Whose Child Is This? Shifting of Dependents Among EITC Claimants Within the Same Household

Finance and Economics Discussion Series Divisions of Research & Statistics and Monetary Affairs Federal Reserve Board, Washington, D.C. Whose Child Is This? Shifting of Dependents Among EITC Claimants Within the Same Household David Splinter, Jeff Larrimore, and Jacob Mortenson 2017-089 Please cite this paper as: Splinter, David, JeffLarrimore, andJacobMortenson(2017). “WhoseChildIsThis? Shifting of Dependents Among EITC Claimants Within the Same Household,” Finance and Economics Discussion Series 2017-089. Washington: Board of Governors of the Federal Reserve System, https://doi.org/10.17016/FEDS.2017.089. NOTE: Staff working papers in the Finance and Economics Discussion Series (FEDS) are preliminary materials circulated to stimulate discussion and critical comment. The analysis and conclusions set forth are those of the authors and do not indicate concurrence by other members of the research staff or the Board of Governors. References in publications to the Finance and Economics Discussion Series (other than acknowledgement) should be cleared with the author(s) to protect the tentative character of these papers.

WHOSE CHILD IS THIS? SHIFTING OF DEPENDENTS AMONG EITC CLAIMANTS WITHIN THE SAME HOUSEHOLD David Splinter, Jeff Larrimore, and Jacob Mortenson July 7, 2017 Using a panel of household level tax data, we estimate the degree to which dependents are “reassigned” between tax units within households, and how these reassignments affect combined tax liabilities. Reassigning dependents reduces combined tax liabilities on average, suggesting some household level coordination. Additionally, when EITC benefits expanded in 2009, reassignments increasingly involved adding a third child to tax returns to claim these new benefits. However, the subgroup reassigning towards three child tax units actually increased total household tax liabilities, suggesting that some taxpayers may prioritize minimizing their own tax burden or focus on particularly salient aspects of tax policy. Keywords: earned income tax credit, tax avoidance, household level tax coordination JEL Codes: D10, H24, H26, H31, H53 David Splinter: Joint Committee on Taxation, U.S. Congress, Washington, DC, USA (david.splinter@jct.gov) Jeff Larrimore: Federal Reserve Board, Washington, DC, USA (jeff.larrimore@frb.gov) Jacob Mortenson: Joint Committee on Taxation, U.S. Congress, Washington, DC, USA (jake.mortenson@jct.gov) 1

Households respond to income taxes along many dimensions. Beyond income or employment responses, one potential margin of response is to alter who is included in a family for tax purposes. In particular, households comprised of multiple tax units often have some discretion about how children are allocated across the tax units within the household.1 Given the size of tax benefits associated with claiming dependents and the non-uniformity in these benefits across tax units, reallocating dependents can result in substantial reductions in tax liability for the household as a whole. This paper investigates the extent to which children are reallocated from one tax unit to another within multiple tax unit households, and the degree to which these reallocations result in lower tax burdens. Understanding household responses along these lines is important for several reasons. First, estimates of behavioral responses to a policy are important when considering the full effects of potential policy changes, and this paper examines an underexplored dimension for such responses. The ability of households to reassign dependents represents an additional challenge to consider when targeting tax benefits and potentially can alter the distribution of benefits across these recipients. Additionally, beyond its value for evaluating responses to policy changes, the frequency of dependent reallocations can inform discussions on the way resources are shared within households. In part due to data limitations, researchers using tax return data often use tax units as proxies for the economic sharing units where consumption, income earning, and savings decisions are made. However, there remains uncertainty over how best to define the true sharing unit. This study suggests that economic sharing units can include multiple tax units, as we provide evidence of coordination across tax units. 1 A tax unit is defined as all individuals filing together on a tax return, including dependents. Individuals that do not appear on a tax return are considered single tax units in this paper. We use the terms “tax unit” and “taxpayer” interchangeably. 2

To explore the question of reassigning dependents, we create a new version of a panel dataset – the Tax Household Sample (THS) – which links tax records with a common address. Using these data, we can observe both the “sending” and “receiving” tax unit whenever the claimant of a child for tax purposes changes from one year to the next, without relying on information external to tax records. Since tax units are linked by addresses in the THS, we can isolate reassignments that occur within a single household from those due to children moving from one household to another. Furthermore, we estimate the impact of these reassignments on total household tax liabilities and observe how the tax implications of reassignments changed with tax laws. This provides a nuanced picture of the tax motivations for reassignments and the extent to which reassignments occur for tax minimization purposes. Researchers have long considered the potential impact of taxes on family formation decisions. While there is not a clear consensus in the literature, most research seems to suggest a small impact of tax policies on family structures. For example, Michelmore (forthcoming), Alm and Whittington (1999), and Sjoquist and Walker (1995) each observe small impacts on marriage patterns in response to the relative tax rates for single individuals and married couples. Milligan (2005) observes that fertility rates rise in response to a large increase in child tax benefits. However, Baughman and Dickert-Conlin (2009) find no positive effect of the earned income tax credit (EITC) on fertility rates among targeted populations. LaLumia, Sallee, and Turner (2015) find a small impact of taxes on the timing of births. While closely related to this earlier research on taxes and fertility, reassigning a child to a different tax unit within the same household – which is the focus of this paper – is likely a simpler behavioral response to the tax system. This is because the decision of who should claim a child for tax purposes has no functional impact on the family, outside of the impact on tax liabilities, if the tax claimants are already living together. 3

Two recent papers have explored such strategic claiming of dependents: Tong (2014) and Jones and O’Hara (2016). Tong (2014) considers the degree to which dependents remain in the same tax unit over time, and estimates that 12 percent of children claimed for EITC purposes are in completely different tax units the following year. An additional 5 percent of EITC children are in tax units where one of the filers is either an addition to the return (as would occur from marriage) or is no longer on the return (as would occur from divorce). However, lacking address information, Tong does not differentiate between instances where the living arrangement of the child changed and those where it did not. Hence, reassignments that are likely motivated by taxes, where living arrangements did not change, are grouped with those that are more likely motivated by other causes. Jones and O’Hara (2016) address the question of residency by using linked Internal Revenue Service (IRS) and Current Population Survey (CPS) data. Their linked data use name and address information to attach administrative records to Census datasets, which includes the household roster and relationship status within the household for linked observations.2 They compare the number of dependents claimed on tax returns to the number expected based on the relationships reported on the CPS questionnaire. They find higher rates of discrepancies between the expected and actual claimant of a child in households with EITC eligible individuals than in non-EITC households. Additionally, the likelihood of discrepancies shifted in 2009 when taxpayers began receiving additional EITC benefits for a third child – which provides further evidence of tax motivated reassignments. Although their work highlights an important and previously underexplored behavioral response to taxation, they acknowledge that their analysis is 2 The linking process is a probabilistic match procedure based on the name, address, date of birth, and gender of the individual. See Wagner and Layne (2014) for details on the matching procedure, and see Bond et al. (2014) for details on the success rate of these matches and an analysis of potential biases from mismatches. 4

limited by the use of cross-sectional data rather than panel data and is constrained by the quality of relationship imputations in the CPS data. We overcome these limitations with the THS dataset. This dataset follows taxpayers over time, includes the physical address where each taxpayer resides, and directly observes the movements of dependents across tax units – along with any associated impacts to tax liabilities. We observe reassignments occurring within multiple tax unit households. On average, these reassignments result in a household level tax benefit, suggesting that income taxes affect reassignment decisions. For example, in 2010 we estimate that households received an average federal income tax benefit from reassignment of $600 per dependent. Consistent with the observations of Jones and O’Hara (2016), we also document a shift towards three child tax units in 2009 among those eligible for EITC benefits. However, this may have been to the financial detriment of these households. Within households where an EITCeligible child was reassigned to a three-child tax unit in 2009, this resulted in an average increase in tax liabilities for the household relative to a counterfactual of the child remaining with their claimant from the prior year. I. BACKGROUND ON EITC RULES FOR CLAIMING A CHILD Although the IRS refers to an eligible dependent for EITC purposes as a child, the dependent does not have to be the biological or adopted child of the claimant. In order to claim a child for EITC purposes, the dependent must meet three rules for eligibility. First, the child must have lived with the taxpayer for at least half of the year. Second, the child must have been under age 19 at the end of the tax year, a full-time student (for whom the age limit is under age 24), or totally and permanently disabled (for whom there is no age limit). Finally, the child must be the 5

taxpayer’s son or daughter (by blood or adoption), stepchild, foster child, brother, sister, stepbrother, stepsister, or their descendant. Hence, a sibling, niece, nephew, or grandchild can legally be claimed as a dependent for EITC purposes on a taxpayer’s return. Importantly, while a dependent may only be claimed for EITC purposes on a single tax return, when several tax units reside together in a household it is possible for a child to meet the current eligibility tests for multiple taxpayers. For example, in the case of an unmarried cohabiting couple, as long as the two parents agree on who will claim the child, either one may legally claim their child for EITC purposes. Similarly in the case where a child lives with both her parents and grandparents, either may generally claim the child as a dependent and receive the associated EITC benefits.3 Furthermore, the level of EITC benefits are determined based only on the income of the tax unit claiming the child. Hence, there is an opportunity for coordination within multiple tax unit households regarding who claims a child, since not all eligible claimants qualify for the same tax benefits.4 Although tax laws introduce the opportunity for some taxpayers to coordinate who in their household claims dependents for tax purposes, it is necessary that these taxpayers have a sufficient understanding of the tax system to recognize tax minimization opportunities. 3 The ability to do so may, however, be subject to additional restrictions depending on the incomes of each tax filer. Typically, if the parent of a child does not claim the child for the EITC, the child can legally be claimed by the coresiding related non-parent who had the highest adjusted gross income (AGI) during the year as long as their AGI is above that of the parent. See IRS publication 596 for additional details on current eligibility rules (www.irs.gov/publications/p596/ch02.html). These rules have changed over time. In 1991, the eligibility rules went from a dependency support test (or household maintenance test if not married) to a residency test with an AGI tiebreaker applied in all cases (Holtzblatt and McCubbin, 2004). Given problems with the universal AGI tie-breaker, since 2002 it only applies in more limited circumstances. 4 While we focus here on the eligibility of children for EITC purposes since that is the tax credit for dependents that requires a shared residence, other benefits for dependents – including the child tax credit and head of household status – have separate requirements. See Holtzblatt and McCubbin (2003) for a discussion of the different definitions of qualifying children used for various tax benefits. 6

Specifically, taxpayers (or their advisors) must understand the regulations regarding who in the household may claim the dependent for EITC purposes, as well as have sufficient knowledge of the tax system to recognize which tax unit within the household would benefit most from the additional dependent. Previous research suggests that taxpayers have at least some understanding of the incentives from the EITC, although it is not obvious whether they have sufficient knowledge to minimize their liabilities with dependent reassignments. Meyer and Rosenbaum (2001) demonstrated that the EITC increases employment rates among targeted populations, which indicates at least a basic recognition of the benefit structure. Additionally, Eissa and Hoynes (1998) demonstrate that while the EITC has a positive impact on labor supply among single women, it has a negative impact on the employment among married women, which is consistent with the phase-out of benefits and an understanding of how employment interacts with the credit. Mortenson and Whitten (2017) find that many tax units report income close to the liabilityminimizing point – which is virtually always somewhere in the EITC plateau region – including in years where this income amount changes by thousands of dollars due to policy changes. Conversely, Chetty and Saez (2013) observe a limited ability for teaching taxpayers the amount that they should work to maximize their credits and suggest that on the intensive margin it is less clear that workers understand how small shifts in income impact their benefits. Similarly, Bhargava and Manoli (2015) use a field experiment and find that many EITC-eligible taxpayers do not claim the credit due to lack of awareness of their eligibility. In summary, while it appears that EITC-claiming tax units often have some knowledge of the EITC benefit structure, there are limits to this knowledge which could serve to reduce any observed effects. 7

II. DATA The Tax Household Sample (THS) data used for this study are drawn from the universe of individual income tax records for tax years 2007 to 2010.5 The sample construction follows the method developed in Larrimore, Mortenson, and Splinter (2017), but observations are sampled at the individual rather than household level, as this study tracks individual children. It also uses an alternative two address matching approach as explained below. The development of the THS consists of three steps: compiling tax records from a panel of individual level tax data for a nearly comprehensive set of U.S. residents, linking individuals into households by the address on their tax return or information return, and extracting a sample at the individual level. The list of individuals is made up of anyone listed on a tax return (e.g., Form 1040 or Form 1040-EZ) for a given tax year, including filers and their dependents, and non-filers with at least one information return for that year who are alive at the end of the year and younger than 100 years old. These individuals are restricted to U.S. residents, and so only include those with an address in one of the fifty states or the District of Columbia. Rather than selecting a single address for each individual, we select up to two addresses: a PO Box address and a street address. For filers, the address on the tax return fills one of the two address slots; addresses on information returns can fill the other slot. For non-filers, all addresses come from information returns, where the first address after sorting that matches the selection criteria is selected for each slot. Street addresses are standardized with over 200 replacements of various character strings. For example, ‘FIRST’ is replaced with ‘1ST’ and ‘STREET’ is replaced with ‘ST’. 5 These tax records include both annual tax return filings as well as information returns such as Form W-2 and Form 1099. These data comes from an individual level dataset of tax records collected by the IRS and which has recently been used for tax related research including Chetty et al. (2014) and Larrimore, Mortenson, and Splinter (2017). 8

These cleaned addresses are used to combine individuals into households. For individuals with only a street address or only a PO Box address, we assign the same household identifier to all individuals sharing the same address and ZIP Code combination. For individuals with both a street address and a PO Box address, we first link individuals sharing the same street address. Next, we link all individuals with the same PO Box address, whether they have only a PO Box address or both types of addresses, as long as street addresses do not conflict. Finally, for unmerged single person households we link individuals who share a PO Box address with the individual, regardless of street address. This final step is included to help address any errors or idiosyncratic reporting that remains following the address cleaning procedure. Using these data, we create a one percent random sample of all U.S. residents by selecting 100 four-digit Taxpayer Identification Number endings, and sample these selected individuals and all other members of their household. The distribution by household size in each year is shown in Table 1. Relative to 2010 Census estimates, the THS has more single person households and fewer two person households, but otherwise closely parallels the Census household size distribution.6 III. SAMPLE SELECTION To observe the reassignment of children between tax units within households over time, we narrow our sample to children ages 16 and younger living in households with at least two tax units for two consecutive years. The age limit of 16 ensures our sample is comprised of children 6 For two-person households not on the same tax return, any deviation in the selected address fields will usually result in the observations failing to merge. In our sample, this will result in two separate one person households. 9

who could be claimed for both the EITC and the child credit.7 In households with only two tax units, we also require that at least both of the primary filers (those adult filers listed first on the tax return) be at least 20 years or older or receive the EITC or refundable child credits. As the sample only includes children claimed as a dependent in both years, we do not measure decisions related to assigning newborns among different tax units. While we do not require that the physical address or the adult roster of a household remain constant over time, we restrict our sample to cases where the primary claimant of the child in year t-1 continued to live with them in year t. Since the relationship test to qualify for claiming a child for EITC purposes is time-invariant as long as a claimant and the child continue to reside together in the same household, this means that the claimant in year t continues to be eligible to claim the child in subsequent years even if they do not do so. Children living in households with multiple tax units are common in the U.S. In 2010, the THS identifies 66.6 million children aged 16 or younger claimed as dependents on tax returns, and around one third of these children (22.3 million) lived in a household that contained more than one tax unit. The likelihood that a child lives in a multiple tax unit household is even higher among the population targeted by the EITC. Of the 31.2 million children living in a household where at least one taxpayer claimed the EITC, over half (16.9 million) lived in a multiple tax unit household. Table 2 outlines the restrictions leading to our final sample in 2010. The most important restriction is the limitation to multiple tax unit households in 2010, which dropped two-thirds of children. Next, we eliminate children living in single tax unit households in 2009, in order to 7 Qualified child requirements for the EITC and child credit overlap substantially, but for the child credit, qualifying children must by younger than 17 years old (hence, our final year age restriction of 16), must be claimed as a dependent, and Individual Taxpayer Identification Numbers (ITINs) are valid, in addition to Social Security Numbers (which are required for the EITC). 10

avoid reassignments that resulted from changes in living arrangements with a new tax unit entering the household. Going further along these lines, we limit the sample to “persistent” multiple tax unit households, which we define as having the primary claimant of the dependent in 2009 continue living in the child’s household in 2010.8 Finally, we limit the sample to children claimed for the EITC in either 2009 or 2010 (“EITC children”), which generates our final sample of 8.2 million EITC children in persistent multiple tax unit households. IV. FREQUENCY OF REASSIGNING CHILDREN Within our sample of EITC children in persistent multiple tax unit households, we classify dependents as being “reassigned” from one tax unit to another within the same household if the claimant in year t-1 does not claim the child in year t. In the case of a married couple claiming a child, the child is only considered reassigned if neither parent claimed the child in the prior year. Of the 8.2 million EITC children in our sample of persistent multiple tax unit households, 410,000 shifted from one tax unit to another in 2010. Hence, approximately 5.0 percent of these EITC children were reassigned across tax units within their household from 2009 to 2010.9 This magnitude of reassignment over time is relatively consistent over the period of analysis from 2007 through 2010. Between 2007 and 2008, 5.2 percent of children in persistent 8 We exclude these individuals from the sample because the tax unit change is likely occurring for broader family formation reasons, such as marriage, divorce, or custody decisions, rather than for tax minimization purposes. The exclusion of these cases largely explains the lower rate that we observe for children moving across tax units than is observed by Tong (2014). 9 Not all children in persistent multiple tax unit households, however, are legally able to be reassigned to someone else in their household, as would be the case if the child’s parent lives only with unrelated individuals. Although the tax data cannot be used to observe how many such children are ineligible for reassignment, this limitation means 5 percent represents a lower bound on the share of eligible children in these households who are reassigned. 11

households with multiple tax units were reassigned from one taxpayer to another. Between 2008 and 2009, 5.1 percent were reassigned.10 A. Types of multi-family households that reassign dependents The IRS data does not contain relationship statuses – either between individuals listed on the same tax return (with the exception of spouses) or between people living in the same household on different tax returns. This limits the ability to observe the precise relationships of taxpayers who reassign dependents. However, using a variant of the Persons of Opposite Sex Sharing Living Quarters (POSSLQ) method, which is commonly applied to infer relationship status when it is unavailable, we are able to obtain broad estimates of relationship statuses in the tax data (see Casper and Cohen, 2000, and Fitch, Goeken, and Ruggles, 2005, for overviews of this approach and see Dokko, Li, and Hayes, 2015, for a modification of this approach that incorporates age bands into the relationship status imputation). We consider three relationship statuses in the data: independent cohabiting couples, multigenerational households, and roommates/other living situations. For these purposes, independent cohabiting couple are assumed to be instances with: exactly two unmarried, non-dependent taxpayers in the household, of opposite sex, and within 15 years of age.11 Multigenerational households are those with at least a 16-year age gap between the primary taxpayer on two returns in the household. Roommates and other living situations are the residual households with multiple tax units, which includes any instances where primary taxpayers within the household are within 15 years of age, 10 While the share reassigned fell slightly from 2007 through 2010, the number of children who were reassigned increased during this period from 340,000 children to 410,000 children. 11 A known limitation of this approach is that it cannot capture unmarried same-sex couples as a cohabiting couple. It also may capture as cohabitating couples roommates of opposite genders. 12

but where there are at least three non-dependent taxpayers within the household or there are two taxpayers of the same gender.12 Table 3 shows the frequency of EITC children across these three multiple tax unit household types. Multigenerational households are the most prevalent, representing 60 percent of these living arrangements, compared to 17 percent for independent cohabiting couples, and 24 percent for roommates/other living situations.13 Reassignments of EITC children occur disproportionately among cohabiting couples, either living independently or as part of a larger household. Just over 8 percent of EITC children living with independently filing cohabiting couples were reallocated from one claimant to the other between 2009 and 2010 (110 thousand of the 1.37 million). This compares to a 3 percent reassignment rate for children in multigenerational households and a 7 percent reassignment rate for children in roommate/other households. Within the multigenerational and roommate/other households, however, about half of reassignments occur between two taxpayers who are potentially a cohabiting couple despite living with others (defined here as two unmarried individuals of opposite sex who are within 15 years of age and live together). In total, cohabiting couples, whether living independently or within larger households, account for about two-thirds of reassignments. V. TAX-EFFECTS OF REASSIGNING CHILDREN 12 It is, of course, possible for there to be multiple distinct types of relationships within a household. However, these three sets are designed to be both exhaustive and mutually exclusive. While the EITC regulations would not allow unrelated roommates to reassign dependents across the household members, if the roommates are closely related (such as a brother or sister), then reassignment would be permitted. 13 Using CPS data – where relationship statuses can be observed directly – Larrimore, Mortenson, and Splinter (2017) also observe that multigenerational households are the most frequent type of household containing multiple tax units, especially those with adult children living with their parents. 13

Although these data indicate that some children are reassigned each year from one tax unit to another within the same household, the presence of reassignments alone does not necessarily demonstrate a tax response. We are therefore interested in whether the reassignments generally reduce the household’s overall tax liability. To do so, we use a tax change calculator created for this study, which accounts for changes in personal exemptions, standard deductions, the difference in tax bracket lengths between single and head of household filing statuses, child tax credits, and the EITC. Specifically, we compute the change in tax liability in year t resulting from removing the child from the year t claimant and adding them to the claimant from year t- 1.14 The change in tax liabilities from reassignment are calculated as the sum of actual tax liabilities in year t for the pair of tax units minus the sum of tax liabilities for the pair if the dependent remained with the claimant from year t-1. In the absence of any tax motivation for reassignment, generally half of children who are reassigned should result in higher tax liabilities and half in lower tax liabilities. To the extent that more than half result in lower total household tax liabilities, it suggests that a contributing factor for the decision to reassign dependents is reducing the total household tax burden. In 2010, twothirds of reassignments of EITC children in persistent multiple tax unit households resulted in a reduction in total tax liabilities. The average federal income tax benefit from these reassignments was $600 per dependent, or $250 million for all reassigned EITC children.15 This tax change calculator result is replicated using the NBER TAXSIM program, which also shows that the 14 Some prior year claimants may be current year non-filers. We estimate the extent to which they would receive the EITC and refundable child credit and pay taxes (assuming single filing status) based on wages only, but limited such that adding a child never causes a non-filer to have a positive tax burden. 15 Although this estimate is above the Jones and O’Hara (2016) 2005-2010 average estimate of about $500 per reassigned child, they estimate a larger aggregate total tax effect of $440 million in 2010. This reflects the fact that they include the entire stock of children claimed by someone other than their parent (or other expected claimant based on CPS relationship status imputations) whereas we focus on the one-year flow of reassigned children and also limit to those in persistent multiple tax unit households. 14

addition of state income taxes increases the average tax benefit of reassigning by about $40 (see Feenberg and Coutts, 2003, for an overview of the TAXSIM program). These tax benefits represent an average of four percent of combined tax unit AGIs. Moreover, they appear to be concentrated among pairs where both tax units have incomes in the EITC range.16 Hence, while there are households that may change the tax assignment of dependent children from one year to the next for non-tax reasons or in ways that increase their tax liabilities, it appears that the minimization of overall tax burdens for the tax unit pair is a contributing factor for some households. A. Impact of 2009 tax changes on reassignment Thus far, we have largely focused on reassignments that occurred between 2009 and 2010, observing the extent to which tax liabilities are reduced for tax unit pairs within a household that reassign dependent children. The tax motivations for reassignments can be further assessed by considering how reassignments shifted in response to changes in tax legislation. In 2009, there were several tax changes that impacted the relationship between the number of dependents and tax liabilities. The most important of these is an expansion of the EITC. Prior to 2009, EITC benefits increased with the number of children, but only for the first two children. As a part of the American Recovery and Reinvestment Act, starting in 2009 there was an added EITC benefit of up to $630 for a third child which may incentivize some reassignments towards tax units with three dependent children. 16 Among the children in our sample who are claimed for EITC benefits in one year, but shift from or to a tax unit with an AGI over $50,000, which is beyond the EITC eligible range, there was an average loss from the reassignment. As a result, the benefits from reassignments generally do not appear to be resulting from children in tax units that are beyond the EITC range shifting into tax units that are in the EITC range, but instead from reassignments among multiple relatively low income tax units. 15

We first analyze potential responses to the EITC change by exploring the extent to which reassignments changed in 2009 when these new policies went into effect. Table 4 shows the share of reassigned children by the number of dependents in the “sending” and “receiving” tax units. Coinciding with the introduction of the three-child EITC credit, in 2009 there was a large increase in the share of reassignments that led to a tax unit having at least three children. In 2008, 10 percent of reassignments of EITC children occurred from a tax unit with two or fewer children to one with at least three children. In 2009, the share increased to 16 percent of reassignments. Overall, the share of reassigned children going to tax units with at least three dependents (irrespective of the number of dependents in their initial tax unit) rose from 14 percent in 2008 to 23 percent in 2009. This is consistent with taxpayers responding to the tax policy change via reassignments. It is also consistent with the observations of Jones and O’Hara (2016) that the EITC changes motivated some EITC taxpayers to claim an additional child in order to take advantage of the expanded tax credits after the policy change. Further supporting that this shift was in response to the EITC policy change, there was no similar reassignment response among children who were not claimed for EITC purposes. In both 2008 and 2009, about 9 percent of these non-EITC children were reassigned from tax units with two or fewer dependents to ones with three or more. Therefore, the increase in reassignments to add a third dependent does seem to be a reaction to the EITC policy rather than a more general trend. A priori, the tax liability implications of reassigning a child in order to receive the threechild EITC are unclear. While the “receiving” tax unit will lower their tax liability, this will be offset to some degree by an increased tax liability for the “sending” tax unit. Reallocating dependents to receive the three-child EITC can be financially beneficial if the two taxpayers 16

have divergent incomes. However, when the two potential claimants have similar incomes, consolidating dependents can often increase household tax liabilities. Consider two tax units, A and B, living in the same household with a total of three children in 2010. Further, assume both tax units have earned income that places them in the EITC’s plateau region, where the EITC is maximized. If one child is claimed by tax unit A and the other two children are claimed by tax unit B, the household would receive $8,086 of EITC ($3,050 for tax unit A claiming one child, and $5,036 for tax unit B claiming two). However, if all three children are reassigned to only one of the two tax units, the total EITC benefits would be $6,009 ($5,666 for the tax unit claiming the children and up to $343 for the childless tax unit).17 Hence, in this example the reassignment to receive the three-child EITC could cost the household over $2,000 in EITC benefits.18 Recognizing that the impact of reassignments on tax liabilities is not always obvious, Table 5, considers the average effect on tax liabilities from observed reassignments. In all cases, the tax liabilities are summed across the tax unit pair and therefore reflect the overall tax implications of a single child reassignment relative to not reassigning. In 2008, reassigning EITC children within the household resulted in an average tax benefit of $810. The positive benefits are concentrated, however, among households reassigning to fewer dependents per tax unit – as tax units that reassigned children to tax units with two or fewer dependents had an average tax benefit from the reassignment of $980 per dependent 17 The maximum credit for a childless individual was $457, although the childless benefit phase-out range ended before the EITC benefits for tax units with children is fully phased in. $343 was the maximum childless credit available in 2010 to someone that is in the fully phased in range of the EITC benefits for those with children. 18 In addition to lost EITC benefits, the reassignment could also prevent the “sending” tax unit from claiming head of household status if they no longer have a qualifying person for head of household status as an exemption on their tax return. They also will lose any child tax credit benefits that they would have otherwise been eligible for, although these benefits may be more or less than the “receiving” tax unit’s child tax credit benefits depending on their specific financial situation. 17

whereas those that reassigned to tax units with three or more dependents had an average loss of $190 per reassigned dependent. In 2009, the average benefits from reassignments fell substantially for tax unit pairs of each size, but remained positive for reassignments to small tax units and negative for reassignments to larger ones. It fell less among those that reassigned to claim the three-child EITC benefit – perhaps because the new EITC benefits reduced tax liabilities for the “receiving” tax unit, which offset some of the loss from the reassignment. However, even with the additional three-child EITC benefit, children reallocated from smaller tax units to those that, after the reassignment, have at least three dependents resulted in an average reassignment loss in 2009 of $240. In 2010, children reallocated in this way resulted in an average loss of $250. This suggests that while tax units showed a desire to claim this additional credit, to the extent that they reassigned dependents in order to do so it was a net financial loss for the household. B. Impact of the child tax credit A remaining puzzle from Table 5 is the substantial drop in average reassignment gains that occurred in 2009 for all reassigning tax units. This can likely be attributed to changes to the Additional Child Tax Credit (ACTC), the refundable portion of the child tax credit. These changes occurred between 2007 and 2009 and reduced the potential benefits from reassignment. The ACTC provides a refundable tax credit of up to $1,000 per child, with a phase in rate of 15 percent for earnings over a refundability threshold. In 2007, this refundability threshold was $11,750 – so if a taxpayer had no tax liabilities they needed to have about $18,400 of earned income to receive the full refundable credit for one child and about $25,100 of earnings to receive it for two children. The refundability threshold was lowered to $8,500 in 2008, and was 18

lowered further to $3,000 in 2009. As a result, in 2009 a taxpayer with no tax liabilities could receive the full refundable child tax credit for one child if they had about $9,700 of earned income and could receive the full refundable child tax credit for two children if they had about $16,300 of earned income. By lowering the refundability threshold, the likelihood increased that both the “sending” and “receiving” tax units qualified for the entire refundable child tax credit – thereby reducing the potential ACTC benefits from reassignment.19 The impact of this can be observed in Table 6, which illustrates the average tax benefits from reassignments in 2009 – but under the counterfactual that prior-year EITC and child credit tax laws remained in effect. When applying the prior year tax policies (no three child EITC and the 2007 child credit refundability threshold of $11,750), reassigning tax units in 2009 would have experienced an average benefit of $750 from the reassignment of the dependent, similar to the observed benefit from reassigning in 2008. When holding the EITC policy constant, the change in ACTC policies between 2007 and 2009 reduced the average benefit to $540, which accounts for most of the decline in reassignment benefits in that year. The remaining reduction in tax benefits from reassignment to $460 came from the new three-child EITC policy. As expected, households shifting to larger tax units saw a relative increase in their household level reassignment benefits from the isolated EITC policy change (from -$460 to -$170) and those shifting towards smaller tax units saw a decline from the new EITC policy (from $840 to $650). VI. DISCUSSION AND CONCLUSION 19 The only case where the tax law change would increase the gap in child tax credits between the two tax units is a case where the earnings of the higher earning tax unit are sufficiently low that the refundable credit was not fully phased in when using the old threshold. In these cases, depending on the earnings of the lower earning tax unit, it is possible that the child credit related tax benefit from reassigning could increase. 19

This paper provides evidence of coordination between separate tax units at the household level when assigning children for tax purposes. Using a new household level panel dataset comprised of tax records, we observe the reassignment of dependents from one taxpayer to another over two year periods. We estimate that on average these reassignments lowered combined tax liabilities. In each year from 2008 through 2010 there was an average tax benefit to reassigning households of between $460 and $810 per reassigned dependent. This suggests that taxes were a contributing factor to the reassignment decision. As further evidence that taxpayers are considering tax policy when making reallocation decisions, changes in EITC rules in 2009 – which allowed for additional EITC credits for a third child – coincided with a greater share of reassigned EITC children to appear on tax returns containing at least three dependents. However, while some taxpayers appear to reassign dependents within their households at least in part to reduce the household level tax liability, there is also evidence that taxpayers may not consider the full repercussions of reallocations. Households who reassigned dependents into larger tax units experienced a net increase in their tax liabilities, on average, because of the reassignment. These findings reinforce the importance of considering the full range of behavioral responses to tax legislation, including those related to household formation and tax unit structure. Further, it emphasizes the notion that tax units are not a perfect proxy for economic sharing units, as we document apparent coordination across tax units within households, especially among cohabiting couples. However, this coordination has limits, as some taxpayers may prioritize minimizing their own tax burden despite increasing their household tax burden. Taxpayers may also respond to particularly salient aspects of tax policy, at least in the short run, without considering all provisions in the individual income tax code. The extent to which any 20

limitations on household level tax minimization are due to each of these two competing factors is beyond the analysis of this paper, but would be a valuable avenue for future study. 21

ACKNOWLEDGEMENTS AND DISCLAIMERS We thank Tom Barthold, Heidi Schramm and participants of the National Tax Association Spring Symposium for helpful comments and discussions. The results and opinions expressed in this paper reflect the views of the authors and should not be attributed to the Federal Reserve Board. This paper embodies work undertaken for the staff of the Joint Committee on Taxation, but as members of both parties and both houses of Congress comprise the Joint Committee on Taxation, this work should not be construed to represent the position of any member of the Committee. CONFLICT OF INTEREST DISCLOSURE The authors have no financial arrangements that might give rise to conflicts of interest with respect to the research reported in this paper. REFERENCES Alm, James and Leslie A. Whittington, 1999. “For Love or Money? The Impact of Income Taxes on Marriage.” Economica 66 (263), 297-316. Baughman, Reagan and Stacy Dickert-Conlin, 2009. “The Earned Income Tax Credit and Fertility.” Journal of Population Economics 22 (3), 537-563. Bhargava, Saurabh and Dayanand Manoli, 2015. “Psychological Frictions and the Incomplete Take-Up of Social Benefits: Evidence from an IRS Field Experiment.” American Economic Review 105 (11), 1-42. Bitler, Marianne, Hilary Hoynes, and Elira Kuka, 2017. “Do In-work Tax Credits Serve as a Safety Net?” Journal of Human Resources 52 (2), 319-350. Bond, Brittany, J., David Brown, Adela Luque, and Amy O’Hara, 2015. “The Nature of the Bias when Studying only Linkable Person Records: Evidence from the American Community Survey,” Census Bureau CARRA Working Paper 2014-08. Casper, Lynne and Philip Cohen, 2000. “How does POSSLQ Measure Up? Historical Estimates of Cohabitation.” Demography 37 (2), 237-245 Chetty, Raj and Emmanuel Saez, 2013. “Teaching the Tax Code: Earnings responses to an Experiment with EITC Recipients.” American Economic Journal: Applied Economics 5 (1), 1- 31. Chetty, Raj, Nathanial Hendren, Patrick Kline, and Emmanuel Saez, 2014. “Where is the Land of Opportunity? The Geography of Intergenerational Mobility in the United States.” Quarterly Journal of Economics 129 (4), 1553-1623. 22

Dokko, Jane, Geng Li, and Jessica Hayes, 2015. “Credit Scores and Committed Relationships.” Board of Governors of the Federal Reserve System Finance and Economics Discussion Series 2015-081. Eissa, Nada and Hillary Hoynes. 1998. “The Earned Income Tax Credit and the Labor Supply of Married Couples.” NBER working paper no. 6856. Feenberg, Daniel and Elizabeth Coutts. 1993. “An Introduction to the TAXSIM Model.” Journal of Policy Analysis and Management 12 (1), 189-194. Fitch, Catherine, Ron Goeken, and Steven Ruggles, 2005. “The Rise of Cohabitation in the United States: New Historical Estimates.” Minnesota Population Center Working Paper 2005-03. Holtzblatt, Janet and Janet McCubbin, 2003. “Whose Child is it Anyway? Simplifying the Definition of a Child.” National Tax Journal 56 (3), 701-718. Holtzblatt, Janet and Janet McCubbin, 2004. “Issues Affecting Low-Income Filers.” In Aaron, Henry and Joel Slemrod (ed.) The Crisis in Tax Administration, 148-200. Brookings, Washington, DC. Jones, Maggie R. and Amy B. O’Hara, 2016. “Do Doubled-Up Families Minimize Householdlevel Tax Burden?” National Tax Journal 69 (3), 613-640. LaLumia, Sara, James Sallee, and Nicholas Turner, 2015. “New Evidence on Taxes and the Timing of Birth.” American Economic Journal: Economic Policy 7 (2), 258-93. Larrimore, Jeff, Jacob Mortenson, and David Splinter, 2017. “Household Incomes in Tax Data: Using Addresses to Move from Tax Unit to Household Income Distributions.” Board of Governors of the Federal Reserve System Finance and Economics Discussion Series 2017-002. Meyer, Bruce D. and Dan T. Rosenbaum, 2001. “Welfare, the Earned Income Tax Credit, and the Labor Supply of Single Mothers.” Quarterly Journal of Economics 116 (3), 1063-1114. Michelmore, Katherine. Forthcoming. “The Earned Income Tax Credit and Union Formation: The Impact of Expected Spouse Earnings.” Review of Economics of the Household. Milligan, Kevin, 2005. “Subsidizing the Stork: New Evidence on Tax Incentives and Fertility.” Review of Economics and Statistics 87 (3), 539-555. Mortenson, Jacob and Andrew Whitten, 2017. “Bunching to Maximize Refunds: Evidence from the U.S.” Working paper. Sjoquist, David L. and Mary Beth Walker, 1995. “The Marriage Tax and the Rate and Timing of Marriage.” National Tax Journal 48 (4), 547-558. Tong, Patricia K., 2014. “Tracking EITC Qualifying Children Over Time.” In Proceedings of the 107th Annual Conference on Taxation, 84-97. National Tax Association, Washington, DC. Wagner, Deborah, and Mary Layne, 2014. “The Person Identification Validation System (PVS): Applying the Center for Administrative Records Research and Applications’ (CARRA) Record Linkage Software.” CARRA Working Paper 2014-01. 23

Table 1 Number of households by household size, 2010 (millions) 2010 Size of 2007 2008 2009 2010 Decennial Household THS THS THS THS Census 1 34.0 35.2 35.7 35.9 31.2 2 33.5 33.2 33.3 33.5 38.2 3 17.9 17.9 17.8 17.9 18.8 4 14.3 14.4 14.6 14.9 15.6 5 7.0 7.1 7.1 7.4 7.5 6 3.2 3.3 3.3 3.5 3.1 7-10 2.3 2.4 2.4 2.7 2.3 Total 112.3 113.4 114.1 115.8 116.7 Notes: THS is the Tax Household Sample. Individuals living in group quarters are excluded, which is defined in the tax data as households with 11 or more individuals. Household weights for each individual in the THS are 99.99 divided by household size, as larger households have more individuals sampled. Source: American FactFinder (Table H13) from the U.S. Census Bureau 2010 decennial census, Tax Household Sample, and authors’ calculations. 24

Table 2 Sample restrictions: number of children in persistent multiple tax unit households (millions) Total children age 16 or younger in 2010 66.58 In household with a single tax unit in 2010 -42.33 In household with a single tax unit in 2009 -10.06 Original claimant no longer resides with child -1.58 Not claimed for EITC purposes in either 2009 or 2010 -4.36 Final Sample 8.25 Notes: Restrictions stack on one another. Source: Tax Household Sample and authors’ calculations. 25

Table 3 Household types and reassigning types among EITC children in persistent multiple tax unit households in 2010 EITC children in persistent multiple tax Reassigned unit households EITC children (millions) (%) (millions) (%) Independent cohabitating couple households 1.37 17 0.11 27 Multigenerational households 4.93 60 0.17 42 Reassignments between potential cohabiting couples 0.09 Other reassignments 0.09 Roommates and other household types 1.95 24 0.13 31 Reassignments between potential cohabiting couples 0.06 Other reassignments 0.07 Total 8.25 100 0.41 100 Cohabitating couple reassignments: Independent & in households with others 0.26 63 Notes: Counts are among children in households where one or more tax unit claims the EITC in either 2009 or 2010. For persistent multiple tax unit households: cohabitating couples live in two unmarried tax unit households, are within 15 years of age, and opposite sex; roommates are any type or number of tax unit where primaries ages are all within 15 years; multigenerational are those with ages more than 15 years apart. For reassigning children, definitions are based only on the two tax units claiming the child in 2009 and 2010. If these tax units are unmarried tax units, within 15 years of age, and of opposite sex, they are considered a potential cohabiting couple – even if they also live with others in the household. Multigenerational reassignments are cases where the child shifts from one tax unit to another where the primary filer is at least 16 years apart in age. Reassignments across roommates are instances where the child is reassigned across two tax units within 15 years of age where either one of the tax units consists of a married couple or the two tax units each consist of unmarried filers of the same gender. Source: Tax Household Sample and authors’ calculations. 26

Table 4 Share of reassigned EITC children in persistent multiple tax unit households, by number of dependents in starting and ending tax units Share of reassigned EITC children initial year <3 3+ All dependents dependents <3 dependents 0.74 0.12 0.86 final year 3+ dependents 0.10 0.04 0.14 2008 All 0.84 0.16 1.00 <3 dependents 0.67 0.10 0.77 final year 3+ dependents 0.16 0.07 0.23 2009 All 0.83 0.17 1.00 <3 dependents 0.64 0.14 0.78 final year 3+ dependents 0.14 0.07 0.22 2010 All 0.79 0.21 1.00 Notes: The initial year is always one year prior to the listed final year. Estimates are shares of children in persistent multiple tax unit households who were reassigned and claimed for the EITC in either of the two years. Source: Tax Household Sample and authors’ calculations. 27

Table 5 Average tax benefits of reassigning EITC children in persistent multiple tax unit households, by number of dependents in starting and ending tax units Tax benefits (dollars) initial year <3 3+ All dependents dependents <3 dependents 910 1,380 980 final year 3+ dependents -160 -260 -190 2008 All 790 930 810 <3 dependents 610 910 650 final year 3+ dependents -240 10 -170 2009 All 450 550 460 <3 dependents 720 1,190 800 final year 3+ dependents -250 20 -150 2010 All 540 790 600 Notes: All values are rounded to the nearest $10. The initial year is always one year prior to the listed final year. Tax benefits compare actual tax liabilities of the reassigning tax unit pair to counterfactual liabilities had the child not been reassigned. Estimates are for children in persistent multiple tax unit households who were reassigned and claimed for the EITC in either of the two years. The number of dependents are based on the total number of dependents in the tax unit claiming the child. Source: Tax Household Sample and authors’ calculations. 28

Table 6 Average tax benefit from reassigning EITC children in persistent multiple tax unit households in 2009 under prior year policies (dollars) All <3 dependents 3+ dependents EITC policy EITC policy EITC policy 2008 2009 2008 2009 2008 2009 2007 750 680 1,050 860 -250 50 Child credit 2008 680 610 980 800 -310 -20 policy 2009 540 460 840 650 -460 -170 Notes: All values are rounded to the nearest $10. Tax benefits compare actual tax liabilities of the reassigning tax unit pair to counterfactual liabilities had the child not been reassigned. Estimates are only for children in persistent multiple tax unit household who were reassigned and claimed for the EITC in either 2008 or 2009. 3+ or <3 dependent groups by 2009 tax unit claimant. Prior year policies only change child credit refundability thresholds and the remove the three-child EITC. Source: Tax Household Sample and authors’ calculations 29