No. 44 December 2004
Child Poverty and Welfare Reform: Stay the Course
by June E. O’Neill and Sanders Korenman
In 1996, the public debate over welfare reform included dire predictions that new work requirements and restrictions on lifetime benefits would thrust millions of children into poverty and leave a lasting stain on the nation’s conscience. Eight years later, with welfare reauthorization pending in Congress, those predictions have proven unfounded.
There is now broad agreement that welfare reform worked—as demonstrated by the large declines in both welfare rolls and child poverty since 1996. But evidence for a direct effect of welfare reform on child poverty is clouded by a number of other trends that coincided with welfare reform authorization—not the least of which was the longest sustained period of economic expansion in the nation’s history.
As the nation’s policymakers debate welfare reform reauthorization, it is important to quantify the impact of welfare reform on child poverty and understand how welfare reform fits into what we already know about poverty and how it can be reduced.
We identify a number of factors that played a significant role in reducing child poverty: increased work participation among single mothers; a rise in the level of wages; increases in parental education; declining family size; and, for some groups, a rise in the proportion of children living with two parents.
Although we cannot attribute all of these developments to welfare reform, we do find that welfare reform may be responsible for as much as half of the decline in child poverty among black and Hispanic households headed by single mothers—groups that had the highest rates of welfare participation and child poverty prior to reform.
In 1996, the public debate over welfare reform included dire predictions that new work requirements and restrictions on lifetime benefits would thrust millions of children into poverty and leave a lasting stain on the nation’s conscience.
Eight years later, with welfare reauthorization pending in Congress, those predictions have proven unfounded. Indeed, there is now broad agreement that welfare reform worked—as demonstrated by the large declines in both welfare rolls and child poverty since 1996. But evidence for a direct effect of welfare reform on child poverty is clouded by a number of other trends that coincided with welfare reform authorization—not the least of which was the longest sustained period of economic expansion in the nation’s history.
As the nation’s policymakers debate welfare reform reauthorization, it is important to quantify the impact of welfare reform on child poverty and understand how welfare reform fits into what we already know about poverty and how it can be reduced. We identify a number of factors that played a significant role in reducing child poverty: increased work participation among single mothers; a rise in the level of wages; increases in parental education; declining family size; and, for some groups, a rise in the proportion of children living with two parents. Although we cannot attribute all of these developments to welfare reform, we do find that welfare reform may be responsible for as much as half of the decline in child poverty among black and Hispanic households headed by single mothers—groups that had the highest welfare participation and child poverty rates prior to reform.
2003 Poverty Data
This report uses data on incomes and poverty through 2002, the most recent year for which data were available. However, as we were going to press, the Census Bureau released poverty and income data for 2003, based on cash income. (Data on non-cash benefits have not yet been released.) A preliminary reading of the new data indicates that child poverty increased somewhat between 2002 and 2003, but remains significantly below what it was in 1995, the year before welfare reform became law and a year of lower unemployment than 2003. Thus the child poverty rate increased from 16.7% in 2002 to 17.6% in 2003, still well below the 1995 poverty rate of 20.8%. The poverty rate of children living with single mothers—the group most vulnerable to poverty and most likely to have been affected by welfare reform—despite rising in 2003, remained more than 8 percentage points below the1995 poverty rate. As we demonstrate in this report, child poverty is strongly related to parental employment. In 2003 parental employment declined somewhat, particularly for single mothers. However, available data show that employment has rebounded in 2004, suggesting that child poverty is not likely to increase in 2004 and may have already begun to decline.
This finding is important because the well-being of children is a subject that has received relatively little attention in the many assessments of welfare reform.
Yet child poverty merits particular attention. Equality of opportunity is a fundamental and broadly held American value, and poverty in childhood can be an impediment to that equality, especially if child poverty and its associated disadvantages have long-lasting developmental consequences. Measuring the impact of welfare reform on child poverty is therefore apt to have broader implications for evaluating proposed revisions to welfare reform.
II. Methodology and Findings
Child poverty showed a rapid decline in the late 1990s. The decline was especially dramatic for black and Hispanic children, among whom the poverty rate dropped by close to one-third between 1993 and 2002, and did not increase significantly during the recession years of 2001–2002.
The fact that the decline in child poverty overlapped large national economic gains in the mid- to late 1990s has been used by welfare reform critics to bolster the argument that declines in child poverty after 1996 were due to broader economic factors and not welfare reform. However, this criticism does not explain why child poverty declined in such a sustained and dramatic fashion, since it had been impervious to improvements in the economy for more than the two decades prior to the mid-1990s.
Because the relation between child poverty and economic growth since the mid-1990s appears to be different from that of recent decades, we must try to identify other potential contributing factors, such as changes in parental characteristics and welfare reform. In pursuit of this goal, we have documented changes in relevant parental and family characteristics, as well as in economic variables, and we use multiple regression analysis to identify the effect of these variables on a child’s poverty status. We then estimate the contribution of changes in these variables to the decline in child poverty between 1995 and 2002.
We find that the important factors contributing to the poverty decline include the increased employment of single mothers, an increase in parents’ education, a decrease in the number of children per family, and an increase in wage rates in the economy. One notable qualification, however, is that the relative importance of these factors differs by race. Among black and Hispanic children, the groups most affected by the Personal Responsibility and Work Opportunity Reconciliation Act (PRWORA), the increase in single mothers’ employment accounts for a particularly large share of the decline in poverty. Among these same groups, but not among white children, there was an increase in the proportion of children living with two parents, which also helped to reduce poverty. In short, groups with the largest declines in poverty, such as black and Hispanic children and children in female-headed households, were more likely than other children to have been positively affected by welfare reform.
The direct effect of welfare reform was to move families from welfare to work. Other factors contributing to the decline in child poverty may very well have been influenced by welfare reform, although the linkages are more complex and difficult to document. For example, the proportion of black and Hispanic children living with married or cohabiting parents increased somewhat between 1995 and 2002, contributing to an additional decline in poverty rates. This is a much welcomed development, as it was hoped that welfare reform would provide an incentive for marriage once single mothers could no longer count on long-term cash assistance without work, as had been the case under AFDC. The recent increase in two-parent families may eventually prove to have been a consequence of welfare reform. However, as yet there is not sufficient evidence to determine whether we have seen the beginning of a trend or simply a short-term fluctuation.
III. Comparing Gains with Historical Trends
The reduction in child poverty during the 1990s is all the more remarkable when seen against the backdrop of trends over the past 50 years. During the 1960s, the child poverty rate dropped sharply, falling from 27% in 1959 to 14% in 1969, a level to which it has never returned (Figure 1). That decline was in step with the rapid economic growth of the 60s and mirrored the general reduction in poverty for all persons and for those aged 18–64. In fact, both the child poverty rate and the generally lower poverty rate for 18–64-year-olds declined by 49% between 1959 and 1969.
However, after 1969 child poverty began to climb, and grew faster than the poverty rate of the total or adult population. Economic recessions in the 1970s and 1980s raised the child poverty rate, while the recoveries that followed produced little decline, not nearly enough to offset the preceding recession-related rise. Thus, after declining by two to three percentage points during the strong economic recovery of the 1980s, the child poverty rate was still 40% above the 1969 rate. During the recession of the early 90s, it rose again, hitting 23% in 1993, a level not seen since 1963, the eve of President Johnson’s declaration of the War on Poverty. By 1996, after several years of economic recovery, the child poverty rate had fallen by only about two percentage points, a slight improvement, in step with the pattern repeated since the 70s.
But the old pattern was not repeated after 1996. By 2001, the child poverty rate had fallen by another 4 percentage points—to 16.2%—for a total decline of more than 6 percentage points from the 1993 peak, and the lowest rate since the mid-1970s. Furthermore, in 2001 and 2002 (a period that includes the 2001 recession and a mild recovery), the child poverty rate increased by only half a percentage point, and currently stands at 16.7%.
The decline in child poverty since 1993 is even more remarkable when compared with the poverty decline experienced by the general population (Figure 1). Whereas the child poverty rate declined by 6.4 percentage points between 1993 and 2001 (a 28% decline), the poverty rate of the population aged 18–64 declined by only 2.3 percentage points (an 18.5% decline), suggesting that the robust economy alone cannot account for the improved economic status of children. As we show below, that relative improvement was the result of a significant increase in the work participation of single mothers, combined with increases in parental education and other factors, such as reduced family size. In addition, the percentage of children living in single-mother families, which had been increasing steadily since the late 60s, reached a peak in 1995 and since then has remained roughly stable. In short, the late 1990s reveal substantial gains in the economic well-being of children that cannot be accounted for by normal economic fluctuations.
IV. Child Poverty: Alternative Measures Show Even More Improvement
Up until now, we have been examining trends in child poverty using the “official definition,” which bases the poverty threshold on the cash income of a child’s family. However, alternative measures of poverty can be constructed using different definitions of the family unit and of the income received by the family that indicate even larger declines in child poverty.
The definition of the family unit is important because cohabitation of single mothers and single fathers has been increasing—a possible effect of welfare reform. Because partners are not treated as relatives, they are excluded from the family unit, and consequently their income is excluded from family income. When the unit is changed from the family to the household, the partner’s income (as well as that of other unrelated members of the household) is included. As a result, the per-capita income of the household tends to rise and poverty to fall, because partners usually work and their earnings add to the household’s income.
The omission of noncash income supplements is another reason that the official poverty definition based on family cash income is likely to understate economic well-being. Particularly since the late 60s, many large-scale government programs were introduced to provide noncash benefits to the poor (for example, food stamps, Medicaid, and housing subsidies). The value of employer-provided health insurance and other fringe benefits also increased. More recently, the income of many families with children has been augmented by the Earned Income Tax Credit (EITC), which provides income supplements to working low-income families. On the other hand, federal, state, and local taxes reduce income available for consumption. Using Census Bureau estimates of the value of noncash income supplements and tax payments, we have calculated an alternative measure of poverty based on a family’s “full income.”
Figure 2 compares the trend in the child poverty rate from 1982 to 2002 using three alternative poverty measures. The first is based on the official definition—the family is the unit, and income includes only cash sources before taxes. The child poverty rate is somewhat lower under the second definition, where the household replaces the family as the income unit. It is substantially lower under Definition 3 (“full income”). The three measures display similar patterns of change over the years shown, although the spread between the official definition and “full income” is somewhat larger during the 90s than during the 80s, reflecting the growing value of government and employer-based health benefits and considerable enhancement of the EITC.
In the analysis of child poverty that follows, we primarily focus on trends in household cash income—Definition 2. This allows us to take account of a more inclusive measure of household income than family income and to document the growth of two-income households.
V. Measuring the Decline in Child Poverty Across Groups
As we noted earlier, the level and trend in child poverty differ considerably among different populations. Here we compare the trend in poverty among groups, focusing particularly on the period 1993–2002, which begins with the trough of the business cycle and the peak in welfare participation and spans the introduction of welfare reform and its aftermath. A comparison of groups of children who were differentially exposed to welfare reform provides insights into the relative impact of welfare reform on child poverty. Consequently, we examine the change in poverty status of groups of children who differ by race and by living arrangements.
Our findings show that those groups with relatively high rates of welfare participation prior to welfare reform experienced the sharpest reductions in child poverty after reform. A greater decline in poverty among groups that formerly were more dependent on welfare strongly suggests that reform had a positive effect on the economic well-being of children in these groups.
A. Differences by Race and Hispanic Origin
As shown in Figure 3, the poverty rates (based on household cash income) of all groups of children declined sharply between the peak poverty year of 1993 and 2002, even though, like 1993, 2002 was a recession year. Although the poverty rate is still much lower among white non-Hispanic children than it is among black non-Hispanic and Hispanic children, a considerable narrowing in the poverty differentials between these groups occurred in recent years, particularly after welfare reform (1996). Among white non-Hispanic children, poverty declined by four percentage points between 1993 and 2002 (from 12% to 8%), with the bulk of the decline occurring between 1993 and 1996—the period of economic recovery from the recession of the early 90s.
In contrast, over the 1993–2002 period, the poverty rate of Hispanic children declined by 13 percentage points (from 39% to 26%) and that of black children declined by 14 percentage points (from 44% to 30%). Moreover, almost all of the decline in poverty among Hispanic children and more than half of the decline among black children occurred after 1996. Because a much larger proportion of black and Hispanic children than of white children live in single-mother families, the improvement in their economic status is more likely to have been driven by welfare reform.
B. Differences by Living Arrangements
Perhaps the most striking change in the child’s home environment over the past four decades has been the increasing proportion of children raised by a single parent, usually the mother, and a corresponding decline in the proportion living with two married parents. As we noted earlier, the poverty rates of children in single-mother families are substantially higher than the rates of children living with two married parents.
However, as displayed in Figure 4, the poverty rate of children living with a single mother declined considerably—from 48% in 1993 to 34% in 2002—a larger decline than that experienced by children in two-parent families (11.5% to 8.4%). Consequently, after welfare reform (in 1996) there was significant convergence in the poverty rates of children in single- and dual-parent households, as indicated in Figure 4.
C. Differences by Age of Children
Families with young children tend to have higher welfare participation, lower earnings, and higher poverty rates than other families with children. (Approximately two-thirds of families on AFDC had a child under age 6.) Figure 5 compares trends in the poverty rates of younger and older children. To delineate the effect of age more clearly, the younger children examined here are selected as aged six and younger, living in households in which no child is older than six. The older children are selected as aged 7–17, living in households in which no child is younger than age seven. As shown, although poverty declined for both groups of children after 1993, the decline was steeper among younger children, particularly after 1996. As a result, there was a striking convergence in the poverty status of the two groups, despite the seemingly greater effect of the recession of 2001–2002 on the household incomes of families with younger children.
By separating out child poverty rates by race, living arrangements, and age, we see that the groups with the highest rates of welfare participation and child poverty were also the beneficiaries of the largest relative declines in child poverty after the welfare reform enactment in 1996. Although this evidence suggests the efficacy of welfare reform for child poverty reduction, we cannot rule out the influence of other variables. We turn to these factors now.
VI. Changes in Parental Characteristics Affecting Poverty
Changes in child poverty are obviously tied to changes in parents’ incomes, which in turn are influenced by many factors, some of which are likely to be related to welfare reform while others are not. Several parental characteristics are particularly important determinants of income: the living arrangements and marital status of parents, parental employment, and parental education. In this section we describe the changes that have occurred in those characteristics, and in the next section we discuss the results of a statistical analysis that provides estimates of their contributions to the decline in child poverty.
A. National Trends in Family Structure
As even the most casual observer of American society would note, the structure of the family has been changing for decades, although those changes have not been evenly distributed across all ethnic and racial groups. In 1959, only 9% of children lived in female-headed households (Figure 6, top panel). That proportion began to rise rapidly in the late 1960s and by 1995 was close to 25%. As shown in the bottom panel of Figure 6, the relatively high poverty rate of children in single-mother households is long-standing. Between 1959 and 1995, it was typically 40 to 45 percentage points above the poverty rate of children in households not headed by a single mother. The long-term increase in children living in single-mother families has been linked to the rise in the overall child poverty rate through the early 1990s.
By 2002, about 69% of all children lived with two married parents, 23% lived with a single mother, 4.6% lived with a single father, and 4% lived with neither parent (Table 1). Most children who are not living with either parent live with a relative, usually a grandparent, and the remainder live with a foster parent or other unrelated adult. Each of the different household types may contain other adults—relatives and nonrelatives—and the presence of those adults affects the poverty status of the household, either by adding more income or by adding another dependent to the household. About 2.6% of all children lived with a mother and her male partner (which is more than 11% of those in single-mother households), and another 1.7% of all children lived with a single father and his female partner (37% of those in single-father households). Consequently, the combined proportion of children living with parents who are either married couples or cohabiting couples was about 73% in 2002.
As shown in Table 1, between 1985 and 1995 the proportion of children living with two married parents declined by 6 percentage points—from 75% to 69%, while small increases occurred in the proportion of children living in single-mother and single-father families and in households with no parent present. However, between 1995 and 2002 the proportion of children living with married parents stabilized, while the proportion in single-mother and in no-parent households declined a little and the proportion in single-father families increased a little. Throughout the period 1985–2002, the proportion of children living with a cohabiting couple increased (both single mother/male partner and single father/female partner) rising from 1.5% in 1985 to 4.3% in 2002.
B. Living Arrangements by Race and Hispanic Ethnicity
The living arrangements of children differ significantly by race and Hispanic ethnicity. In 2002, 77% of white children, 65% of Hispanic children, and 38% of black children lived with parents who are married (Table 1). The percentage of children living with married parents has been falling for many years in each of these groups. However, as shown in Figure 7, over the second half of the 1990s this decline seems to have come to an end for black children as the proportion living with married parents underwent a reversal and rose by five percentage points (upper panel). For whites, marriage continued to decline slightly, and for Hispanics, the trend flattened.
The lower panel shows the trend for children living with parents who are either married or cohabiting. Although the trend line for whites looks similar to that in the upper panel (though level in the late 90s instead of slightly declining), there is now evidence of a 4–5 point increase in “couple-headed” families for Hispanic children after 1994 and of a more pronounced (6–7 point) increase for black children. The increases for black and Hispanic children were sufficiently large to fully reverse the decline in couple-headed families from 1984 to the mid-1990s. Although we cannot attribute these changes with certainty to welfare reform, the timing is certainly suggestive and merits further study.
C. Living Arrangements by Age
Figure 8 compares the trends in living arrangements for younger and older children, again presenting the trend in the percentage of children living with two married parents and the trend for those living with parents who are either married or cohabiting couples. The population shown in the upper panel includes only children in families in which all resident children are under the age of seven. Between 1984 and 1995, there is a steep nine-percentage-point decline in the percentage of young children living with married parents, and that decline is followed by a small, 2 percentage-point rise between 1995 and 2002. However, among younger children, the proportion living with cohabitating parents has increased considerably, rising from about 2% in 1984 to 4% in 1995 and to 7% in 2002. Consequently, when we examine the trend using the broader definition of two-parent family (that is, including cohabiting as well as married parents), the decline from 1984 to the mid-90s is less pronounced and the reversal from 1995 to 2002 is stronger, a 4-percentage-point rise.
By contrast, among children in households in which all the children are older than age 7 (bottom panel of Figure 8), the percent living with two married parents declines by only 4 percentage points between 1984 and the mid-90s and continues the downward trend, but only weakly, after that. Because the proportion of older children living with cohabiting parents is quite small (accounting for only about 3% of these children in 2002), there is only a slight difference between the trend in the proportion of children living with married parents and the trend for those living with either married or cohabiting parents. Again, the group most affected by welfare reform exhibited more positive results.
D. Parental Employment
Whether or not a child’s parents are employed is bound to have a significant impact on whether a child is poor. The proportion of all children who live with at least one working parent has increased modestly since the mid-1990s, rising from about 85% in 1995 to 89% in 2002, after declining slightly during the 2001–2002 recession (Table 2). However, among children who are more vulnerable to poverty and who consequently have higher welfare participation rates—black and Hispanic children, young children, and those living with a single mother—the gains in parental employment have been much larger. As shown in Table 2, between 1995 and 2002, the proportion of black and Hispanic children living with a working parent increased by about 10 percentage points—to 77% and 85%, respectively—bringing them closer to the proportion of white non-Hispanic children living with a working parent (91% in 1995, 92% in 2002). These gains are also larger for younger children than they are for older children, increasing by 15 percentage points for black children under the age of seven.
The increase in the probability of living with a working parent is due in large part to increases in the proportion of mothers who work. The mothers of black and Hispanic children were much more likely to work in 2002 than they were in 1995, which was not the case for white non-Hispanic mothers. However, the proportion of children living with a working parent is also influenced by their living arrangements. Children living with two parents are much more likely to have a working parent than children living with a single mother; single fathers are somewhat more likely to work than single mothers. As discussed above, the proportion of black and Hispanic children living with two married parents and with single fathers increased while the proportion living with single mothers decreased, and these changes have contributed to the increases in children living with a working father shown in Table 2, as well as the increase in children with at least one working parent.
Table 3 provides the details on parental work participation separately for children living with two married parents and for children living in single-mother families, and the information is given by race and Hispanic ethnicity. In most two-parent families, at least one parent works. (The proportion of children in these families with no working parent was 6% for blacks, 7% for Hispanics, and 4% for whites in 2002.) Both parents work in about 60% of black and white families, and that proportion has been quite stable since the mid-90s. However, although the proportion of Hispanic families with both parents working has increased since 1995, it was still only 46% in 2002, considerably below other groups. Hispanic families also differ in that the proportion in which only the father works—the so-called traditional family—is well above that of blacks and whites. In 2002, among children in two-parent families, 42% of Hispanic children were in families in which only the father works, compared with 21% of black children and 32% of white children. It is much more unusual for only the mother to work in two-parent families. Only about 5% of Hispanic and white children were in that category. However, the proportion of black children in mother-only-works families has been increasing slowly and stood at 13% in 2002.
The work participation of mothers changed much more dramatically for children living in single-mother families—traditionally a welfare-dependent group. For children living in single-mother families, the proportion with a working mother increased between 1995 and 2002 by about 15 percentage points among black children, and by a startling 24 percentage points among Hispanic children, but by only 8 points among white children.
E. Parental Education
The education level of resident parents has also been increasing, and more so for mothers than for fathers. Over the period 1985–2002, the percentage of fathers completing at least one year of college rose from 46% to 57%; for mothers, that percentage rose from 36% to 55% (Table 4). Although by 2002, mothers were almost as likely as fathers to have attended college, they were still less likely to have completed college. The pattern of larger educational gains for mothers than for fathers was repeated among Hispanics and whites. But among blacks, fathers and mothers gained at about the same rate.
Parental education differs considerably by race. White mothers and fathers are much more likely to have attended and completed college and less likely to have dropped out of high school than their counterparts who are black or Hispanic. However, the level of schooling completed by Hispanic parents is considerably below that of either black or white parents. For example, about 43% of Hispanic mothers had not completed high school in 2002, compared with 15% of black mothers and 7% of white mothers. The educational level of Hispanic parents also increased much less than that of black or white parents between 1985 and 2002. However, the Hispanic population has grown rapidly in recent years through the immigration of young adults with relatively low levels of education. The influx of these immigrants reduced the average educational attainment for Hispanics as a whole.
Parents are not randomly selected from the population. Women who have high educational attainment are more likely to delay marriage and childbearing. Selection may also be at work sorting out single mothers and married mothers. As shown in Table 5, which provides detail on parental educational attainment by marital status, among both black and white parents, married mothers are considerably more educated than single mothers. Although single mothers have gained more schooling over time, they are still more likely to have failed to complete high school and much less likely to have gone to college than married mothers. Among Hispanics, however, there is little difference in the years of school completed of single and married mothers.
VII. Accounting for the Decline in Child Poverty
As we noted above, the dramatic fall in child poverty rates between 1995 and 2002 coincided with welfare reform as well as other changes in family characteristics. It also was a period of strong economic performance. All of these factors may help explain the decline in child poverty in those years. In order to arrive at a more complete accounting of the relative importance of these factors to the substantial decline in child poverty, we conducted a set of multivariate regression analyses. We used the analyses to identify the effects of the relevant variables on poverty change and to estimate the contribution of these factors to the decline in poverty. This section summarizes the results; the underlying regression analyses are presented in the Appendix.
In the first step, we regressed the poverty status of a child on a standard set of demographic and geographic controls: race; education and age of mother or of other caretaker; number of children and adults in the household; six regions; and an indicator for urban residence. We also included controls for the level of welfare benefits in the state and for local labor market indicators most pertinent to low-income households: the state unemployment rate and hourly wage rate for workers with no more than a high school education. To capture the pathways through which welfare reform affected child poverty, we also included a set of controls for 14 categories of combined family living arrangements and work status. Specifically, the 14 categories span two employment statuses (currently employed, not employed and four combinations of mother/father employment in married-couple families and six living arrangements—two-parents, independent single mother, single mother with an unmarried male partner, single mother with other adults such as a grandmother, single father, no parent present but lives with other relatives or nonrelative adults).
The mean values of these variables in 1995 and 2002 are displayed in Table 6 for all children and Table 7 for children in single-mother families.
We regressed the 1995 child poverty rate on the control variables in 1995, for the entire population of children and for groups such as Hispanic children, black children, or children in single-parent families. We then used the 1995 regression equations to predict the change in the child poverty rate between 1995 and 2002 that would have occurred from changes in the average values of the control variables between 1995 and 2002 (e.g., changes in work and family arrangements; and increases in wage rates for less-educated persons.). In this way, we can break down the change in the child poverty rate for the entire population and for each group into parts contributed by each category of control variables. Table 8 summarizes the results of this exercise.
For example, the first entry in the first column of the table indicates that the overall child poverty rate—the rate for all children combined—fell by 4.2 percentage points between 1995 and 2002 (from 19.2% to 15%). Looking down the first column, the most important contributor to that decline was the change in the hourly wage rate, which alone accounted for about half the decline in child poverty in the period. Other key factors in the overall decline in child poverty were increases in educational attainment (although not shown separately, particularly the fraction of mothers with a college degree), decreases in the number of children per family, and increases in the proportion of children living with an employed single mother, an employed single father, and, to a lesser extent, two working parents. Notably, the gain in the share of children living in households with a working parent offset the decline in the proportion of children living with a nonworking single mother, most of whom are welfare recipients, and in doing so had a strong poverty-reducing effect.
Some factors were important in reducing poverty in nearly every group: specifically, increases in parental education, declining family size, and the rise in the real wage rate. However, the weight of these factors varied by group. Thus, the rise in the wage rate accounted for three-quarters of the small (2-percentage-point) decline in poverty among white children but only about one-fifth of the larger declines in poverty among black and Hispanic children. Improved education, too, was a much more important factor in the poverty decline among whites than the other groups.
The contributions of other factors also varied across groups. Increased work participation among single mothers was hugely important for all children in single-mother families, but especially in black and Hispanic single-mother families, where this change accounted for nearly 40% of the very substantial decline in child poverty in this period.
The combination of an increase in the proportion of children residing in two-parent families and an increase in work participation within these two-parent families was also quite important for Hispanic children. The fraction of all Hispanic children living with two married parents increased from 63% to 65% and the subset living with two employed parents increased from about 31% to about 35% over the 1995–2002 period. These changes accounted for 19% of the reduction in poverty among all Hispanic children.
A reduction in family size (the average number of children in a family) accounted for nearly one-quarter of the 14.1-percentage-point decline in the child poverty rate among black single-mother families.
What was the role of welfare reform? Although we cannot say precisely, we can make rough judgments about the size of the effects. Some of the forces that led to reductions in poverty are unlikely to be by-products of welfare reform. For example, the increase in average hourly wages that occurred in most states in the U.S. is attributable to the unprecedented increase in productivity growth since the late 1990s and is not likely to be related to welfare reform.
Other factors contributing to the reduction in child poverty may have been influenced by welfare reform although to date there is little research evidence to support a relation. One example is the increase in educational attainment of parents. Education has been generally trending upward over time, and we cannot discern an acceleration in that trend related to welfare reform. Welfare reform may provide the motivation to pursue advancement in education and to make other investments in human capital that eventually will raise the wages of less skilled workers. But it is unlikely that we would have seen those effects so soon after the passage of welfare reform.
It is also possible that welfare reform played a role in the reduction in the average number of children in single-mother families. There has been a significant decline in the birthrate of teenage girls, particularly black teens among whom most births are out-of-wedlock and in the past were frequently followed by long spells of poverty and welfare participation. Although that decline began in the early 1990s, several years before the passage of reform legislation, the decline continued after the passage of welfare reform, dipping below the level that predated the sharp rise in teen births in the late 1980s. However, it is difficult to establish a causal role for reform, and thus far, statistical studies have not found evidence of a substantial effect of welfare reform on fertility.
The increase in the proportion of children in two-parent families, which was most evident among Hispanic families, in theory also could be an effect of welfare reform. Again, evidence is lacking. However, demographic forces involving fertility and marriage are strongly influenced by cultural factors that may well change in the long term as economic incentives for early childbearing and single parenthood are altered by welfare reform.
Having said that, welfare reform certainly deserves credit for much of the reduction in child poverty that came about from increased work among single mothers, the group most affected by welfare reform. Although the employment effects of welfare reform might account for only about 10% of the decline in poverty among all children, it could plausibly account for up to 35% of the very large decline in poverty among black and Hispanic children in single-mother families, and about 21% of the decline among white single-mother families. Since the local area unemployment rate and low-skill wage rate are controlled in the model, these increases in employment (and living arrangements) are above what would be expected to occur based on the strong economy alone.
In sum, the most important factor driving the decline in poverty for all children appears to have been the increase in the wage level that began in the late 1990s. However, the accounting is quite different for children in single-mother families, who have always had the highest poverty rates and who experienced much larger reductions in those rates than children in other families. Perhaps as much as half of the decline in poverty among children in single-mother families could be attributed to changes in parental behavior spurred by welfare reform, if we also attribute changes observed in family size and living arrangements to incentives produced by welfare reform. It is notable in this regard that black and Hispanic children in single-mother families also experienced the sharpest declines in welfare participation (see Appendix Figure A3).
VIII. Concluding Comments
We have seen that child poverty rates, especially among children in single-parent families, decreased dramatically in the second half of the 1990s and slightly increased in the recession that began early in 2001. The decline in poverty is partly attributable to the productivity driven increases in wage levels that gained momentum after 1995. But welfare reform also played a significant role, particularly in reducing poverty among children in single-mother families and among children who are black or Hispanic. The sharp increase in work participation by single mothers was to a large extent spurred by welfare reform and was a major factor contributing to the decline in child poverty for these groups. Increases in parental education, declining family size, and, for some groups (notably Hispanics), an increase in the proportion living with two parents also contributed to the decline in child poverty. Although these other poverty-reducing characteristics may also have been influenced by welfare reform, that linkage is not well established at this time.
The income gain associated with the decline in child poverty is in itself beneficial to children. However, increases in education and employment, decline in long-term welfare dependence, and decreases in family size among disadvantaged families all confer additional long-term developmental advantages on children beyond those provided by income alone.
What, then, are the lessons for welfare reform reauthorization currently being debated in Congress? We believe that the evidence indicates that TANF (Temporary Assistance to Needy Families), the new welfare system created by the welfare reform act of 1996 (PRWORA), has been a clear improvement over the old AFDC program that it replaced. We doubt that the employment and earnings gains among single mothers would have been achieved without welfare reform. Moreover, we see no evidence of adverse impacts on child well-being and considerable evidence of promising developments. Our (bipartisan) advice to policymakers, then, is this: if it’s working, don’t fix it. We see no evidence that retreating from the emphasis on work requirements and time-limited benefits would be beneficial for children; neither do we see a need to legislate increase in the number of hours of work required of welfare recipients. The states currently have the flexibility to be more stringent than current law requires and a number have taken the option to do so. On the other hand, current proposals to broaden the range of activities qualifying as primary work to include, for example, support for college education, undermine the work incentives that appear to be responsible for the positive effects of reform. This is not to say that subsidizing the education of disadvantaged young women and men is a bad idea. Quite the contrary. But it should not be tied to welfare participation.
In sum, it seems to us that the current orientation of welfare reform has struck an appropriate balance between providing a helping hand to the needy and improving incentives for self-sufficiency. PRWORA appears to be one of the rare policy prescriptions that actually exceeded expectations. Our advice to policymakers, therefore, is to maintain the course they’ve charted.