Table of Contents | Previous | Next |
Chapter Five
Employment and Earnings
5.1. BACKGROUND
Beyond reducing welfare dependence, one of the key goals of PRWORA is to increase work. By all accounts, work has risen in recent years. The fraction of welfare recipients working rose from 7 percent in 1992 to 33 percent in 1999. In 1998, the fraction of welfare recipients starting jobs exceeded 50 percent in 10 states. The fraction of job entrants still employed after three months exceeded 75 percent in 29 states (USDHHS, 2000).
Among single mothers more generally, recent trends in employment and earnings are the mirror image of recent trends in welfare use. Employment among single mothers rose from 69 percent in 1993 to 83 percent in 1999, a gain of 20 percent. Weeks worked during the year, a broader measure of work effort that we will refer to as "labor supply," rose from an average of 29.5 to 36.7 over the same period. Mean trends in earnings have also been favorable. Measured in 1998 dollars, average annual earnings among female family heads (including those without earnings) stood at roughly $12,300 in 1993. By 1999, they had risen to nearly $16,600, a gain of 35 percent (Grogger, forthcoming).
Although these average trends are favorable, some low-income families may have lost economic ground in recent years. Earnings among the lowest 20 percent of female-headed families fell by roughly 10 percent between 1995 and 1997, but appear to have rebounded in the meantime (Primus et al., 1999; Haskins, 2001). Fifteen welfare-leavers studies sponsored by USDHHS show that roughly 50—65 percent of persons leaving welfare are employed in their first quarter off aid (USDHHS, 2001a). Approximately 60—70 percent find employment at some point within the first year. However, many of these jobs are fairly unstable, since only about 40 percent of welfare leavers on average are employed in each of their first four quarters off aid.
These studies also show that earnings among employed welfare leavers average roughly $1,800 to $3,400 in their first quarter off the rolls. There is some evidence of earnings growth, but it is small. By their fourth quarter off welfare, leavers with earnings typically make between $2,100 and $3,900 (USDHHS, 2001a). Such relatively flat earnings trajectories could stem either from low wage growth or from intermittent work. Recent research shows that the wages of low-income women rise with work experience in a manner that is comparable to other workers when experience is measured as actual hours of previous work (Gladden and Taber, 2000; Loeb and Corcoran, 2001). This suggests that low earnings growth is due to sporadic employment.
Although this descriptive evidence provides an invaluable context for interpreting the effects of recent policy changes, our interest lies not in merely describing the experience of low-income families in the post-reform era, but rather in assessing how welfare reform in general, and specific reform measures in particular, have affected employment and earnings. The economic model discussed in Chapter 2 predicts that nearly all the recent reforms should increase the employment of single parents. Specifically, because financial work incentives increase the amount of earnings a working recipient may keep, they promote employment. Other reforms encourage recipients to seek work as well, either as a condition for receiving aid or in anticipation of reaching their time limit. The one possible exception involves work requirements that mandate participation in an education-focused welfare-to-work program. Recipients participating in such a program may actually work less than they would have otherwise, at least while they are taking part in the program.
However, the earnings effects of some of the reforms are ambiguous, because of the countervailing incentives discussed in Chapter 2. As noted there, financial work incentives increase the rate of return from work, but they also raise income for a given level of employment. The effect on returns, known as the "substitution effect" in the economics literature, acts to increase labor supply, but the effect on income, known as the "income effect," may induce the consumer to reduce her labor supply. The net effect of the financial work incentive on labor supply and earnings is thus ambiguous, depending on the relative magnitudes of the substitution and income effects. All else equal, we would expect smaller income effects among families with lower levels of income, simply because such families are less able to "afford" to reduce their labor supply in response to a financial incentive.
To illustrate this point, consider two women, each of whom earns $6/hour. The first works 40 hours per week, the second works 10 hours per week. A decrease in the benefit reduction rate from 67 to 50 percent raises each worker’s net wage by $1/hour. If the workers do not change their hours, this would result in a monthly earnings gain of $160 for the first worker and $40 for the second worker. Alternatively, the first worker could reduce her labor supply by 10 hours per week and still enjoy greater earnings than she made before the benefit reduction rate reduction. However, if the second worker reduced her hours by the same amount, her earnings would fall to zero.
The logic of this example is sometimes used to argue not only that substitution effects should be larger among families with lower income, but that, among welfare families, the substitution effect should dominate the income effect, causing reductions in benefit reduction rates to increase labor supply and earnings as well as employment. However, some of the evidence below suggests that, even at the levels of income received by some welfare recipients, the income effect may outweigh the substitution effect.
Other polices with complex effects on labor supply and earnings include work requirements and time limits. Both policies encourage employment, either directly or indirectly. However, both policies may hasten job search, which could lead recipients to accept jobs with lower wages and perhaps other characteristics that might result in a less durable match between the worker and the employer. Thus, it is conceivable that these policies could result in fewer hours of work and lower earnings.
As in Chapter 4, we consider evidence on the effects of reform from both random assignment and econometric studies. The two types of studies typically employ different types of employment and earnings data. Although some random assignment studies collect survey data, most use administrative data from the states’ unemployment insurance (UI) systems. Every quarter, all employers covered by the UI system are required to report the earnings of all their employees who earn $50 or more. This information constitutes the earnings data used in most random assignment studies, although the studies sometimes analyze annual earnings or even earnings over a multiyear follow-up period instead. Because the units of measurement vary among studies, we report not only the impacts from the studies themselves, but also a normalized monthly impact measure. Typically, a study participant is considered to be employed if she has any reported earnings in a given quarter, although some studies adopt an annual measure of employment instead.
The main problem with such administrative data is that they miss some earnings. The UI system covers about 90 percent of all jobs in the United States; uncovered sectors include self-employment, federal government employment, some state and local government employment, some domestic jobs, and some agricultural jobs (U.S. Bureau of Labor Statistics, 1989). Of course, the informal sector is uncovered as well, which means the UI data miss income from people who provide informal child care, take in laundry, and otherwise work for cash. A comparison of employment data from administrative and survey sources suggests that administrative data underestimate self-reported employment among welfare leavers by about 10—20 percent (Isaacs and Lyon, 2000, Table 2C).
All the econometric studies are based on data from the March CPS. They therefore use fairly similar measures of employment and earnings. Employment is typically measured as a dummy variable that equals one if the survey respondent worked for pay in the year preceding the survey and equals zero otherwise. Respondents also indicate the number of weeks they worked for pay in the previous year, which provides a useful measure of labor supply. In addition, respondents report their income from earnings in the previous year. As in the econometric literature on welfare use, some researchers analyze the individual-level data directly, whereas others first aggregate the data into cells defined by the respondent’s education, age, state, and year.
Many of the analytical issues that arise in econometric studies of the caseload also arise in studies of employment and earnings. In particular, welfare reform is but one of several factors that contributed to recent changes in the labor market behavior of single mothers. The strong economy played an important role, as did the EITC (Meyer and Rosenbaum, 2001; Grogger, forthcoming; Hotz, Mullin, and Scholz, 2001). Other, largely unobservable factors may have played a role as well. As in the caseload literature, researchers have attempted to account for such confounding influences by including controls in their regression models for the economy, characteristics of survey respondents, other policy variables, and state-specific fixed-effects and trends.
In this chapter, we review the evidence on how much welfare reform has increased employment among the welfare-eligible population. We also examine the effects of welfare reform on earnings, which represent the primary means of support for low-income families not receiving welfare. The effects of welfare reform on earnings bear importantly on its consequences for income and other broad measures of family well-being that we address in later chapters.
As in Chapter 4, we begin with a discussion of the random assignment studies, including a brief summary of the results from Appendix A with analyses by subgroups. Following that, we discuss the results from a number of econometric studies. We then synthesize the studies to convey what is known about the effects of welfare reform on employment and earnings. We conclude with a summary of our findings.
5.2. RANDOM ASSIGNMENT STUDIES OF THE EFFECTS OF WELFARE REFORM ON EMPLOYMENT AND EARNINGS
Most of the evidence on the effects of welfare reform on employment and earnings comes from random assignment studies. Table 5.1 summarizes their findings. As in Chapter 4, when the estimated effects of a program seem to change with time since random assignment, or when time limits begin to bind, we report multiple estimates from the same program. Otherwise, we present a single estimate.
5.2.1. Programs That Focus on Financial Work Incentives
Results from the three programs that focus on financial work incentives are presented in Panel A of Table 5.1. According to Becerra et al. (1998), CWPDP reduced employment by 2 percentage points. Although the program impact is insignificant, it appears to contradict the prediction from the economic model discussed in Chapter 2. However, results from a reanalysis of the CWPDP data show that the program increased employment (by 3.1 percentage points in the third year of the follow-up). Moreover, the employment impacts from that reanalysis are statistically significant (Hotz, Mullin, and Scholz, 2002). However, the authors provide no reconciliation of their results with those of Becerra et al. (1998). Since it is beyond the scope of this synthesis to provide a reconciliation of these contradictory findings, we omit the results of CWPDP from the discussion below.
| Employment | Earnings | |||||||||
|---|---|---|---|---|---|---|---|---|---|---|
| Name | Cases served | Data | Measure | Control mean |
Impact | % | Control mean |
Impact | % | Normalize to monthly |
| A. Programs that focus on financial work incentives | ||||||||||
| CWPDP | Single parent recipients | A | Avg. employment, earnings, year 3 | 37.0 | -2.0 | -5.4% | $2,372 | -$160 | -6.7% | -$13 |
| WRP-IO | Single-parent recipients and applicants | A | Ever emp, avg. quarterly earnings, last 3 mos. of FU | 48.4 | 2.8 | 5.8% | $1,502 | $14 | 0.9% | $5 |
| MFIP-IO | Urban single parents recipients | A | Avg. quarterly employment, earnings, year 1 | 32.8 | 7.0*** | 21.3% | $537 | $50 | 9.3% | $17 |
| A | Avg. quarterly employment, earnings, year 3 | 44.7 | 3.6* | 8.1% | $1,298 | -$48 | -3.7% | -$16 | ||
| Urban single parents applicants | A | Avg. quarterly employment, earnings, year 1 | 48.8 | 2.7* | 5.5% | $1,216 | -$66 | -5.4% | -$22 | |
| A | Avg. quarterly employment, earnings, year 3 | 55.3 | 0.0 | 0.0% | $2,017 | -$136 | -6.7% | -$45 | ||
| B. Programs that focus on financial work incentives tied to hours of work | ||||||||||
| New Hope | Poor families employed FT at RA | A | Ever employed, total earnings, year 1 of 2-yr FU | 94.7 | 2.5 | 2.6% | $10,480 | -$253 | -2.4% | -$21 |
| A | Ever employed, total earnings, year 2 of 2-yr FU | 91.8 | 2.6 | 2.8% | $11,550 | -$889 | -7.7% | -$74 | ||
| Poor families not employed FT at RA | A | Ever employed, total earnings, year 1 of 2-yr FU | 77.9 | 9.9*** | 12.7% | $4,380 | $916*** | 20.9% | $76 | |
| A | Ever employed, total earnings, year 2 of 2-yr FU | 76.7 | 6.6*** | 8.6% | $6,129 | $473 | 7.7% | $39 | ||
| SSP (a) | Single-parent recipients | A | Monthly emp. and annual earnings, year 2 | 30.4 | 9.8*** | 32.2% | $3,198 | $1,254*** | 39.2% | $105 |
| A | Monthly emp. and annual earnings, year 3 | 32.5 | 7.2** | 22.2% | $3,852 | $865*** | 22.5% | $72 | ||
| SSP Plus (a) | Single-parent recipients | A | Employment, earnings, Q5 | 31.1 | 16.2*** | 52.1% | $221 | $120*** | 54.3% | $120 |
| SSP Applicants (a) | Single-parent applicants | A | Employment, earnings, Q5 | 38.1 | 4.1** | 10.8% | $552 | $78** | 14.1% | $78 |
| A | Employment, earnings, Q9 | 42.8 | 12.1*** | 28.3% | $610 | $242*** | 39.7% | $242 | ||
| C. Programs that focus on mandatory work-related activities | ||||||||||
| LA Jobs-1st GAIN | Single-parent recipients and applicants | A | Ever emp., avg. total earnings, 2-year FU | 57.6 | 9.6*** | 16.7% | $6,385 | $1,627*** | 25.5% | $68 |
| Atlanta LFA | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 61.6 | 4.5*** | 7.3% | $5,006 | $813*** | 16.2% | $34 |
| Grand Rapids LFA | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 70.1 | 7.6*** | 10.8% | $4,639 | $1,035*** | 22.3% | $43 |
| Riverside LFA | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 45.0 | 15.1*** | 33.6% | $4,213 | $1,276*** | 30.3% | $53 |
| Portland | Recipients and applicants; no cases with substantial barriers | A | Ever emp., avg. total earnings, years 1 and 2 | 60.9 | 11.2*** | 18.4% | $5,291 | $1,842*** | 34.8% | $77 |
| Atlanta HCD | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 61.6 | 2.8** | 4.5% | $5,006 | $496** | 9.9% | $21 |
| Grand Rapids HCD | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 70.1 | 5.3*** | 7.6% | $4,639 | $580** | 12.5% | $24 |
| Riverside HCD | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 38.9 | 9.3*** | 23.9% | $3,133 | $317 | 10.1% | $13 |
| Columbus Integrated | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 72.2 | 1.7 | 2.4% | $6,892 | $673** | 9.8% | $28 |
| Columbus Traditional | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 72.2 | 1.3 | 1.8% | $6,882 | $677** | 9.8% | $28 |
| Detroit | Recipients and applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 58.2 | 4.1*** | 7.0% | $4,001 | $367* | 9.2% | $15 |
| Oklahoma City | Applicants | A | Ever emp., avg. total earnings, years 1 and 2 | 65.0 | -0.9 | -1.4% | $3,514 | $5 | 0.1% | $0 |
| IMPACT Basic Track | Recipients and applicants-Basic Track | A | Employed in Q4; earnings in year 1 | 44.6 | 0.7 | 1.6% | $2,345 | -$146 | -6.2% | -$12 |
| D. Programs that focus on financial work incentives and mandatory work-related activities | ||||||||||
| WRP | Single-parent recipients and applicants | A | Ever emp, avg. quarterly earnings, last 3 mos. of FU | 48.4 | 8.7*** | 18.0% | $1,503 | $135* | 9.0% | $45 |
| MFIP | Urban single-parent recipients | A | Avg. quarterly employment, earnings, year 1 | 32.8 | 13.3*** | 40.5% | $537 | $163*** | 30.4% | $54 |
| A | Avg. quarterly employment, earnings, year 3 | 44.7 | 11.5*** | 25.7% | $1,298 | $143* | 11.0% | $48 | ||
| Urban single-parent applicants | A | Avg. quarterly employment, earnings, year 1 | 48.8 | 3.0** | 6.1% | $1,216 | -$70 | -5.8% | -$23 | |
| A | Avg. quarterly employment, earnings, year 3 | 55.3 | 2.8** | 5.1% | $2,017 | $15 | 0.7% | $5 | ||
| TSMF | Recipients | A | Avg. annual emp., earnings over 4-yr FU | 36.2 | 1.6*** | 4.4% | $3,120 | $223*** | 7.1% | $19 |
| Applicants | A | Avg. annual emp., earnings over 1-yr FU | 41.5 | -0.6 | -1.4% | $3,109 | $134 | 4.3% | $11 | |
| A | Avg. annual emp., earnings over 2-yr FU | 39.0 | 1.0 | 2.6% | $3,426 | $12 | 0.4% | $1 | ||
| FIP | Recipients | A | Avg. quarterly emp., earnings in Q4 of 2-yr FU | 48.2 | 1.3 | 2.7% | $916 | $94** | 10.3% | $31 |
| A | Avg. quarterly emp., earnings in Q8 of 2-yr FU | 56.1 | 1.2 | 2.1% | $1,334 | $57 | 4.3% | $19 | ||
| Applicants | A | Avg. quarterly emp., earnings in Q4 of 2-yr FU | 56.8 | 4.8*** | 8.5% | $1,757 | $152** | 8.7% | $51 | |
| A | Avg. quarterly emp., earnings in Q8 of 2-yr FU | 58.4 | 3.1* | 5.3% | $2,097 | $121 | 5.8% | $40 | ||
| E. Programs that focus on other individual reforms | ||||||||||
| F. Programs that focus on TANF-like bundle of reforms (time limits with financial incentives, work-related activities, or both) | ||||||||||
| EMPOWER (b) | Recipients | A | Quarterly emp., earnings, Q1-Q10 | 39.1 | -1.1 | -2.8% | $886 | -$29 | -3.3% | -$10 |
| S | Employment, monthly earnings at mid-1998 interview | 51.3 | 3.4 | 6.6% | $597 | $60 | 10.1% | $60 | ||
| IMPACT Placement Track |
Recipients and applicants-placement track | A | Employed in Q4; earnings in year 1 | 50 | 7.6*** | 15.2% | $3,139 | $815*** | 26.0% | $68 |
| A | Employed in Q8; earnings in year 2 | 54.4 | 2.8 | 5.1% | $4,944 | $559*** | 11.3% | $47 | ||
| VIP/VIEW | Recipients | A | Avg. emp., earnings in year 2 of 2-yr FU | 51.3 | 2.9*** | 5.7% | $2,777 | $193* | 6.9% | $16 |
| ABC | Recipients and applicants | A | Any emp., total earnings, Q1-Q4 | 48.0 | 9.4*** | 19.6% | $3,378 | $446 | 13.2% | $37 |
| FTP | Recipients and applicants | A | Avg. annual emp, earnings in year 2 | 43.2 | 6.5*** | 15.0% | $3,278 | $661*** | 20.2% | $55 |
| A | Avg. annual emp., earnings in year 3 | 44.6 | 6.7*** | 15.0% | $3,852 | $910*** | 23.6% | $76 | ||
| A | Avg. annual emp., earnings in year 4 | 48.0 | 1.8 | 3.8% | $4,640 | $567*** | 12.2% | $47 | ||
| JOBS First | Recipients and applicants | A | Ever emp., avg. earnings in Q7 | 48.6 | 8.6*** | 17.7% | $1,424 | $149*** | 10.5% | $50 |
| A | Ever emp., avg. earnings in Q8 | 50.3 | 7.7*** | 15.3% | $1,531 | $198*** | 12.9% | $66 | ||
| A | Ever emp., avg. earnings in Q16 | 53.1 | 7.6*** | 14.3% | $2,149 | $129* | 6.0% | $43 | ||
| NOTES: For full program names and citations, see
Table 3.4. Abbreviations: A=administrative data; S=survey data; E=earnings;
W=cash welfare payments; FS=Food Stamp payments; EITC=Earned Income
Tax Credit; CC=out-of-pocket child care expenses; FU=follow-up; HH=household;
Q=quarter. * = statistically significant at the 10 percent level; ** = statistically significant at the 5 percent level; *** = statistically significant at the 1 percent level. (a) Results in Canadian dollars. (b) Phoenix site only, cash assistance. |
WRP Incentives-Only had positive but small and insignificant effects on employment and earnings. This seems consistent with the weak incentive that was offered by the program. The MFIP Incentives-Only program, with its stronger financial incentive, significantly increased employment among both recipients and applicants in the first year after random assignment. Although the effects fade over time, they remain significant among recipients, for whom the first-year effects were stronger as well.
As for earnings, MFIP Incentives-Only had no significant effect, and most of the insignificant estimates are negative rather than positive. This may happen because the income effect, which provides an incentive to decrease hours of work, outweighs the substitution effect, which provides an incentive to increase hours of work.45 As noted in Chapter 2, we would expect the income effect to be larger at higher levels of income. MFIP has a relatively high benefit level, providing $9,228 per year to a family of three with no other income. Compared to other welfare programs, this represents a relatively high level of income.
5.2.2. Programs That Focus on Financial Work Incentives Tied to Hours of Work
The results of the New Hope program, presented in Panel B, differ between families who were working full time at the beginning of the program and those who were not. Among families working full time at random assignment, New Hope had no effect on employment over the follow-up period. It reduced earnings in both years, albeit insignificantly. Among families not working full time at random assignment, New Hope increased employment and earnings substantially.
The difference between these two groups may have to do with the relative strength of the income and substitution effects arising from New Hope’s financial incentive. The average control-group earnings among families working full-time at random assignment was about $11,000. This is comparable to the income provided by MFIP. Moreover, these participants satisfied the work requirement already when the program began and had employment rates over 90 percent throughout the follow-up period. Bos et al. (1999) show that the program led these participants to reduce their annual hours of work, which suggests that the income effect arising from New Hope’s financial incentive may have dominated the substitution effect. In other words, the opportunity to enjoy greater income from the earnings supplement, while at the same time spending fewer hours away from home, outweighed the opportunity to enjoy even greater income gains by working more. For families who were not working full-time at random assignment, who worked and earned substantially less, the opposite may have been true: The opportunity to enjoy additional income may have outweighed the cost of more hours away from home. The requirement to work full time to receive the supplement would have reinforced the positive substitution effect among these participants as well.
The SSP and SSP Plus programs required recipients to be working full time within one year of random assignment to qualify for the programs’ earnings supplement. Both programs increased employment and earnings substantially in the period following the one-year mark, although SSP’s effects faded a bit in the following year. The SSP Applicant study required recipients to be working full-time within two years after random assignment to qualify for the supplement. It increased employment and earnings substantially in the quarter following the two-year mark, but also raised employment and earnings in the interim period.
5.2.3. Programs That Focus on Mandatory Work-Related Activities
The employment and earnings effects of programs that focus on mandatory work-related activities are presented in Panel C of Table 5.1. Of the 13 programs, 12 had positive effects on employment over a two-year follow-up period, and the one negative effect is quite small. Moreover, nine of the 12 positive estimates are significant.
On average, these programs increased employment by 5.6 percentage points during the first two years, which amounts to an average 10 percent gain over the control groups. The LFA programs, which emphasize job search, resulted in larger average employment gains than the HCD programs, which emphasize skill-building and generally require the recipient to participate in classroom activities. The average employment increase among the search-oriented programs was 9.2 percentage points, compared to 3 percentage points among the skills-oriented programs.
The earnings results from these programs were similarly positive. Twelve of thirteen programs produced positive effects on earnings, nine of which were significant at least at the 5 percent level. The one negative effect was insignificant. The average earnings gain over the first two years of the follow-up exceeded $700; only four of the programs failed to increase earnings by at least $400. Again the gains were greater for the search-oriented programs than the skills-oriented programs. Among the four work-first programs, two-year earnings gains averaged $1,188. Among the human-capital programs, they averaged $371.
Five-year employment impacts from NEWWS are presented in Figure 5.1; five-year earnings impacts are shown in Figure 5.2. As in Chapter 4, there is evidence of program fade-out. Annual employment impacts in years four and five averaged 2.0 percentage points, compared to 4.8 percentage points in years one and two. Long-term annual earnings impacts averaged $324, compared to short-term impacts of $378.
|
[D] |
[D] |
In the three sites that ran LFA and HCD programs simultaneously, the gap between the LFA impacts and the HCD impacts faded over time as well. For the first two years of the program, the average LFA impacts on annual employment and earnings were 8.7 percentage points and $561, respectively. The average HCD impacts were 4.7 percentage points and $267. For years three through five, the average LFA impacts on annual employment and earnings were 2.8 percentage points and $355, and the average HCD impacts were 2.2 percentage points and $291. The longer-term LFA-HCD differentials are much smaller than the differentials that appeared during the first two years of the program, suggesting that some of the early differential was due to the fact that LFA participants were looking for work during the initial period of the program rather than taking part in training activities.46
5.2.4. Programs That Combine Financial Work Incentives and Mandatory Work-Related Activities
The results from four programs that combine financial work incentives and mandatory work-related activities are summarized in Panel D of Table 5.1. All these programs report at least some significant and positive employment effects. However, the effects vary across studies.
Similarly, all four programs produced at least some significant earnings gains. Here too, however, the effects are heterogeneous. In some cases, the effects change with time since random assignment; in others, they appear to differ between ongoing recipients and recent applicants.
Although these programs exhibit some consistent patterns, some of their effects seem attributable to their specific designs. The nature of the various mandated work-related activities seems likely to play an important role. For this reason, we discuss the results from each program in some detail.
Both the MFIP and WRP programs have stronger effects on employment and earnings than their incentives-only counterparts. The results from WRP suggest that a requirement to work in exchange for welfare may be effective in increasing employment and earnings, even when it has little effect on welfare use, as was seen earlier in Table 4.1. MFIP’s work requirement seems to decrease the extent to which the program’s effects fade over time. The earnings results from MFIP also suggest that work requirements, which applied to ongoing recipients throughout the follow-up period, may provide a means of overcoming the negative income effects that arise from the combination of a strong financial incentive and relatively high benefit level.
The FIP program combined a fairly strong financial incentive with an education-focused welfare-to-work program, which may explain its results. The FIP evaluation showed that the program increased the fraction of participants who combined work and welfare, precisely as the standard economic model would predict. Moreover, this effect was similar in magnitude and significant for both recipients and applicants. However, the program had quite different effects on the rate at which it moved participants into the workforce and off welfare entirely, which is the other means by which it could have increased employment. Whereas the program had essentially no such effect on recent applicants, it significantly decreased the fraction of ongoing recipients who abandoned welfare for work. As a result, the program increased overall employment for both ongoing recipients and recent applicants, but the effect was significant only for recent applicants. One possible explanation for this pattern is that applicants were more likely to satisfy the work-related activity mandate by working, while ongoing recipients were more likely to satisfy it by taking part in the state’s welfare-to-work program. However, since the evaluation provides no information on welfare-to-work participation rates, this explanation is speculative.
Michigan’s TSMF program had small positive effects on employment, mirroring its small negative effects on welfare use. These effects may result from features of the program’s design. TSMF had a weak financial incentive, and its welfare-to-work program was not particularly work-focused during its first two years. When the program did adopt a work-first approach, the work-related activity mandate was applied to both the treatment and the control group (although the control group was subject to lesser sanctions for violating the work mandate). With little effective difference in the conditions applying to the treatment and control groups, we might expect to find small effects.
5.2.5. Programs That Focus on TANF-Like Bundles of Reforms
Most of the programs that included TANF-like bundles of reforms had positive effects on employment and earnings. The exception is Arizona’s EMPOWER program. Administrative data suggest that the program may have caused a slight decline in employment and earnings, whereas survey data show small increases. However, none of the estimates from EMPOWER are statistically significant. This could be the result of substantial confusion regarding the program’s time limit, which was its principal policy reform. Indiana’s Placement Track program significantly raised employment in the first year, but the effect faded over time. Its effects on earnings were larger and somewhat more persistent.
VIP/VIEW, ABC, FTP, and Jobs First generated increases in employment and earnings, most of which were significant. Most of the gains were fairly sizeable as well, particularly for earnings. The smallest effects stem from the VIP/VIEW program, which may be attributable to its phased-in evaluation design.
The pre— and post—time limit impacts of FTP and Jobs First are fairly similar. Both programs had positive and sizeable impacts on employment during the pre—time limit period. In FTP, the impacts were slightly larger in year three, whereas in Jobs First, they were slightly smaller. Their earnings impacts rose slightly in the immediate post—time limit period, but then fell by the fourth year. Neither employment nor earnings changed a great deal as recipients began to exhaust their benefits.
5.2.6. Evidence on Other Employment-Related Outcomes
Beyond providing evidence on the primary employment and earnings outcomes, a subset of the studies listed in Table 5.1 provide estimates on how the various programs have affected other employment-related outcomes, such as weekly hours, wages, job characteristics, and wage growth. However, many of those estimates are nonexperimental in nature, since the outcomes are functions not only of the recipient’s treatment status, but also of her post-randomization employment status. To illustrate this point, note that wage growth can be calculated only for persons working at two specified time points (unless we wish to attribute a zero wage to persons not working). However, persons employed at a point in time do not represent random subsamples of the treatment or control groups. Thus, simple comparisons of wage growth between the control and treatment groups does not in general provide estimates of the causal effect of the treatment. Moreover, some of the estimates are based on small survey samples, rather than the larger samples for which administrative records are available.
Partly because of these reasons, the results are not very conclusive. Although there are individual exceptions, there is no clear evidence from the random assignment studies that welfare reform increases wages, wage growth, or the fraction of workers who receive employer-provided health benefits, paid vacation, or paid sick leave. There is even less basis for discriminating between the effects of different types of reforms.
5.2.7. Subgroup Differences
Although only a subset of the random assignment studies provides subgroup analyses, the subset is larger in the case of employment and earnings than it was in the case of welfare use. These results from the available studies are discussed in Appendix A. These studies provide no clear evidence that any of the reforms consistently works to the benefit or detriment of relatively disadvantaged groups in terms of employment and earnings. Some of the results are mixed, however, and the number of studies on which this conclusion is based is generally small.
5.3. ECONOMETRIC STUDIES OF THE EFFECTS OF WELFARE REFORM ON EMPLOYMENT AND EARNINGS
Only five econometric studies estimate the impact of reform on employment and labor supply. As in the caseload literature, these studies typically use "modified dummy variables" to represent welfare reform policies. Therefore they provide moderate-quality evidence on the effects of those policies. Some studies represent some reform policies using measures that capture additional dimensions of policy variation, thus providing high-quality evidence on the effects of such policies.
The results of these studies, reported in Table 5.2, are fairly consistent. Four studies estimate the effect of reform as a bundle. O’Neill and Hill (2001) find reform to have positive and significant effects on the employment of single mothers, with the effects of TANF being stronger than the effects of waivers. Both Moffitt (1999) and Schoeni and Blank (2000) find waivers to have their largest effects among the least-educated women, increasing both their employment and their labor supply by about 4 to 5 percent. Schoeni and Blank report that TANF has similar effects on weeks worked but not on employment. Grogger (forthcoming) finds that reform increased both employment and labor supply among single mothers with infants, by about 4 and 8 percent, respectively, although those effects decrease with the age of the youngest child.
| Other controls | |||||||||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Study | Data | Sample population | Begin | End | Outcome | Dep. var. | Policy var. | Coeff. (s.e.) | % effect | Economy | Demogr. | Fixed Effects | Policy |
| A. Employment | |||||||||||||
| Schoeni and Blank (2000) | CPS aggregated | women 16-54, educ<12 | 76 | 98 | Fraction working in previous year | Level | Any waiver | 0.020 (0.007) | 3.7 | U, U-1, EG, & each *E | A, E, A*E, R | S, Y, state time trends, Y*E |
B, B*E |
| women 16-54, educ=12 | Any waiver | -0.005 (0.006) | -0.6 | ||||||||||
| women 16-54, educ>12 | Any waiver | -0.002 (0.005) | 0.2 | ||||||||||
| women 16-54, educ<12 | TANF | 0.015 (0.017) | 2.9 | ||||||||||
| women 16-54, educ=12 | TANF | 0.004 (0.015) | 0.5 | ||||||||||
| women 16-54, educ>12 | TANF | -0.024 (0.011) | -2.9 | ||||||||||
| Grogger (forthcoming) | CPS micro data | female headed families 16-54 | 78 | 99 | Employed last year | Level | Any reform (waiver or TANF) |
0.026 (0.012) | 3.7 | U | A, E, R Young child A, # kids |
S, Y, State time trends |
B, MW, EITC |
| Any reform * Age of youngest child |
-0.002 (0.001) | -0.3 | |||||||||||
| Modified time limit dummy | -0.014 (0.017) | -2.0 | |||||||||||
| Time limit dummy * (age*) | -0.003 (0.002) | -0.5 | |||||||||||
| O'Neill and Hill (2001) | CPS micro data | single mothers 16 - 54 | 83 | 2000 | Employed last week | Level | Any waiver | 2.34 *** | 3.9 | U, W, college wage premium | A, E, R, No. kids, age youngest kid, ever married, urban-rural | S, trend and trend squared, state-year trends | B |
| TANF | 6.59 *** | 11.0 | |||||||||||
| Meyer and Rosenbaum (2001) | CPS micro data | female family heads, 19-44 | 84 | 96 | Employed last year | Level | Modified dummy for any time limit waiver | 0.014 (0.007) | 2.3 | A, E, R # kids, any kids< 6,3,2,1 |
S, Y, month |
B, MW, EITC, Medicaid, training, child care | |
| Dummy for any time limit terminations | 0.022 (0.011) | 3.7 | |||||||||||
| Welfare benefit for working recipients (in $1,000s) | 0.077 (0.007) | 12.8 | |||||||||||
| B. Labor Supply | |||||||||||||
| Moffitt (1999) | CPS aggregated | women 16-54 | 77 | 95 | Annual weeks worked | Level | Any waiver | 0.3 (0.3) | 1.0 | U, U-1 | S, Y, State time trends | B | |
| CPS aggregated | women 16-54, educ<12 | 77 | 95 | Annual weeks worked | Level | Any waiver | 1.5 (0.6) | 5.0 | U, U-1 | A, E | S, Y, State time trends | B | |
| women 16-54, educ=12 | Any waiver | 0.5 (0.6) | 1.8 | ||||||||||
| women 16-54, educ=13-15 | Any waiver | 0.5 (0.6) | 1.6 | ||||||||||
| women 16-54, educ>16 | Any waiver | 0.2 (0.3) | 0.6 | ||||||||||
| Schoeni and Blank (2000) | women 16-54, educ<12 | Annual weeks worked | Any waiver | 0.732 (0.355) | 4.0 | U, U-1, EG, & each *E | A, E, A*E, R | S, Y, state time trends, Y*E |
B, B*E | ||||
| women 16-54, educ=12 | Any waiver | -0.175 (0.291) | -0.5 | ||||||||||
| women 16-54, educ>12 | Any waiver | 0.031 (0.239) | 0.1 | ||||||||||
| women 16-54, educ<12 | TANF | -0.090 (0.832) | -0.5 | ||||||||||
| women 16-54, educ=12 | TANF | 0.300 (0.728) | 0.9 | ||||||||||
| women 16-54, educ>12 | TANF | -0.377 (0.553) | -1.0 | ||||||||||
| Grogger (forthcoming) | CPS micro data | female headed families 16-54 | 78 | 99 | Annual weeks worked | Level | Any reform (waiver or TANF) |
2.425 (0.600) | 8.0 | U | A, E, R Young child A, # kids |
S, Y, State time trends |
B, MW, EITC |
| Any reform * Age of youngest child |
-0.216 (0.073) | -0.7 | |||||||||||
| Modified time limit dummy | -0.830 (0.923) | -2.7 | |||||||||||
| Time limit dummy * (age*) | -0.116 (0.098) | -0.4 | |||||||||||
| C. Earnings | |||||||||||||
| Moffitt (1999) | CPS aggregated | women 16-54 | 77 | 95 | Annual earnings | Level | Any waiver | 274.3 (161.8) | 2.8 | U, U-1 | S, Y, State time trends | B | |
| CPS aggregated | women 16-54, educ<12 | 77 | 95 | Annual earnings | Level | Any waiver | 87.8 (318.6) | 0.9 | U, U-1 | A, E | S, Y, State time trends | B | |
| women 16-54, educ=12 | Any waiver | 560.0 (318.6) | 5.7 | ||||||||||
| women 16-54, educ=13-15 | Any waiver | 441.4 (318.5) | 4.5 | ||||||||||
| women 16-54, educ>16 | Any waiver | 154.7 (318.7) | 1.6 | ||||||||||