Abstract
The authors use a unique data set of federal tax returns to analyze usage and participation patterns of the Earned Income Tax Credit (EITC) over the period 1989–2006. The authors find that most EITC recipients claimed the EITC for short periods, 61% for 1 or 2 years. Over the period examined, the EITC reached approximately 50 percent of the taxpayers with children. Finally, the authors find considerable income mobility among the EITC eligible population. Only 11 percent of those claiming the EITC in 1990 and in the third decile of income were in the same decile in 2003. They also find that 20 percent of EITC claimants claim the EITC for more than 5 years.
The Earned Income Tax Credit (EITC) was enacted in the Tax Reduction Acts of 1975 as a mechanism for providing tax relief to working low-income families suffering from the 1974–1975 recession. Proponents argued that the EITC was a tool that could be used to encourage work by “low-income families who might otherwise need large welfare payments” (U.S. Joint Committee on Taxation 1979, 51). Thus, the EITC was designed with two major objectives: first, to act as a safety net and provide relief to the working poor hit hard by the 1974 and 1975 economic recession and, second, to act as a work incentive for individuals and families to encourage them to move into the labor force.
Eissa and Liebman (1996) and Hotz, Mullin, and Scholz (2005) and others have found evidence that the EITC modestly increases labor supply. 1 Grogger (2004) and Dowd (2005) found a negative relationship between EITC and welfare. In combination, these results suggest that the EITC acts as a work incentive, mitigating the usage of other safety nets. Finally, Auerbach and Feenberg (2000) and Dowd (2005) found that the EITC acts countercyclicaly and is correlated with downturns. The EITC, therefore, also appears to be fulfilling its other major obective of being a safety net for individuals who are hit by individually specific or economy-wide shocks.
In this article, we try to disentangle the extent to which the EITC is used as a short-term safety net or a long-term income support. If a household’s income remains stable from year to year within the ranges of EITC eligibility, then we assume that the EITC acted as a long-term income support. We assume that the EITC primarily acts as a short-term safety net for households who receive the EITC for short amounts of time. Next, we investigate the causes that contribute to the movement from one portion of the income distribution to another.
We look at usage of the EITC over the period 1989–2006, with particular attention to the length of time that an EITC recipient claims the credit and on how likely a recipient is to reclaim the credit after a period of not claiming the credit. All else equal, there are four major reasons that a taxpayer might move in and out of receiving the EITC: (1) changes in income, (2) changes in the eligibility of children, (3) changes in family structure such as the gain or loss of spousal income, and (4) law changes that make the EITC phase-out more generous. These are not mutually exclusive. A divorce or adding a child to the family might result in short-term lower income.
We find that most EITC recipients claim the EITC for short amounts of time. Sixty-one percent have spells of one or 2 years. However, at the same time, we find that 20 percent of EITC recipients starting a spell, conditional on observing the taxpayer in 1989, claim the credit 5 or more years. Therefore, for some taxpayers, the EITC acts as a temporary safety net during periods of either anticipated or unanticipated income or family structure shocks. But the EITC also acts as a long-term mechanism of providing assistance to taxpayers with children who are entrenched in the lowest-income brackets.
Data
We use a representative random panel of individual income tax returns provided by the Statistics of Income of the Internal Revenue Service (IRS) that were filed between 1989 and 2006, the Continuous Work History Sample (CWHS). Most of the time we restrict our analysis to tax returns that claimed a dependent child either in the tax year being analyzed (the reference year) or, in the case of the panel data, in at least one of the years of the panel.
Selection into the CWHS panel is based on the randomly assigned last four digits of the primary taxpayer’s Social Security number (SSN). 2 The selection process results in an automatic refreshment characteristic; any taxpayer filing a return with the selected SSN as the primary taxpayer is included in the CWHS sample each year that they file as the primary taxpayer. 3
Unfortunately, because selection into the panel is based on the SSN of the primary taxpayer, and women tend to be listed as the secondary taxpayer on most married filing joint tax returns, the CWHS panel is biased toward oversampling men. 4 As a consequence, marital changes can result in no longer observing women. For example, if a single woman in the panel marries and files as the secondary taxpayer, she will no longer be sampled (unless her husband also has a CWHS SSN) or if a married woman divorces but does not have the selected SSN, she will not be sampled in subsequent years. For purposes of spell-length estimates, the bias against sampling women will result in shorter observed spell lengths.
Approximately 3 percent of tax returns are filed for years prior to the current tax year. We include these taxpayers who file untimely returns in the panel for the appropriate tax filing year. However, while we try to correct for untimely filers, the sample is censored in both 1989 and 2006. This censoring will reduce the number of times we observe a taxpayer claiming the EITC. Moreover, some taxpayers do not file returns in every year of the sample (22 percent of taxpayers observed in 1990 were either missing in 2003 or had changed marital status and were not observed). Generally, we keep these observations because both the observable and the missing returns tell us about the usage of the EITC.
The CWHS sample of tax returns is a rich data set that contains information on actual EITC receipt and a wide variety of income and demographic variables. Among some of the demographic variables are information about the number of dependents and the age of the taxpayer and dependents (starting in 1995).
Table 1 shows the sample sizes of each year of the CWHS data set based on the two-number sample. The sample sizes range from a low of 22,460 in 1989 to a high of 26,810 in 2006. Restricting the sample to taxpayers that have a child in at least 1 year cuts the sample size roughly in half. Further restricting the sample to taxpayers that have a child in the year of interest, or the reference year, eliminates almost two-thirds of the CWHS observations. Finally, the last two columns show how the CWHS weighted number of EITC recipients and the actual number of EITC recipients (last column) had a sharp increase from 15 million in 1993 to 19 million in 1994. This increase reflects the 1994 expansion of the EITC to taxpayers without a qualifying child and a more generous credit.
CWHS Panel Sample Sizes, and Number of EITC Recipients
Source: Authors' calculations and Internal Revenue Service.
Note: CWHS = Continuous Work History Sample; EITC = Earned Income Tax Credit.
Table 1 shows a steady increase in the number of returns that are included in the panel. However, these increases belie the underlying level of attrition in the panel. From one year to the next, there is approximately 8 percent attrition of CWHS observations. Attrition from the sample is due to deaths, nonfiling, and marital status changes. As discussed previously, marital status changes introduce a bias to the sample; rather than dropping out of the filing population some returns are simply not observed (e.g., single filer with CWHS SSN marries and becomes secondary on joint return). We estimate that 2,899 CWHS taxpayers died between 1989 and 2003, representing almost 12 percent of the CWHS tax returns in 1989; approximately equal to the average annual U.S. population death rates of slightly less than 1 percent (Xu et. al. (2010)). 5 Finally, we estimate that total attrition from the sample between 1989 and 2006 is approximately 40 percent.
Claim Rates
In order to understand whether the EITC acts primarily as a short-term safety net or as a long-term income support, it is necessary to understand how different groups claim the credit over time. In particular, we look at the percentage of taxpayers who claim the credit in any given year and over a longer period for which one might expect to have a qualifying child. For taxpayers who generally claim the credit only once or twice, we argue that the credit primarily acts as a safety net for brief income, employment, or family structure shocks. Alternatively, if taxpayers generally claim the credit for longer periods, then we argue that the credit is largely helping low-income families make ends meet while at the same time encouraging labor force attachment.
Figure 1 shows four different probabilities of claiming the EITC based on the CWHS panel data. The series in the figure include taxpayers for whom we might have missing returns for some of the years. All of the series are conditional on the taxpayer having a child in at least 1 year between 1989 and 2006. The top line, solid diamonds, of the figure indicates that roughly 50 percent of taxpayers who have a child in at least 1 year of the 18-year sample, claim the EITC at some point over the 18-year period! Obviously, the EITC has very broad reach to families with children who file tax returns. In 2006, for all taxpayers with children, the median wage income was $37,808; the median adjusted gross income was $3,174. Given that the zero-credit income for two or more children was $38,348, it is not surprising that a large portion of taxpayers with children claim the credit.
Probability of claiming the Earned Income Tax Credit (EITC) Source: Authors' calculation based on Continuous Work History Sample sample of tax returns from 1989 to 2006, conditional on tax payer having a dependent child sometime between 1989 and 2006. Note: Claim rates do not include the childless EITC. TP indicates taxpayer.
The second from the bottom line, open diamonds, in Figure 1 shows the percentage of taxpayers who have a child for at least 1 year who claim the EITC in the reference year. 6 There is a slight upward trend from 18 percent in 1989 to 23 percent in 1996, leveling off until an uptick in 2002 from 22 to 24 percent. The increase in 2002 is the combined result of the changes made in the Economic Growth and Tax Relief Reconciliation Act of 2001, which increased the phase-out threshold for taxpayers filing married jointly and the 2001 recession. The difference between taxpayers who claim the credit at least once, the solid diamond series, and taxpayers who claim the credit in the reference year, the open diamond series, shows the extent to which there is observed churning of taxpayers claiming the EITC. If all of the EITC recipients claimed the credit every year with a qualifying child, these two lines would be the same. Thus, the gap between these two lines is an indicator of the degree to which there is churning of recipients and dropping of panel members due to the sampling structure. The flip side of churning is that the gap shows the degree to which the EITC might be acting as a short-term safety net.
Because the CWHS sample potentially undersamples women, and especially women who change marital status, the amount of churning indicated by the two diamond lines in Figure 1 could be overstated. To get a sense of the degree to which our basic estimate overstates the amount of churning, the other two series presented in Figure 1 show the probability of claiming the credit for taxpayers who file joint returns. Restricting the sample to taxpayers who file joint returns reduces the number of observations from 12,155 in 1989 to 4,353. The solid square series presents the probability of claiming the credit at least once during the 18-year period, while the open square series presents the probability of claiming the credit in the reference year. Not surprisingly, overall claim rates vary by filing status, and are lower for taxpayers who file joint returns. The difference among joint filing taxpayers who claim the credit at least once, the solid square series, and joint filing taxpayers who claim the credit in the reference year, the open square series represents the amount of churning of claiming the credit for joint-filing taxpayers. The gap between the two joint-filer series ranges from 15 to 20 percentage points. Thus, even for taxpayers who file a joint return, there is still quite a bit of churning.
Figure 2 shows the probability of claiming the EITC conditional on either claiming the credit in 1991 or in 2004. The solid square markers are for taxpayers who claimed the credit in 1991, and the solid circle markers are for taxpayers who claimed the credit in 2004. The open square markers only include taxpayers who filed a joint return and claimed the credit in 1991 while the open circle markers only include taxpayers who filed a joint return and claimed the credit in 2004. 7
Forward and backward probability of claiming the Earned Income Tax Credit (EITC) Source: Authors' calculation based on Continuous Work History Sample sample of tax returns from 1989 to 2006, conditional on taxpayer having a dependent child sometime between 1989 and 2006. Note: Claim rates do not include the childless EITC.
Figure 2 shows a very steep drop off from either 1991 or 2004, with only 70 percent of those that claim the credit in 1991 also claiming the credit in 1992. After 15 years, in 2006, only 20 percent of those that claimed the credit in 1991 make a repeat claim in 2006. For the series conditional on claiming the credit in 2004, a similar pattern holds with less than 15 percent of the taxpayers claiming the credit in 1989. The two series that only included joint filers have a similar pattern to those of all filers. However, for the 2004 joint filers there appears to be a greater tendency to claim the credit in earlier years. Thus, for the 2 to 3 years surrounding 1991 and 2004, female sample bias does not appear to be a major part of the story. However, there does appear to be a problem with joint filers that we observe claiming the EITC in 2004; they have a much higher probability of claiming the credit in the early years than all filers do. This could be due to female sample bias, although the fact that the 1991 joint-filer series does not also exhibit this problem suggests that it may be something else. In particular, joint filers that we observe as joint filers in 1990 and as joint filers in 2004 and claiming the EITC may be a significantly different subpopulation.
As noted previously, one reason for changes in claim rates for taxpayers over time is the result of changes in family structure. In particular, because EITC eligibility is directly related to the presence of children in the household, we might suspect that the claim rates are related to the age of the child. Unfortunately, we only have the age of the child starting in 1995. Figure 3 shows the time path of claim rates as the qualifying child ages, conditional on observing the taxpayer in the year the child is born. In the initial year that the child is born, there is approximately a 43 percent probability of claiming the EITC. For the three starting years that we look at, claim rates decline in a fairly similar pattern. Some of the decline likely represents the normal anticipated shock that having a newborn has on family labor income in the year of birth, thereby reducing income in the birth year and increasing eligibility. Later, in the section on income mobility, over the period 1995–2002, we find that the loss in wage earnings associated with having a newborn for those who claim the EITC ranges from a low in the birth year 1999 of 7.5 percent to a high in the birth year 2002 of 15.6 percent.
Newborn Earned Income Tax Credit (EITC) claim rate transitions Source: Authors' calculations based on cross-sectional tax return data from 1995 to 2006. Note: calculations do not include the childless EITC.
Figures 1 through 3 indicate the broad reach of the EITC to taxpayers who file tax returns. However, if most eligible families are unable, or unwilling, to claim the credit, then arguably the EITC is a poor mechanism for either providing a safety net or for encouraging labor force attachment and upward mobility. Scholz (1994) estimates a participation rate of between 80 and 83 percent. More recently, Blumenthal, Erard, and Ho (2005) find that 89 percent of those taxpayers legally required to file a tax return and who are eligible for the EITC claimed the credit and that between 30 and 39 percent of families not legally required to file a tax return and eligible for the EITC claimed the EITC. Their combined participation rate for 1988 is between 69 percent and 74 percent. Dickert-Conlin, Fitzpatrick, and Hanson (2005) survey EITC participation studies and find that participation estimates for the EITC range from between 42 percent and 96 percent. In sum, participation research seems to indicate that a large portion of those that are eligible for the EITC do in fact claim it.
Description of Spell Length
The high degree of turnover in the EITC population indicated by Figures 1 through 3, suggests that taxpayers either claim the EITC for short periods of time or cycle on and off fairly frequently. Table 2 shows the length of EITC spells for taxpayers that we observe in 1989 and that have a child present at some point during the period 1989–2006. A spell is the time during which a family actually received the EITC. Calculations regarding spell length are based on Bane and Elwood (1983, 1994). 8 Column 1 shows the length of a spell in years. Columns 2 and 3 report statistics for families beginning a spell. Column 4 is for families at a point in time. Point in time estimates are a snapshot at a particular time, and as a result they sample more longer spells.
Distribution of EITC Spells
Source: Authors' calculations and 1989–2006 Individual Statistics of Income. Tabulations are conditional on observing the taxpayer in 1989 and having a child at some point in the 18-year period.
Note: EITC = Earned Income Tax Credit.
D(t) is the fraction of families who have EITC spells that last exactly t years. Forty-two percentage of spells last 1 year and 19 percent last 2 years for a total of 61 percent of spells completed in 2 years or less. Overall, the average spell length is roughly 3 years. This is slightly longer than Horowitz (2002) who found an average spell length for families newly claiming the EITC was 2.1346 years and 50.56 percent of spells lasted only 1 year. Using data covering the years 1989–2003 and not restricting the data to those with children, Dowd (2005) finds that 41 percent of recipients receive the credit for 1 or 2 years and 49 percent for 3 or fewer years.
In column 3, p(t) is the probability that a household in their tth year of EITC receipt ends their spell at the end of the year. p(t) is the exit probability. p(1) is the probability that a family in the first year of a spell does not continue the second year. In Table 2, D(1) is the 42 percent of households who received the EITC in a spell that lasted only 1 year, and p(1) is the 42 percent of households that received the EITC the first year but who do not claim the EITC the second year.
To calculate D(t), the fraction of families beginning a spell who eventually have a spell that lasts t years, calculate the fraction of families who last t − 1 years and multiply by the probability of exiting after t years. D(1) = p(1); D(2) = p(2)(1 − D(1)); D(3) = p(3)(1 − D(1) − D(2)); and
Column 4, F(t) shows the fraction of all households who receive the EITC at a point in time and who received the EITC for exactly t years. Equation 3 assumes a no growth steady state, so the number of families starting a spell each year is assumed to be constant. The numerator tD(t) is the fraction of families who have EITC spells that last exactly t years weighted by duration t. The denominator is the sum of all tD(t).
The average spell length for a family at a point in time is 5.3 years, which is approximately 2.25 years longer than the average spell duration for a family beginning a spell. About 63 percent of spells are 5 years or less.
Horowitz (2002) found an average spell length of 3.55 years for families at a point in time versus our data that find an average spell length of 5.3 years. The average spell is about a year and a half longer for families using administrative receipt data rather than survey eligibility data. Horowitz’s results for eligible families between 1975 and 1992 found somewhat more mobility than we find for our later period with tax data. The average spell length for families beginning a spell is 2.135 using the earlier eligibility data and 3 years using actual recipients. This is almost a full year longer. Some of this difference might be attributed to the different time period of analysis; Horowitz’s (2002) data were for the years 1975−1992 while the data in this study are for the years 1989−2006. The later sample years captures a period where the EITC was a much larger program with many more claimants and a period when welfare reform was encouraging labor force participation.
Table 2 indicates that almost a third of EITC claimants exit each year. This suggests that there is considerable movement of taxpayers onto and off the EITC. However, it does not tell us anything about whether these taxpayers claim the EITC again after a period of not claiming the credit.
Table 3 column 2 shows the reentry rate R(t) for persons exiting the EITC, for whom we observe filing a tax return in 1989 and have a child present at some point during the 18-year period. Column 1 shows the year since the exit and column 2 shows the probability that a family who has not received the EITC for t years will start a new spell. Approximately, 45 percent of those EITC recipients who did not receive the EITC for 1 year will claim the EITC again the next year. Approximately, 35 percent of those who did not receive the EITC for 2 years will receive it again the 3rd year. In other words, there is considerable churning, with taxpayers claiming the credit for short frequent spells.
Reentry for EITC Exits
Source: Authors calculations and 1989–2006 Individual Statistics of Income. Tabulations are conditional on observing the taxpayer in 1989 and having a child at some point in the 18-year period.
Note: EITC = Earned Income Tax Credit.
Column 3 shows the percentage of households who after the tth year reclaim the credit again. About 20 percent of households start another spell after 1 year, and about 8 percent of households start a spell after 2 years of not claiming the credit. Column 4 shows the cumulative percentage of households that reclaim the EITC. Forty-four percentage of households reclaim the EITC in the first 16 years, with 38 percent reclaiming the credit in the first 5 years.
Large expansions of the EITC increase the likelihood that families will remain in the EITC longer and those who do exit are more likely to reenter. In order to understand how policy and programmatic changes could affect claim rates, Table 4 examines what the completed spell distributions would look like if the 1989 programmatic details were held constant throughout the period 1989−2006. 11
Distribution of EITC Spells (1989 Parameters and Completed Spells Only)
Note: EITC = Earned Income Tax Credit. EITC benefits calculated using constant law 1989 parameters. Tabulations are conditional on observing the taxpayer in 1989 and having a child at some point in the 18-year period.
Using 1989 parameters and only including spells that were completed by 2006, the average spell length decreased only slightly from 3.0 years (Table 2, column 2) to 2.7 years (Table 4, column 2). Looking at the completed spell distribution, the average spell length decreased from 5.3 years in Table 2 to 4.8 in Table 4. Changing parameters and right-censored data do not seem to have had major effects on the results. Using a slightly different sample, Dowd (2005) also found that for persons with a child, the probability of claiming the credit was little affected by the changes to the EITC since 1989.
Table 5 looks at how reentry for the EITC might have been affected by the programmatic changes over time, by holding the program details constant at the 1989 levels for completed spells only. 12 The reentry rate decreases from 44 percent to 35.5 percent. Instead of 38 percent returning to the EITC after 5 years, 32 percent return to the EITC. These results are much closer to those of Horowitz (2002) for the earlier time period suggesting that the changes enacted in the 1990s have led to an increase in repeat usage.
Reentry for EITC Exits (1989 Parameters and Completed Spells Only)
Note: EITC = Earned Income Tax Credit. EITC benefits calculated using constant law 1989 parameters. Tabulations are conditional on observing the taxpayer in 1989 and having a child at some point in the 18-year period.
Families who are in the phase-out range of the EITC may have different spell lengths than families in the phase-in range or the maximum credit range. Table 6 shows the distributions of EITC spells based on whether the taxpayer was in the phase-in, maximum credit, or the phase-out ranges the first year of claiming the credit. Spell lengths are recorded for any EITC claim, regardless of which range they were in during subsequent years. According to Table 6, families beginning a spell and in the phase-out range on average experience shorter spells claiming the EITC. Families in the maximum credit range had average spells of 3.9 years, while families in the phase-in range had average spells of 3.2 years, and those in the phase-out range had average spells of 2.7 years. In sum, families are quite mobile and generally do not stay on the EITC for long periods, regardless of whether they initially claimed the credit in the phase-in or in some other programmatic range.
Distributions of EITC Spells in the Phase-in range, Phase-out, and Maximum Credit Ranges
Source: Authors' calculations and1989-2006 Individual Statistics of Income. Tabulations are conditional on observing taxpayer in 1989 and having a child at some point.
Note: EITC = Earned Income Tax Credit.
Income, Missing Data, and EITC Receipt
To be eligible for the EITC, taxpayers must have earned income and adjusted gross income below certain thresholds. Consequently, changes in claim rates are likely to be related to changes in income. Using CWHS data, Table 7 contains transition matrices for Adjusted Gross Income in 1990 and 2003. Column 1 presents the 1990 decile breakpoints. Column 2 indicates how many observations are found in each decile in 1990.
Transition Matrix for AGI Between 1990 and 2003
Source: Authors' calculations and1989–2003 Individual Statistics of Income.
Note: AGI = Adjusted Gross Income; EITC = Earned Income Tax Credit.
Twenty two percentage of the observations are missing in 2003 (Table 7, column 3 of matrix 1). The largest percentages of missing observations are in the lower income deciles. In 2003, 31 percent of the observations from the poorest decile are missing, while only 9 percent of the observations are missing from the richest decile. Some of the missing data could be the result of marital status changes. However, some of the missing observations are the result of taxpayers falling below the filing threshold and no longer required to file a return.
Of those taxpayers who file a return in both 1990 and 2003, 35 percent of the taxpayers move to a lower decile than the one they occupied in 1990 (column 4), 17 percent stay in the same decile and a quarter move to a higher decile. Kopczuk, Saez, and Song (2007) find that the probability of staying in the top and the bottom two quintiles after 1 year are on the order of 90 and 80 percent, respectively, and has remained relatively stable over much of the latter half of the twentieth century. They also report that after 10 years, the probability of moving from the bottom two quintiles to the upper quintile is roughly 10 percent. In comparison, Table 7 indicates that slightly more than 10 percent of those in the bottom two quintiles appear in the upper quintile.
Gottschalk (1997) finds that 42 percent of earners stayed in the bottom quintile, 36 percent stayed in the second quintile, 32 percent stayed in the third quintile, 32 percent stayed in the fourth quintile, and 54 percent stayed in the top quintile. Looking at matrix 1, we observe the same U-shaped pattern of increased persistence at the bottom and the top of the income distribution. Using AGI instead of earnings and using a 13-year period instead of their 17-year period, we estimate that 55 percent stay in bottom quintile, 46 percent stay in the second quintile, 44 percent stay in the third quintile, 55 percent stay in the fourth quintile, and 97 percent stay in the fifth quintile. 13
Matrix 2 of Table 7 only includes households who claim the EITC in 1990. As a consequence, the number of observations drops from 11,083 to 2,114. Not surprisingly, there are no observations with incomes in the top five deciles in 1990. Because lower-income households were less likely to file tax returns, missing data increased from 22 to 38 percent. Because taxpayers who claimed the EITC in 1990 often had missing data in later years or changed marital status, the percentages staying in the same deciles appear to fall relative to matrix 1. However, after adjusting for the missing observations, the number staying in the same decile is essentially identical at 23 percent (0.14/.62) for those claiming the EITC and 22 percent (0.17/0.78) for the entire population. Thirty-six percentage (0.22/0.62) of taxpayers that claimed the EITC in 1990 were in a higher-income decile, while 33 percent (0.26/0.78) of the entire population moved up. Households who received the EITC in 1990 and continued to file a tax return were more likely to have higher incomes than the population as a whole. However, these numbers include very low rates of movement into higher deciles at the upper portion of the income distribution. If we look only at the first four deciles, income growth for taxpayers who did not claim the EITC in 1990 was faster than for those claiming the EITC in 1990; 53 percent compared to 36 percent for taxpayers who claimed the EITC in 1990. 14
In 1990, the beginning of the phase-out of the EITC occurred with income starting at $10,730 and ended with an income of $20,254. Thus, the third decile in 1990 was solidly in the phase-out portion of the EITC. In 2003, the beginning of the phase-out for taxpayers with qualifying children was at $13,730, and the ending income was $33,692. In 2003, the third decile for AGI is between $22,838 and $31,631, again solidly in the phase-out of the EITC. About 44 percent (598/1108) of households in the third decile were also EITC recipients. Comparing the transitions for the third decile for all taxpayers (matrix 1) with those claiming the EITC (matrix 2), households who were in decile 3 and received the EITC in 1990 were more likely to be missing in 2003 and to have slipped into a lower decile. Moreover, these households were much less likely to have moved up to a higher income decile.
Matrix 3 shows the AGI transitions for taxpayers that claim the EITC in 1990, but who did not claim the EITC in 1989, that is, “first-time” claimants. Taxpayers who claim the EITC for the first time in 1990 and are in the third decile are less likely than all taxpayers that claimed the EITC in 1990 to be missing in 2003 or to move to a lower decile. However, they are more likely to either remain in the same decile or move to a higher-income decile.
These three matrices indicate that there is less income mobility for the EITC population than for the population as a whole and that the EITC population is less likely to file a tax return in subsequent years. 15 Matrix 3 indicates that “first-time” claimants in 1990 had faster income growth.
Is there a pattern to income growth that is related to the number of times that a taxpayer claims the EITC? Figure 4 shows the average real wages for first-time EITC claimants in the year prior to claiming the credit and thereafter, excluding missing years in our calculation of average real wages, by whether the taxpayers claimed the credit one to two times, three to five times, or six or more times. For all three groups, wages drop in the first year of claiming the credit and then increase after the first year. Not surprisingly, the largest increases in wages are associated with taxpayers that claim the credit only once or twice (representing 41 percent of the families claiming the EITC for the first time), and the smallest increases in wages are associated with taxpayers that claim the credit six or more times. Moreover, families claiming the credit more than twice but less than six times (28 percent of the families claiming the EITC for the first time) wages appear to approach the wages of those who claim the credit only once or twice. Perhaps most interesting is that the bulk of the wage growth differential between the three groups occurs in the year after first claiming the credit. This suggests that quite a bit of the wage growth is the result of short-term shocks for the infrequent users of the EITC. 16
Average real wages one year prior to first claim year and thereafter Source: Authors' calculation based on Continuous Work History Sample (CWHS) sample of tax returns from 1989 to 2006, conditional on taxpayer having a dependent child sometime between 1989 and 2006. Note: Missing observations are not included in the calculation of average wages.
One reason that we might see wages decline in the first year of claiming the EITC is the birth of a newborn child. The birth of a newborn not only potentially creates eligibility for the EITC; it also is often associated with lapses in employment and a decline in wages. Figure 5 shows average real wages for tax returns that claim the EITC in the same year as the birth of a child. Like Figure 4, the results exclude missing years from the calculation of average wages. As expected, wages initially drop to a low in the birth year and then grow thereafter. Real wage income growth from 1996 to 2006 for families claiming the EITC in 1995 and having a newborn child in 1995 was roughly 5.9 percent annually. Using cross-sectional survey data, the Congressional Budget Office (CBO; 2007) found that real income growth for low-income households with children increased by 35 percent between 1991 and 2005. Moreover, CBO also found that real incomes increased by about 45 percent for low-income households with children. We show real wage growth was 27 percent between 2001 and 2003 for those taxpayers who both received the EITC in 2001 and had a newborn child in 2001. For each of the other family birth cohorts, the real wage growth was less.
Average wages for Earned Income Tax Credit (EITC) claimants with newborn child Source: Authors' calculation based on Continuous Work History Sample of tax returns from 1989 to 2006, conditional on taxpayer having a dependent child sometime between 1989 and 2006. Note: Missing observations are not included in the calculation of average wages.
How Spell Length is Affected by Family and Other Determinates
The previous analysis has been strictly univariate. To understand more clearly how spell length is affected by family and other determinates we calculate a multiple regression with spell length as the dependent variable. Equation 4 shows the structure of the regressions.
Variable and Spell Definitions
1Southern states include: Alabama, Arkansas, Florida, Kentucky, Georgia, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, Tennessee, Texas, Virginia and West Virginia.
Table 9 includes descriptive statistics for family and other potential determinates of spell length. The results in columns 2, 5, 8, and 11 are for spells that are completed by 2006. About 77 percent of families completed their spell by 2006. Including only completed spells introduces a bias toward shorter spells. We also include summary statistics for those who did not complete their spells in 2006 (columns 3, 6, 9, and 12) and for all spells completed or uncompleted (columns 4, 7, 10, and 13). We restrict the sample to observations where the primary taxpayer’s age at the beginning of the spell is over the age of 19. Teen primary taxpayers with children are likely to have very different economic and family dynamics than adults. As with the analysis presented in the prior tables and figures, we restrict the sample to taxpayers who have children. In particular, each spell is defined as a spell of claiming the EITC with children in the household. For variables that indicate a change over time, we assume no change for those taxpayers who do not file a tax return.
Descriptive Statistics for Spells 1 through 3 and Pooled Spells
Of the 14,791 spells that occurred between 1989 and 2006, 62 percent were in the first spell that we observed. 17 Twenty-nine percentage of the spells were in the second spell, and 9 percent were in the third spell. We restricted our analysis to the first three spells.
For completed spells, spell length is between 2.2 and 2.6 years, depending on whether it is the first, second, or third spell. Spells that are not completed by 2006 are between 1.2 and 2.4 years longer. For all spells, spell length is 3.04. This is just slightly more than the spell length of 2.985 in Table 2. 18
The average age of the primary taxpayer is 35.4 years when pooling all of the spells together. Naturally, the age of the primary taxpayer increases as there are more spells. Families starting a second and third spell were a little older, and the age of the primary taxpayer was younger for those in uncompleted spells. Seventy percentage of the primary taxpayers in families that receive the EITC are in their twenties or thirties. Only about 11 percent of primary taxpayers are in their fifties and sixties. Primary taxpayers starting their first spell also tend to be younger than those starting their second or third spells. Forty-five percentage start their first spell when the primary taxpayer is in their twenties compared to 28 percent of families starting their second spell and 13 percent of families starting their third spell.
Female heads of household are more likely to receive the EITC than married heads of household. Female heads are also relatively more likely to be in the first spell (43 percent) than the second spell (40 percent) or the third spell (35 percent). On the other hand, married heads are slightly more likely to have more spells.
The average number of children increases from 1.36 children in the first spell to 1.66 in the third spell. This may be because EITC benefits are more generous with more children.
In 2006, 37 percent of households lived in the South, and the South had lower median household income ($43,884) than the rest of the country ($48,201). 19 A somewhat larger percentage of households from the South are represented in the EITC claiming population, with 43 percent residing in the south. There is a slight increase in the percentage in the south going from 41 percent of first spell claimants to 46 percent of third spell claimants.
Thirty-one percentage of EITC claimants are married at the start of their spell, with a slight increase in the proportion married as the number of spells increases. About 6 percent (0.04/0.69) of EITC claimants who are not married at the start of the spell are married after 3 years. Divorce is much more common than getting married; 25 percent (0.08/0.31) of primary taxpayers who started their first spell married are divorced after 3 years.
The mean real wage in 2006 was $14,810 for families starting their first completed spell, $16,740 for the second completed spell, and $17,130 for the third completed spell. The maximum credit range for joint filers was between $11,340 and $16,180 and between $11,340 and $14,810 for nonjoint filers in 2006. This means that the $14,810 for families starting their first spell would be in the maximum credit range for joint filers and in the phase-out range for nonjoint filers. In each case, the average taxpayer is close to where the maximum credit range and the phase-out range meet. As a result, families with small increases in real wages would still be eligible to receive the EITC so increases in real wages probably would not have much effect on spell length.
Fourteen in row 19 means that the average uncompleted spell started in 2002. For families who completed their spells by 2006, the average family started their first spell in 1996, their second spell in 1998, and their third spell in 2000.
Thirty-eight percentage of completed spells end when the primary taxpayer does not file a return. First spells are slightly more likely (39 percent) to end because the taxpayer did not file a return at the end of the spell. Families with more spells tend to be older, be married, have more children, have higher real incomes, and obviously file tax returns relative to the average single spell.
Table 10 shows multiple regression results with spell length as the dependent variable. Since we have included an intercept, each of the other dummy parameters is relative to the excluded dummy group of 30-year-old male primary taxpayers starting their first spell in 1990 (starting their second spell in 1991 and starting their third spell in 1993).
Regressions for Spells 1 through 3 and Pooled Spells
Note: ** and * indicate that the parameter is significant at the 1 and 5 percent level, respectively. Numbers in parenthesis are standard errors.
Older families have shorter spells. Looking at the pooled results, families where the primary taxpayers are in their forties and fifties have spells that are about 3 and 4 months shorter and primary taxpayers in their sixties have spells that are approximately 9 months shorter. This may be because, as shown in Figure 5, families with older children tend to have higher real wages. The shorter spells of the older families suggest that for these families, the EITC acts more as a short-term safety net.
In contrast, younger families and those with a female head tend to have longer spells. The pooled and spell 1 results indicate that female heads have spells that are approximately half a year longer. Families with more children have slightly longer spells; an additional child at the beginning of the spell will add between 2 and 4 months to completed spells and between 0 and 2 months to uncompleted spells. An increase in the number of dependent children after 3 years increases spell length by approximately 3 and 4 months. Living in the South increases spell length by about 2 months. These slightly longer spells suggest that for these families, the EITC is more likely to act as long-term income support.
Marital status at the start of the spell is not statistically significant. Moreover, getting married after 3 years is only statistically significant in the completed pooled results, where it reduces spell length by about 2 months. However, getting divorced after 3 years decreases spell length for the first completed spell by 7 months and by almost 9 months for the second completed spell. Getting divorced was not statistically significant for spell 3 and for the uncompleted spells. Again, we need to be careful in interpreting these results about marriage and divorce because divorce results in many dropped female observations.
A $1,000 increase in real wages reduces spell length by less than a month. As mentioned above, a $1,000 increase in the real wage probably does not have much effect because the average taxpayer is in the phase-out range near the maximum credit range. According to Table 10, the average taxpayer with a completed spell earned $15,540. The zero credit income in 2006 for a taxpayer with one qualifying child is $32,001. Thus, to be phased out of the EITC, a taxpayer must earn $16,461 more than the average taxpayer with a completed spell. If she had two qualifying children, she must earn even more to become phased out of the EITC. Not filing a tax return reduces spell length by about 4 months in the first spell, 2 months in the second spell, and is not statistically significant in the third spell.
Summary and Conclusion
Transitions into and out of the EITC are frequent; 61 percent of families beginning a spell are likely to have completed the spell after 2 years. At any point in time, 57 percent of families have been on the EITC for 5 or fewer years, with 39 percent having received the EITC for 3 or fewer years. The average spell length for a family beginning a spell on the EITC is 3.0 years, but the average spell length for families on the EITC at any particular point in time is almost 5.3 years. Most recipients of the EITC claim the EITC for short periods (42 percent for only 1 year at a time).
The combination of high usage by tax filers with children, short spells of usage and high turnover suggest that the EITC has very broad reach for the filing population. About 50 percent of taxpayers who have a child at some point over the 18-year period claim the EITC at some point. We find considerable movement from one portion of the income distribution to another for the overall population of taxpayers, but those who claimed the EITC in 1990 are less likely to move up the income distribution than the overall population. 20
The percentage of households that return to claiming the EITC decreases as the number of years off of the EITC increases. Twenty percentage reclaim the EITC after 1 year, while only 2.3 percent reclaim after 5 years. Nevertheless, we found that there is considerable reentry, with 44 percent reclaiming the EITC. Even though the EITC significantly expanded in the 1990s, we found that if families had faced an unchanged EITC program since 1989, there would be little difference in the length of time on the EITC. However, as a result of this increase in the eligible population, there would be a significant increase in reentry.
Younger families headed by women, with more children at the beginning of the spell tend to have longer spells than older male headed families. Being married has little impact on spell length. In contrast, getting divorced has a significant effect on spell length. This last result may be the consequence of the sample design that disproportionately drops women who get divorced from the sample.
In conclusion, the EITC acts as a short-term safety net to many taxpayers who claim the EITC for short periods during shocks to income or family structure. At the same time, the EITC acts as a long-term income support for a significant portion of EITC recipients who claim the credit for long periods of time, and for multiple spells, cycling on and off.
Footnotes
The views in this article are those of the authors and should not be attributed to the staff of the Joint Committee on Taxation, any Member of Congress, or Ball State University. We wish to thank Stacey Dickert-Conlin, Lori Stuntz, Jeff Larrimore, anonymous referees, and participants at seminars at the National Tax Association meetings, Ball State University, Purdue University, and the Tax Economists Forum in Washington DC for their helpful comments. All remaining errors are our own.
The author(s) declared no potential conflicts of interests with respect to the authorship and/or publication of this article.
The author(s) received no financial support for the research and/or authorship of this article.