returns of separately filing spouses instead of dividing the return of jointly filing spouses.

2. Timing of marriage. Marriage penalties slightly delay the reported timing of certain marriages. Sjoquist and Walker (1995) showed this with time-series data, considering tax-sensitive marriage delays from the last quarter of one year to the first quarter of the next year. Alm and Whittington (1995) used panel data, finding an elasticity of short-term marriage delays with respect to marriage penalties of 0.78-1.54. Alm and Whittington (1997) found that marriage penalties increased the probability of delaying reported marriage timing by nearly 5%. Welfare eligibility rules may also affect the timing of marriage (Alm, Dickert-Conlin, and Whittington 1999). Teitler et al. (2009) estimated TANF participation delayed marriage by 12 to 16 months, but Isaac (2020) found little effect on marriage or divorce.

3. Marriage rates. While this study focuses on the timing of reporting marriages, the prior literature mostly examines the relationship between marriage penalties and marriage formation rates. Using average marriage penalties, Alm and Whittington (1997) found marriage penalties decreased marriage rates by about 2% in 1985. Alm and Whittington (1999) emphasized that marriage penalties more strongly affected women's marital decisions. Isaac and Jiang (2025) found small marriage formation responses to the marriage bonuses and penalties in the Affordable Care Act. Holtzblatt and Rebelein (2000), Lin and Tong (2012, 2014), and Maag and Acs (2015) considered cohabiting couples, whose decision to marry may be sensitive to marriage penalties. For a broader review, see Friedberg and Isaac (2024).

4. EITC. When spouses have more similar earnings, marriage penalties tend to worsen. Couples with children are especially affected because of EITC phase-outs (Congressional Budget Office 1997; Crandall-Hollick and Hughes 2018). Holtzblatt and Rebelein (2000) found that EITCs increased marriage penalties by nearly one-tenth, mostly from phase-outs. Although the EITC is usually associated with lower-income groups, the larger penalties in the phase-out range imply EITC marriage penalties mostly affect middleincome couples. Dickert-Conlin and Houser (2002) and Isaac (2020) found few EITC effects on marriage, but the latter study found that a more generous EITC encourages lower-earning women to divorce. We find that EITC-claiming newlyweds are more likely to misreport their marital status. This is consistent with interviews of EITC recipients, who said marriage could "mess up" their tax refund; however, rather than forgoing marriage, recipients admitted they (and their tax preparers) sometimes misreported marital status to increase tax refunds (Edin, Tach, and Halpern-Meekin 2014).

5. Race. The sensitivity of marriage penalties to the similarity of spousal earnings may create systematic differences in marriage penalties by race. Due to more similar earnings among black spouses, marriage penalties tend to be worse for this group, at least in studies using survey data (Brown 2021; Alm, Leguizamon, and Leguizamon 2023; Holtzblatt et al. 2024). Using tax data, however, Costello et al. (2024) found that nearly all income groups below $200,000 had lower marriage penalty rates for black joint filers than for white joint filers. They suggested this different result was from many more marriages being reported in survey data than tax data.

6. Tax Noncompliance. IRS (2024) estimated filing status errors explain $7 billion of noncompliance and show noncompliance rates by type of income. Our average maritalstatus misreporting rate of about 2% suggests newlyweds misreport their marriage at rates similar to the noncompliance rate of wages, which have substantial information reporting. But for newlyweds subject to large marriage penalties, the marital-status misreporting rate can reach the 15% noncompliance rate for income subject to only some third-party information reporting (e.g., partnership income), although well below the 55% noncompliance rate for income with little or no information reporting (e.g., sole proprietor income). Marital-status misreporting rates among those with tax penalties are therefore usually below misreporting rates for other forms of noncompliance.

Misreported filing status was the third-largest reason for estimated EITC noncompliance in 2006–2008, accounting for up to $3.3 billion annually in tax credit overclaims (IRS 2014).⁵ This filing-status noncompliance was mostly from married taxpayers incorrectly filing as unmarried, as nearly one-tenth of EITC claimants with children likely misreported their marital status.⁶ We find that large credit overpayments from marital status noncompliance can persist for many years. The weak enforcement against this form of noncompliance occurs because the IRS does not observe up-to-date marital status, which is administered at the state and local levels, and therefore the IRS only corrects this noncompliance through resource-intensive audits. If misreporting noncompliance is sensitive to increased information reporting—as are other forms of noncompliance—then information reporting could mitigate marital-status misreporting. Additionally, if the IRS had up-to-date marriage-status data, it could use marital-status misreporting as part of its e-filing rejection criteria.⁷

B. Marriage Mismeasurement and Administrative Data Linkages

Our linked data also highlight broader implications of marriage mismeasurement. Survey data include many misreported marriages, but are the basis for estimates of the decline in marriage rates (Kearney 2025) and estimates of the causal impact of housing costs on marriage (Bowmaker and Emerson 2015; Chiocchio 2025) and marriage on fertility (Hayford, Guzzo, and Smock 2014; Smith 2019). While marriages are modestly underreported in tax data, marriages appear to be widely overreported in survey data. Linking individuals across the Current Population Survey and tax return data suggests a 10% higher net marriage rate in survey data.⁸

National estimates using administrative records only report annual flows of marriages and divorces. There are no parallel administrative estimates of the stock of overall marriages, which require tracking each individual's marriages, divorces, immigration, emigration, and death over time. Although marriage and divorce data are collected at the state and local level, these data are not aggregated and therefore multi-year tracking is generally not feasible. We address this data limitation by focusing on newlyweds and using administrative marriage records. Publicly available marriage records, however, do not include Social Security Numbers. We therefore develop a matching algorithm based on spousal names. This resembles the method used by Census to link tax and survey data using Protected Identification Keys and the Splinter et al. (forthcoming) matching of tax data to loan data by business name. However, we do not observe divorces and therefore limit our main estimates to newlyweds.

III. Data Construction and Marriage Penalty Calculation

This study uses administrative data from state marriage records and federal tax returns. The state marriage records cover almost all marriages in Minnesota. The tax return data are from IRS administrative data containing the population of federal tax returns and information returns (e.g., Forms W-2). This section describes how we link these data, create the sample to cover the 22-year period from 2001 to 2022, and calculate marriage penalties by comparing federal income tax liabilities of actual returns and returns with counterfactual filing statuses. The sample includes over two hundred thousand couples and enables analysis of the relationship between marital-status misreporting and marriage penalties.

A. Linking Marriage Records to Tax Data

Marriages conducted in Minnesota are publicly available on the Minnesota Official Marriage System (https://moms.mn.gov). Marriage records date back to the early 19th century and are continuously updated. Although these records are public, the data cannot be downloaded directly. We create a bot with web-scraping functionality and collect all marriage records between 2001 and 2022, resulting in about 600,000 marriage records.

Tax data include extensive demographic information, such as name, sex, taxpayer identification number, and mailing address. Marriage records only include the county of marriage, certificate number, date of marriage, and the full names of both spouses. This leaves limited options to link these data. The location of a marriage may not determine where a couple files their taxes, the certificate number is only for internal use, and maritalstatus misreporting on tax returns means the dates of marriage cannot be used for linking. Therefore, we rely solely on names to link marriage records to tax data.

In the marriage records, names appear in two fields: "LAST" and "FIRST MIDDLE". To prepare the second field for matching, we classify the first name as all characters before the first space, and the middle initial as the first character after the space. In the tax data, if the names of a fictional couple are "JANE H DOE" and "JOHN G DOE" then their names are usually recorded as "JANE H & JOHN G DOE" when filing jointly. For the first spouse, we classify the first name as all characters before the first space, the middle initial as the first character after the first space, and the last name as all characters after the last space. For the second spouse, we classify the first name as all characters between the "&" and second-to-last space, the middle initial as the first character after the second-to-last space, and the last name as all characters after the last space.

To link marriage records to tax returns, we use a strict-name match, employing both spouses' full names including middle initials. We require an exact match to the first name and middle initial of both spouses, along with an exact match to one last name, which allows for last-name changes upon marriage. Because we rely on both spouses' names, the couple must have filed a joint or married filing separately tax return at some point to be matched. For this purpose, we use the tax return (from tax years 2001 to 2024) that is the first year a couple filed with a married filing status (married filing jointly or married filing separately). This implies that we do not match a couple's marriage record to their tax returns if there is (a) any deviation in spelling of the first name or middle initial for either spouse, (b) any deviation in spelling of the last name on the joint return, (c) a middle name is not provided, (d) the last name contains a suffix, such as "JR" or "III", or (e) the couple has never filed as married. The strict-name approach matches 36% of newlyweds in 2001- 2022 Minnesota marriage records to their federal tax returns. Unmatched marriages are from spouses never filing a married return by 2024 or without middle names, as well as differences in names in the two datasets (e.g., misspellings, nicknames, suffixes, or data errors), which should be uncorrelated with marriage penalties. The online appendix shows that misreporting rates appear to be largely random with respect to the year and month of marriage (except for December marriages) and the county of marriage. The match identifies 214,443 couples.⁹

B. Sample Construction

After matching both spouses' names on marriage records to tax returns, we observe their taxpayer identification numbers (usually Social Security numbers) and follow them over time within the tax data. We primarily focus on the year of marriage to study newlywed marital-status misreporting. This allows us to compare couples who are reacting to a new required filing status, where some are compliant and others misreport. For each couple, the year of legal marriage is:

1. The first year they file taxes as married (compliant)
2. Before they first file taxes as married (delayed reporter)
3. After they first file taxes as married (early reporter)

Delayed reporters represent 2.3% of newlyweds, while early reporters represent 1.4%. Our estimates focus on the more common form of marital-status misreporting, that due to delayed marriage reporting. The online appendix discusses early reporters, who have larger-than-average marriage bonuses. To ensure comparability and accurate marriage penalty estimates, we drop early reporters and couples where neither spouse filed an individual income tax return. We do allow for couples where only one spouse filed a return and the other was a non-filer. Implementing these restrictions narrows our dataset to 208,602 couples. Table A1 shows misreporting and compliant newlyweds are similar across various characteristics when both spouses file, although misreporters tend to have more dependents and face marriage penalties. Misreporting newlyweds where one spouse is a non-filer are generally lower-income, single-earner couples facing (but not receiving) marriage bonuses, although missing non-filer income may bias estimates among this group.

C. Calculating Marriage Penalties

Calculating marriage penalties requires estimating counterfactual tax liabilities using a tax calculator and counterfactual returns.

1. Tax Calculator. We use the TAXSIM calculator from the National Bureau of Economic Research to estimate tax liabilities for both actual and counterfactual returns using about three dozen variables. The federal income tax liability amounts in the tax data are replaced with calculated tax liabilities to ensure a common baseline. The calculated federal income tax liabilities deviate from reported liabilities in the tax data by a median of $654 (mean of $2,135) and from reported EITCs by a median of $0 (mean of $107). These differences can result from tax calculator simplifications, especially for high-income returns.¹⁰

2. Counterfactual Returns. To estimate marriage penalties, the combined tax liability of a couple's two separate tax returns needs to be compared to the tax liability of a joint tax return, where one is a simulated counterfactual. Different approaches are required to create counterfactual returns based on whether a couple filed as unmarried (misreported) or filed as married (compliant). For misreporting married couples, a counterfactual joint return is constructed by summing items (income, itemized deductions, dependents, etc.) across the two observed returns. If one of the spouses was a non-filer, a counterfactual single return is first constructed for the non-filer and then a counterfactual joint return is constructed for the couple, where the non-filer's income items are retrieved from Form W-2 wages, Form 1099-MISC/NEC nonemployee compensation, and Form SSA-1099 Social Security and disability benefits.

For compliant spouses filing jointly, the one joint return needs to be split into two counterfactual separate unmarried returns. Whereas combining misreporting returns is straightforward, dividing the components of compliant returns is less so. We start by applying a data-informed approach to allocate certain income sources. Wages are allocated based on individual Form W-2 wages. For example, if the primary filer earns 80% of the couple's combined Form W-2 wages, then that share of the couple's wage amount reported on their tax return is allocated to the primary filer and the remaining to the secondary filer. Sole proprietorship income and farm income are allocated based on Schedules C and F, which are attributable to only one of the two filers, as this income is subject to individual self-employment taxes. Remaining taxable income sources, such as capital income and deductions, are allocated 60% to the primary filer and 40% to the secondary filer. While information returns could inform the capital income attributable to each spouse, the individuals listed for these information returns provide an imperfect measure of how much each spouse would claim had they filed individually. Finally, dependents are split evenly between spouses, with the oldest going to the primary filer. When there are an odd number of dependents, the primary filer is allocated the additional dependent. These assumptions likely result in an overly equal split of income, which decreases the average compliant marriage bonus and suggests that the average federal marriage penalty gap between compliant and misreporters could be larger than our estimate of $1,000.

IV. Results

First, we report misreporting rates by marriage penalty levels and then show the sources of marriage penalties and how they vary by the number of dependents. Next, we show the strong correlation between EITC marriage penalties and misreporting rates, where both are larger among spouses with similar earnings. Finally, we discuss how marriage penalties among compliant and misreporting filers have changed over time with policy changes.

Delayed reporting of marriages is consistent with strategic tax minimization (and credit maximization). The overall newlywed marital-status misreporting rate was 2.3%. But the misreporting rate was only 1.5% for couples with marriage bonuses and 3.3% for couples with marriage penalties.¹¹ Besides this average difference, misreporting rates increase as marriage penalties increase. Figure 1 shows the misreporting rate of 4.3% for couples with a marriage penalty of $4,000 increases to 8.2% for couples with a marriage penalty of $6,500 and then increases to 14% for couples with a marriage penalty of around $8,000.¹² When marriage penalties are measured as a percent of income, misreporting rates increase in a similar pattern. Compared to under 2% for couples with no marriage penalty, the misreporting rate doubles to 4% when penalties reach two percent of income and doubles again to 8% when penalties reach eight percent of income. See the online appendix for details, as well as evidence that the lowest-income group has the highest marital-status misreporting rate.¹³

Compliant newlyweds tend to have marriage bonuses, while misreporting newlyweds tend to have marriage penalties. Table 1 shows that compliant newlyweds received an average marriage bonus of $376, but misreporters faced an average marriage penalty of $599. This main analysis only considers federal income taxes, but adding state income taxes reinforces the main finding, increasing the average gap between compliant and misreporting newlyweds from $975 to $1,225.

Table 1: Marriage penalties and bonuses: Newlyweds & prior-year misreporters