A Profile of Social Security Child Beneficiaries and their Families: Sociodemographic and Economic Characteristics

Introduction

The US Social Security system provides benefits to children of insured workers who die, become disabled, or retire. At the end of 2008, approximately 3.1 million children under age 18 received monthly benefits from Social Security, representing about 6 percent of Social Security beneficiaries (SSA 2010a, Table 5.A1.4).

In-depth analysis of Social Security child beneficiaries is surprisingly limited. Data published by the Social Security Administration (SSA) provide useful aggregate information, but offer little insight into the characteristics of children receiving benefits and the importance of these benefits to the financial security of their families. In this study, we use a restricted-use dataset that matches respondents in the Census Bureau's Survey of Income and Program Participation (SIPP) calendar-year 2004 file with Social Security (Old-Age, Survivors, and Disability Insurance, or OASDI) benefit records. These matched data allow more accurate identification and measurement of Social Security child beneficiaries and their benefit types, benefit amounts, and family income than survey data alone.

Our primary objective is to provide a portrait of the sociodemographic and economic characteristics of recent Social Security child beneficiaries. Newcomb (2003/2004) analyzes child beneficiaries using a matched 1996 SIPP-SSA file; however, additional analysis using more recent data is needed, given potential demographic and economic changes in the population. In addition, the current study adds new empirical insight by disaggregating child benefit income from total (adult and child) benefit income within a family unit. This study also uncovers important heterogeneity in outcomes across the three main child benefit types.

The results yield a number of important insights. Although Social Security rules treat persons with identical earnings histories and life events (for example, the death of a parent) uniformly, we find that the likelihood that a child will receive a benefit, and the amount of his or her benefit, vary widely across different population segments. The results also advance our understanding of the economic status of child beneficiaries in the OASDI program. In general, child beneficiaries have lower economic status than all children, in part because the qualifying event (death, disability, or retirement) likely relates to diminished capacity to generate earnings and other forms of income. Social Security benefits, therefore, constitute a substantial portion of the family income of many child beneficiary families, even when only child benefits are counted. Although attention often centers on how Social Security reforms might affect the economic status of aged adults, policy changes may also influence child recipients because child benefit computations are based on the parent's benefit. Children constitute an important and often economically vulnerable share of OASDI beneficiaries.

The next section provides a brief background on Social Security child benefits. A description of our data and analytic approach follows. Then, we present our empirical results. The final section summarizes our findings and suggests further avenues of research.

Social Security Child Benefits

Just 4 years after the passage of the Social Security Act, the 1939 Social Security Amendments established child benefits and began to change Social Security from a retirement program for workers to an economic security program for the whole family (DeWitt, Béland, and Berkowitz 2008; Martin and Weaver 2005). Under the current program, monthly benefits are payable to children of insured workers who have either died, become disabled, or retired. This article focuses on minor children; that is, children under the age of 18. To be eligible, a child must be the biological or adopted child of an insured worker or, in some cases, a dependent stepchild or grandchild.¹ In addition, a child generally must be unmarried to receive child benefits.² Child benefits are additional to payments to the disabled or retired workers themselves.

Table 1 outlines the eligibility requirements for Social Security child benefits. Three types of Social Security benefits are available to qualifying persons under age 18. The first of these is for children of deceased workers. To qualify, the child's deceased parent must have earned at least one of the following: (a) 40 quarters of coverage throughout his or her lifetime, (b) 1 quarter of coverage for every year between age 21 and death, or (c) 6 quarters of coverage over the 13 calendar quarters prior to death.³ The benefit amount for the surviving child equals up to 75 percent of the base amount of the parent's benefit, called the primary insurance amount (PIA). The second type of benefit is for children of disabled workers. To be entitled, the child's parent must be receiving Social Security disability benefits. Program rules limit the benefit for children of a disabled parent/guardian to 50 percent of the disabled worker's PIA. The third benefit is for children of retired workers who are fully insured (have earned 40 quarters of coverage) and are receiving Social Security benefits. Program rules also limit the benefit for a child of a retired worker to 50 percent of the parent's PIA.

Child benefits are subject to a family maximum, which limits the total monthly amount payable from the primary beneficiary's earnings record. The family maximum for retirement and survivor benefits ranges from 150 percent to 188 percent of the worker's benefit, and the family maximum for disability benefits ranges from 100 percent to 150 percent of the worker's benefit. If the total amount payable to the primary and auxiliary recipients exceeds the family maximum, then child benefits, along with any spouse or widow benefits, are reduced proportionally.⁴ For example, if a disabled worker's PIA is $1,000 and he has a spouse and 1 child who are each eligible for $500 (50 percent of the workers' PIA), then their total benefits plus his worker benefit would equal $2,000 (200 percent of the PIA). If his family maximum were 150 percent of his benefit, however, his spouse and child would each receive only $250 (25 percent). Although it is not as common, child benefits can be reduced by the earnings test if the child earns more than the allowed amount or if the parent providing the child benefit is a retired beneficiary who is younger than the full retirement age and earns more than the allowed amount.

Data and Methods

We use data from the 2004 SIPP calendar-year longitudinal file matched one-to-one with SSA's benefit record for that year.⁵ The SIPP is a large panel survey of a nationally representative sample of the civilian noninstitutionalized US resident population conducted by the Census Bureau. The survey collects detailed information on an array of topics including demographics, income, labor force participation, and government program participation. Respondents are interviewed in staggered 4-month cycles called waves, and individual SIPP panels last between 2 ½ years and 4 years (Census Bureau 2006).

The SIPP calendar-year 2004 file combines monthly data for several waves, allowing us to sum monthly observations to derive annual estimates.⁶ This technique likely provides more accurate measures of family income and poverty than annual recall data from a single interview (Bound and Krueger 1991).

Data on Social Security benefits come primarily from SSA's Master Beneficiary Record, with some additional data from the Supplemental Security Record.⁷ The Master Beneficiary Record provides data on monthly OASDI benefits throughout 2004. The Supplemental Security Record provides similar information on Supplemental Security Income (SSI) payments. Linked administrative data provide the information needed to determine the precise types and amounts of child benefits received. Household surveys such as SIPP and the Current Population Survey provide large samples and rich demographic detail at the microdata level, but do not allow for the precise identification of children receiving benefits or the reasons for benefit receipt.

One limitation of the matched dataset is that not all SIPP respondents can be linked to their SSA benefit records.⁸ The unweighted match rate for the calendar-year 2004 SIPP file is approximately 82 percent (84 percent for individuals aged 18 or older and 77 percent for individuals under age 18). If the propensity to be matched varies in a systematic, nonrandom way, it could introduce bias into the estimates. To address this issue, we treat nonmatched SIPP respondents as a unit nonresponse (Nicholas and Wiseman 2009; Holt and Elliot 1991). Preliminary analyses found that the respondent child's race and the family head's education and income significantly predicted whether data for individuals under age 18 would match. We perform a logistic regression to estimate the effect of the aforementioned characteristics on the logarithmic odds of being matched. We then use these estimates to adjust the sample weights (Groves and Couper 1998; Carlson and Williams 2001). The appendix provides more detailed information on this procedure.

We restrict our sample to all children under the age of 18 (and their families) in the matched SIPP-SSA calendar-year 2004 file. Our unit of analysis is each child, not the child's family. The sample of Social Security child beneficiaries consists of individuals under age 18 in December 2004 who received a child benefit every month of the calendar year. We exclude individuals who did not receive a child benefit in every month in 2004 to gauge the program's economic impact over a calendar year rather than at one point in time.⁹ This method avoids conflating children who receive low benefits because their parents had low earnings and those who receive low annual amounts because they only receive benefits for part of the year. With these restrictions, we identify 778 child beneficiaries from the 17,152 individuals under age 18 (unweighted) in the SIPP-SSA matched dataset. Using SIPP methodology we define families as persons living in the same household and related by blood, marriage, or adoption.¹⁰ If only one parent or guardian is present in the family, we consider that person the family head; if two parents or guardians are present, we consider the individual with higher personal annual earnings the family head.

Information on sociodemographic and economic characteristics is from the SIPP. We use the administrative data to identify benefit receipt, amounts, and benefit type (child of deceased worker, child of disabled worker, and child of retired worker). Prior research has shown that survey respondents, particularly those in families with child recipients, underreport OASDI and SSI income (Huynh, Rupp, and Sears 2002; Nicholas and Wiseman 2009). To account for this, we replace matched respondents' self-reported OASDI and SSI income with their SSA-reported income. This technique leads to higher quality estimates of both Social Security and total family income.

Our analysis uses descriptive methods to profile the Social Security child beneficiary population. We examine the pattern of child benefit receipt across key demographic and socioeconomic subgroups at the individual and family levels. To add perspective, we include comparative estimates for the overall child population. In addition, to ascertain how the characteristics of child beneficiaries may have changed in recent years, we compare our findings with estimates based on the 1996 SIPP, as reported by Newcomb (2003/2004). A final point of the analysis considers the ratio of Social Security income to total income in the families of child beneficiaries. Unlike previous work, we disaggregate a family's reliance on OASDI child benefits alone, which reflect payments to all children in the family, from their reliance on total OASDI benefits, which reflect payments to all adults and children in the family.

Our analysis does not capture all of the ways in which Social Security may affect children. Children may receive Social Security as students between ages 18 and 19 or as disabled adult children. We omit these groups to focus on minor children, the predominant share of child beneficiaries. The analysis also does not assess children who do not receive Social Security benefits directly, but nevertheless live in a family in which at least one member receives Social Security. Children who also qualify for other forms of assistance, such as SSI, are also outside this study's scope.

All statistics reported herein apply the adjusted sample weights. Estimates based on fewer than 30 unweighted observations would raise disclosure and statistical reliability concerns and are not reported. Because of SIPP's complex sample design, we computed standard errors (available upon request) using STATA's survey command procedure, which takes into account both stratification and clustering within the sample. Unless otherwise noted, all declarations of difference between estimates underwent statistical tests and were found significantly different at a 90-percent level.¹¹

Results

Our findings are described below, with separate sections addressing the sociodemographic and the economic characteristics of child beneficiaries.

Sociodemographic Characteristics of Child Beneficiaries

Table 2 compares child beneficiary information from our matched SIPP-SSA file with SSA administrative counts published in the 2004 Annual Statistical Supplement to the Social Security Bulletin. A comparison of these two data sources verifies that our sample is reasonably representative of the overall child beneficiary population. The weighted count of Social Security minor child beneficiaries in 2004 in the matched SIPP-SSA sample was about 3.0 million, a figure similar to the published administrative total. Of these, 1.2 million (40.2 percent) were children of deceased workers, 1.5 million (50.5 percent) were children of disabled workers, and 278,972 (9.3 percent) were children of retired workers. Published SSA administrative totals based on a 10-percent sample of the Master Beneficiary Record indicate a slightly higher percentage of children of deceased workers and a slightly lower percentage of children of disabled workers; average monthly benefit amounts, however, are similar for all child beneficiaries and for each benefit type.¹² Modest differences are expected because of differences in the two populations and the time frames of the data.¹³ Overall, the relative trends in data from both sources are analogous, suggesting that the matched SIPP-SSA sample accurately approximates the child beneficiary population in 2004.

Table 3 documents the incidence of child benefit receipt and the average amount of child benefits, and compares child beneficiaries with all children, by selected sociodemographic characteristics. Table 4 shows the distributions and average benefit amounts for each benefit type by these characteristics. In total, we find 4.4 percent of US children received Social Security child benefits throughout 2004.¹⁴ The average individual monthly benefit in 2004 was $252 for children of disabled workers, $426 for children of retired workers, and $600 for children of deceased workers; the overall average was $408. Children of disabled workers may have had lower benefits because their parents had less time to accumulate earnings and because the family maximum limits are lower for disability benefits (100–150 percent) than for retirement or survivor benefits (150–188 percent). Surviving children receive larger benefits in part because the child survivor benefit is more generous (up to 75 percent of insured worker's PIA) than those of the other benefit types (up to 50 percent).

As Tables 3 and 4 show, the sociodemographic characteristics of the child, the family head, and the entire family associate with different probabilities of child benefit receipt and average benefit amounts for entitled children.

Characteristics of child. Table 3 shows important associations between the race and age of the child and both the likelihood of benefit receipt and the average benefit amount. A larger share of black children received a child benefit in 2004 (5.8 percent) than did white children (4.3 percent) and Hispanic children (3.6 percent). Table 4 shows that black children were also disproportionately likely to receive all types of benefits, particularly children of a deceased or retired worker. This could be because black individuals, as a group, exhibit relatively high mortality and disability rates (Bound, Schoenbaum, and Waidmann 1995; DeCesaro and Hemmeter 2008; Dunlop and others 2007; Garrett 1995).

Because the probability of parental death, disability, or retirement cumulates over time, older children were more likely than younger children to receive a child benefit in 2004. In addition, older children received higher average benefits than their younger counterparts (Table 3), which could reflect more extensive earnings histories among their parents.

Characteristics of family head. The education and marital status of the family head associate with child benefit receipt and amount. Children whose family head lacked a college degree were more than twice as likely to receive benefits as were children whose parents graduated from college (Table 3). Further, among those child beneficiaries living in families headed by a person with a college degree, the average amount was higher ($579) than that for children of high school graduates ($389) and for children of parents who did not finish high school ($329). Compared with 1996, child beneficiaries in 2004 were more likely to have a family head with at least a high school diploma (Newcomb 2003/2004, Table 2).

The educational attainment of the family head also varied according to the child benefit type (Table 4). Children of disabled workers had the lowest share of college-educated family heads, which likely coincides with lower disability incidence among higher-educated adults (Bound, Schoenbaum, and Waidmann 1995). Children of retired workers had relatively higher proportions of family heads who were high school dropouts or who were college graduates, reflecting a somewhat bimodal distribution between low and high socioeconomic status. However, because of the small sample size, the standard error for educational attainment of the family head of children of retired workers is quite large; therefore, we urge caution when comparing children of retired workers by family head's education.

We also find important variation by the marital status of the family head (Table 3). Compared with all children, child beneficiaries were more likely to have a divorced or separated family head (26.1 percent versus 15.0 percent). This pattern could reflect, in part, higher divorce rates in the disabled population (Martin and Davies 2003/2004; Singleton 2009). Not surprisingly, given the survivor component of child benefit entitlement, a substantial share of child beneficiaries lived in families headed by a widow or widower (14.4 percent).

Characteristics of family. In 2004, 34.1 percent of child beneficiaries lived in single-child families, compared with 22.0 percent of all children (Table 3). Likewise, 9.0 percent of child beneficiaries lived in families with 4 or more children, versus 14.3 percent of all children. This trend may occur because the life events through which children become eligible for benefits (parent's death, disability, or retirement) eliminate or reduce the likelihood of further childbearing.

Families with only 1 or 2 child beneficiaries received higher average individual child benefits ($480 and $405, respectively) than families with 3 or 4 (or more) child beneficiaries ($260 and $157, respectively). These disparities may capture the effects of the family maximum, which is more likely to affect families with multiple Social Security child beneficiaries, and to affect them more deeply.

Interestingly, the majority of child beneficiaries in 2004 (53.6 percent) lived in families without an adult beneficiary (Table 3). Table 4 shows that children of disabled and retired workers were far more likely to live with an adult beneficiary than were children of deceased workers. The absence of an adult beneficiary in a child beneficiary's family can occur for a number of reasons. For example, a child beneficiary could live in a different household from the beneficiary parent; or, a survivor child beneficiary could live with a surviving parent who is not eligible for benefits.¹⁵

Economic Characteristics of Child Beneficiaries

Table 5 compares the economic characteristics of child beneficiaries with those of the overall child population. The families of child beneficiaries, as a whole and by benefit type, have substantially lower incomes than those of all children.

Table 5 indicates that the family income disparities between child beneficiaries and all children are driven in part by reduced labor earnings in child beneficiary families. By all measures, the personal earnings of heads of families with children receiving Social Security were sharply lower than the earnings of the heads of all families with children. For example, at the 25^th percentile, personal earnings for the family heads of child beneficiaries were zero, compared with $15,300 for the family heads of all children. In fact, 19.4 percent of child beneficiaries had no family members with any labor earnings in 2004, compared with only 4.3 percent of the overall child population. Because this study does not include the earnings records of workers prior to their receipt of Social Security benefits, it is difficult to pinpoint the cause of reduced labor earnings for child beneficiary families. Lower earnings could relate to the income shock that child beneficiaries' families experienced when a parent retired, died, or became disabled. Conversely, low-income workers may be more likely to experience a qualifying life event, so their children are more likely to become child beneficiaries. Additionally, a higher rate of SSI receipt among families of child beneficiaries (12.9 percent) than the families of all children (5.9 percent) suggests relatively low assets and low income, as does child beneficiary families' Food Stamp receipt rate (26.3 percent) relative to that for all children (18.1 percent).

Table 6 presents the annualized family incomes of child beneficiaries and all children relative to the federal poverty threshold. Somewhat similar shares of child beneficiaries (16.7 percent) and all children (15.5 percent) lived in poverty in 2004.¹⁶ However, the economic status of Social Security child beneficiaries was generally lower than that of all children. This is evident in the lower mean and median incomes relative to the poverty threshold. It is also indicated by the prevalence of near-poverty (income between 100 and 150 percent of the threshold), where the share of child beneficiaries (17.6 percent) exceeds that of children overall (12.1 percent). As further evidence, family income at the upper end of the distribution—that is, at least 300 percent of the poverty threshold—was less prevalent among beneficiary children. Nevertheless, child beneficiaries were more protected from deep poverty. Only 3.9 percent of child beneficiaries had family income below 50 percent of the poverty threshold, compared with 5.6 percent of all children.

Economic characteristics also vary within the child beneficiary population. The poverty rate is greatest among children of disabled workers. Poor health conditions, among other factors, make it difficult for disabled workers to engage in labor market activity. This low labor force participation, along with the stricter family maximum on disability benefits, results in lower Social Security benefits.

For children of retired workers, family income trends appear somewhat bimodal—relatively low at the 25^th percentile, but relatively high at the 75^th percentile (Table 5). Moreover, a relatively large segment of the children of retired workers was decidedly poor (4.1 percent were in families with income below 50 percent of the poverty threshold), but at the same time, a much larger segment was decidedly well off (41.8 percent in families with income at least 300 percent of the poverty threshold) (Table 6). These findings are consistent with Table 4, which shows that the family head of children of retired workers often had relatively low or high (as opposed to midrange) educational attainment.

Children of deceased workers experienced the lowest poverty rate (14.8 percent) (Table 6). This could be in part because children of deceased workers live in families whose head has higher levels of personal earnings, as shown in Table 5. In addition, children of deceased workers receive higher benefit amounts than other child beneficiaries, further protecting them from poverty.

Table 7 reports the share of family income derived from Social Security for the 2004 child beneficiaries. Social Security child benefits constitute an important part of the income of child beneficiary families. For example, child benefit payments alone comprised an average of 23.4 percent of total family income.¹⁷ For 67.1 percent of beneficiaries, child benefits constituted less than 25 percent of family income; but for 11.1 percent, they accounted for 50 percent or more of family income. This shows that child benefits alone constitute a substantial portion of family income for a sizable segment of child beneficiaries.

Not surprisingly, families are more reliant on total (adult and child) Social Security benefits: An average of 37.4 percent of the income of child beneficiary families originates from Social Security.¹⁸ For 28.3 percent of child beneficiaries, Social Security comprised more than half of family income. The greater reliance on Social Security income when also counting adult benefits, rather than child benefits alone, underscores the fact that 46.5 percent of child beneficiaries lived with an adult beneficiary.

Among beneficiary types, child benefits comprised a greater average source of total family income for children of deceased workers (30.2 percent) than for children of disabled workers (18.1 percent) and retired workers (22.0 percent). Almost 17 percent of children of deceased workers relied on child benefits for 50 percent or more of their family income, while only about 7 percent of families with children of disabled or retired workers did so. Survivor child benefits are more generous (75 percent of the worker's PIA) than those of the other benefit types (50 percent of the worker's PIA), partially explaining the greater reliance on child benefit income among surviving-child families.

The families of children of retired workers receive the highest proportion of their income from total Social Security benefits. For 12.5 percent of these child beneficiaries, Social Security accounted for almost all of their annual family income (95 percent or more). Only 7.0 percent of the families with children of deceased workers and 5.7 percent of the families with children of disabled workers reported such a high reliance on Social Security.

Concluding Remarks

Using rich data that matches SIPP respondents with their SSA administrative records, this article provides empirical insights into the Social Security child beneficiary population. We find noteworthy variations in the incidence of child benefit receipt across different population segments. The analysis also sheds light on the economic characteristics of child beneficiaries. An important overall finding is that Social Security child benefits, although not targeted toward low-income families, provide income maintenance for many such families, in part because the conditions that give rise to child benefit eligibility—death, disability, and retirement—often lead to family income loss.

To summarize the relevant findings: First, child beneficiaries in 2004 differed from the overall child population in several demographic and economic status measures. Social Security child benefits were particularly important to black children. Children with unmarried family heads represented a disproportionate share of child beneficiary recipients. Compared with all children, child beneficiaries had relatively similar poverty rates but lower economic status in general, marked by a higher share with family incomes at 100–149 percent of the poverty line and lower family income and earnings levels. Families with child beneficiaries also had higher SSI and Food Stamp receipt rates than the families of all children. Conversely, deep poverty (family income below 50 percent of the poverty threshold) was more prevalent among all children than among child beneficiaries.

Second, important demographic and economic differences are also evident among child beneficiaries across benefit types. Children of disabled workers live in families headed by individuals least likely to have higher educational attainment. Accordingly, family incomes are somewhat lower than those of the other beneficiary types. Children of retired workers exhibit bimodal distributions in terms of parental education, income, and poverty outcomes. Although some children in this category are well off, a substantial segment exhibits financial vulnerability. Children of deceased workers, the subgroup that receives the highest average amount from child benefits, exhibit the lowest poverty rates of the beneficiary types.

Third, Social Security is a major source of family income for many child beneficiaries. The family income for child beneficiaries is lower than that for the general population, and child beneficiaries would be far worse off without the Social Security benefits. Direct OASDI child payments constituted an average of 23.4 percent of total annual family income of child beneficiaries. For children of deceased workers, child benefits alone accounted for an average of 30.2 percent of family income. For families in which a child beneficiary lives with a parent or guardian who also receives benefits, the reliance on Social Security income is obviously even greater.

Our results raise a number of policy implications. The data show major differences in the economic well-being among the three child beneficiary types. The economic challenges for children of disabled workers are particularly notable. One way to address differences across beneficiary types would be to reconsider the formula underlying child benefit payments (percentage of workers' PIA), which is lower for children of disabled or retired workers than for children of deceased workers. Another option to consider is adjusting the family maximum, which is lower for disabled worker benefits than for retirement and survivor benefits. Other proposals may include a minimum benefit for child beneficiaries in very low-income families; or determining child benefit payments using percentages of a worker's PIA based on family income, rather than using a flat percentage based solely on the qualifying life event.

A promising avenue for future work would be to examine the factors that influence the relative economic importance of Social Security to child beneficiaries. One approach to examining these factors would be to complement our empirical work here with multivariate regression analysis. An open question is why some children who received OASDI remain in poverty and others do not. One important variable not fully explored in this article is family structure. Another fruitful line of inquiry may be to explore the extent to which changing demographics, including higher adult disability rates or later childbearing, may affect future Social Security child beneficiary receipt. Choosing the best policy options for Social Security's child programs and understanding their distributional consequences requires further study.

Appendix

We use a logistic regression to determine which factors predict whether SIPP responses for individuals under age 18 match with SSA administrative records. After preliminary analyses, we found that child's race, family head's education, and family income were significant match predictors for individuals under age 18. Accordingly, we use the following model for the logistic regression:

where p equals the probability of a given SIPP respondent being present in the matched sample, α is an intercept, and ε is an error term. The betas refer to the relative impact of a given variable on the log odds of being matched: β₁ refers to family income, β₂ refers to the respondent's race, and β₃ refers to the family head's education.

We use the logit coefficients to calculate the probability of being matched:

We multiply the inverse of this probability by the SIPP calendar-year 2004 person weight, which yields a SIPP person weight adjusted for nonmatches.