ACF: OPRE Approaching Time Limits

Chapter 5

How Time Limits Affect Employment, Welfare Receipt, and Other Outcomes

Time limits are not designed simply to reduce long-term welfare receipt but also to change the behavior of current or potential welfare recipients — to encourage them to get jobs, hold jobs, or seek other sources of support instead of welfare. For example:1

The existence of a time limit might encourage potential welfare recipients to try harder to keep a job, change their living arrangements, delay childbearing, get married, or take other steps to avoid applying for benefits and using up months of eligibility;
Individuals who go onto welfare might try to find jobs and leave welfare more quickly, even before reaching the time limit, to bank some months for the future;
Individuals who reach a time limit and have their welfare benefits canceled might try harder to find or keep jobs, rely more heavily on other forms of public assistance, or take steps to reduce expenses.

The pattern of these effects may determine how time limits affect family income and material well-being. For example, if individuals respond to time limits by finding relatively well-paying jobs, they could end up better off financially; if not, they could end up with lower income and higher levels of material hardship. Of course, there could also be nonfinancial benefits or costs associated with relying less on welfare and more on other sources of support.

Key Findings

Because time limits have generally been implemented as part of a package of other welfare reforms, it is difficult to isolate their effects. Nevertheless, data from evaluations and econometric studies suggest several tentative conclusions:

There is some evidence that time limits can cause welfare recipients to find jobs and leave welfare more quickly, even before reaching the limit; however, the magnitude of this effect is not clear.
It does not appear that the cancellation of welfare benefits at a time limit induces many recipients to go to work in the short term.
Welfare reform initiatives with time limits have generated few overall effects on family income, material hardship, or household composition in the period after families began reaching the limits, although it is difficult to isolate the effects on families whose benefits were terminated.

In considering the implications of these results, it is important to note that none of them pertains directly to the 60-month federal time limit. Moreover, all the studies from which the data were drawn were conducted before the most recent recession began.

Measuring the Impacts of Time Limits

In general, the best way to measure the impact of a policy change such as a time limit is to conduct a random assignment study in which eligible individuals are assigned, by chance, to a group that is subject to the change (the program group) or to a control group that remains subject to the preexisting policies. Both groups are then followed over time, and any differences that emerge between them can reliably be attributed to the policy change being tested.

In fact, when states began to impose time limits under federal waivers in 1994 (see Chapter 1), they were required to conduct evaluations of this type, and several of the states elected to continue those studies after the 1996 federal welfare reform law passed. These random assignment studies provide some of the most reliable evidence about the effects of time limits. However, the studies are limited in several respects:

Almost all states imposed time limits as part of a “package” of reforms that also included expanded earned income disregards, broader work requirements, or other measures. Almost all of the studies were designed to measure the impact of the entire package, not to isolate the impact of the time limits.
In part, time limits (or other welfare reform measures) may affect people’s behavior by changing broad, community perceptions about welfare receipt. It is impossible to isolate a control group from this indirect but potentially important effect; as a result, the studies probably underestimate the effects of the reforms.2
The waiver evaluations tested the earliest time-limit programs, during a period when time limits were new and unfamiliar. The implementation components of the studies found that many recipients and staff were skeptical about whether the time limit would really be imposed. These perceptions might be different today.
None of the random assignment studies was designed to measure the impact of welfare reform or time limits on welfare applications. Thus, the studies provide little evidence about the first potential effect described at the beginning of the chapter.3

With these cautionary notes in mind, this chapter discusses the results of several random assignment studies of welfare reform programs that included some form of time limit. The key features of the programs and studies are summarized in Table 5.1. In general:

The Connecticut and Florida studies provide the most complete data at this point. Both of the programs included a benefit termination time limit, and both studies collected four years of follow-up data, measuring effects long after families began reaching the time limits. Both also collected data on the well-being of children. The Florida program was a relatively small pilot project, while the Connecticut program operated statewide (but was studied in two welfare offices).

The Delaware and Virginia programs also included benefit termination time limits, but the studies’ follow-up periods were cut short when the states decided to apply welfare reform rules to the control groups.4

Welfare Time Limits

Table 5.1

Selected Information About the Waiver Evaluations Discussed in This Chapter
State/Evaluation	Time Limit		Evaluation
State/Evaluation	Months	Type	Follow-Up	Child Impacts	Evaluator
Arizona	24	Reduction	2-3 yearsc	None	Abt Assoc.
Connecticut	21	Termination	4 years	Extensive	MDRC
Delaware	48a	Termination	2.5 yearsd	Some	Abt Assoc.
Florida FTP	24 or 36 b	Termination	4 years	Extensive	MDRC
Indiana	24	Reduction	2 yearse	Extensiveg	Abt Assoc.
Texas	12, 24 or 36b	Reduction	19 months	None	U. of Texas
Virginia	24	Termination	2-3 yearsf	None	Mathematica

SOURCES: Arizona: Kornfeld et al., 1999; Connecticut: Bloom et al., 2002; Delaware: Fein et al., 2001, and Fein and Karweit, 1997; Florida: Bloom et al., 2000; Indiana: Fein et al., 1998; Texas: Schexnayder et al., 1998; Virginia: Gordon and James-Burdumy, 2002.

NOTES:
aDelaware had a 48-month time limit when the study was conducted. In addition, recipients had to be working in order to receive assistance after 24 months of benefit receipt.(back)

bIn Florida and Texas, the length of the time limit depends on individual client characteristics.(back)

cEmployment impacts are reported for 10 quarters, and welfare impacts are reported for 36 months.(back)

dThe Delaware study reports 2.5 years of follow-up, but the results after the first year probably underestimate program impacts because the control group became subject to welfare reform policies.(back)

eThe Indiana study will eventually include 5 years of follow-up, but only 2 years are currently available.(back)

fThe Virginia study collected 3.5 years (42 months) of follow-up data, but this includes 3-18 months of data (depending on the site) from before welfare reform was implemented. Also, results in the last 15 months of follow-up underestimate program impacts because the control group became subject to welfare reform policies.(back)

gThe Indiana study includes a child impact analysis, but no data are available yet.(back)

The Indiana and Arizona time limits applied only to adults. The Indiana study will include long-term follow-up and a child impact study, but only two years of follow-up data are currently available. Two to three years of follow-up data are available for the Arizona study.
The Texas study was the only one designed to isolate the impact of a time limit. However, the Texas time limit applied only to adults, and only short-term follow-up data are available from the study.5

The chapter also discusses the results of other studies that do not use random assignment. Most of those studies take advantage of the natural variation in state welfare policies, examining the association between the timing or content of state policies and state welfare caseloads (and, in some cases, state-level data on employment) to estimate how much of the decline in the caseload was attributable to welfare reform. A few of the studies use individual-level data from national surveys. The studies attempt to control for other differences across states that may explain the caseload decline (for example, differences in economic conditions). A few studies try to isolate the impact of specific welfare reform provisions, including time limits.

A key advantage of these econometric studies is that they account for effects on both welfare exits and welfare applications. Also, in principle, they can measure impacts generated by changes in community perceptions of welfare that accompany the reforms. On the other hand, the studies usually rely on general information about state welfare policies, as opposed to data on how the policies are actually implemented. This can create a misleading impression of the policy environment in a particular state. In addition, the statistical methods used in these studies may or may not succeed in controlling for other factors that affect caseloads or employment.

Anticipatory Effects of Time Limits

Many people believe that the imposition of time limits played a key role in generating the large welfare caseload declines in the second half of the 1990s. Since few families actually reached a time limit during that period, these effects must have been anticipatory — that is, people must have left welfare more quickly (or decided not to apply for welfare) in order to avoid using up months of eligibility. Much of the evidence for this belief is anecdotal, but several eral studies have examined whether time limits generate anticipatory impacts on both employment and welfare receipt.6

Effects on Employment and Earnings

Table 5.2 shows results from five of the random assignment studies described earlier.7 The table focuses on the end of the first year after individuals entered the studies, before anyone had reached a time limit. The first column shows the percentage of program group members who were employed at that point; the second column shows the percentage of control group members who worked; and the third column shows the difference — the impact of the programs.

All five programs increased employment at the end of year 1.8 Although not shown in the table, most also increased average earnings. However, it is not clear what role the time limits played in generating these effects.9 As noted earlier, all the programs included other components that were also designed to boost employment. In the past, studies of welfare-to-work programs that included neither time limits nor enhanced earnings disregards have found similar effects on employment.10 It is also notable that the Connecticut, Florida, and Virginia programs, which included the most stringent time limits, did not consistently have the largest early impacts on employment.11

Welfare Time Limits

Table 5.2

Impacts on Employment at the End of Year 1 in Five Waiver Evaluations
State/Evaluation	Employed (%)		Difference
State/Evaluation	Program Group	Control Group	Difference
Connecticut	52.6	44.6	8.1 ***
Delaware	48.9	43.5	5.4 **
Florida FTP	45.2	40.8	4.3 *
Indianaa	57.6	50.0	7.6 ***
Virginiab
Lynchburg	57.9	48.8	9.1 **
Prince William	51.4	47.9	3.5
Petersburg	64.6	52.6	12.0 ***

SOURCES: Connecticut: Bloom et al., 2002; Delaware: Fein et al., 2001; Florida: Bloom et al., 2000; Indiana: Fein et al., 1998; Virginia: Gordon and James-Burdumy, 2002.

NOTES: In all studies, employment data come from unemployment insurance wage records.

A two-tailed t-test was applied to differences between outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; and *** = 1 percent.

aIndiana results are for sample members in the "placement track," who were subject to all welfare reform policies.(back)

bResults for Lynchburg and Prince William are for the fourth quarter after each county implemented welfare reform. Results for Petersburg are for the third quarter after implementation because the control group became subject to welfare reform in the fourth quarter. In each case, impacts are probably understated because some sample members had left welfare by the time the reforms were phased in.(back)

A few of the caseload studies described earlier estimated the effects of welfare reform on employment among single parents. Like the random assignment studies, most of the caseload studies concluded that the waiver programs increased employment. Results for the post-1996 period are more mixed. However, these studies generally did not attempt to sort out the effects of time limits on employment.12

Effects on Welfare Receipt

One might assume that effects on welfare receipt would simply be the converse of effects on employment — increases in employment would lead to decreases in welfare receipt. The reality, however, is more complex.

Random assignment studies. Tables 5.3 and 5.4 focus on the random assignment studies discussed in the previous section, showing effects on cash assistance receipt rather than employment. Table 5.3 shows the percentage of each group receiving welfare benefits at the end of the first year of follow-up. Table 5.4 shows, for several of the programs, the average number of months of benefits received in the period before program group members began reaching the time limits.13

The effects on welfare receipt are much more modest than the effects on employment. Most of the programs either increased welfare receipt or had no effect.14 At first glance, these results suggest that little or no “banking” was going on, but this is not necessarily the case. In fact, the pattern of welfare impacts is largely attributable to expanded earnings disregards and other policies that allowed a greater proportion of working recipients in the program groups to continuing receiving benefits; as a result, the programs increased the proportion of people who mixed work and welfare. The one program that substantially reduced welfare receipt — Indi-ana’s — did not have an expanded disregard.15 Of course, it is impossible to isolate the impact of the time limit in that case, and it is worth noting that Indiana’s time limit was designed in a way that did not provide an incentive for banking months of assistance.16 Also, the impact largely disappeared by the end of year 2.

Welfare Time Limits

Table 5.3

Impacts on Welfare Receipt at the End of Year 1 in Five Waiver Evaluations
State/Evaluation	Receiving Cash Assistance (%)		Difference
State/Evaluation	Program Group	Control Group	Difference
Connecticut	73.1	65.1	8.0 ***
Delaware	61.0	59.0	1.8
Florida FTP	56.6	54.4	2.2
Indianaa	43.3	52.6	-9.3***
Virginiab
Lynchburg	65.7	60.7	5.0
Prince William	38.1	40.7	-2.6
Petersburg	42.5	49.0	-6.5*

SOURCES: Connecticut: Bloom et al., 2002; Delaware: Fein et al., 2001; Florida: Bloom et al., 2000; Indiana: Fein et al., 1998; Virginia: Gordon and James-Burdumy, 2002.

NOTES: In all studies, employment data come from unemployment insurance wage records.

aIndiana results are for sample members in the "placement track," who were subject to all welfare reform policies.(back)

Welfare Time Limits

Table 5.4

Impacts on Cumulative Months of Pre-Time-Limit Benefit Receipt in Selected Waiver Evaluations
State/Evaluation	Months of Receipt		Difference
State/Evaluation	Program Group	Control Group	Difference
Connecticut (Months 1-21)	15.1	13.5	1.7 ***
Delaware (Year 1)	9.1	9.1	0.0
Florida FTP (Years 1-2)	11.9	11.7	0.0
Indiana (Years 1-2)	9.0	10.6	-1.6 ***

SOURCES: Connecticut: MDRC calculations; Delaware: Fein and Karweit, 1997; Florida: Bloom et al., 2000; Indiana: Fein et al., 1998.

NOTES: A two-tailed t-test was applied to differences between outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; and *** = 1 percent.

Data are drawn from state administrative records except in Delaware, where they are drawn from a survey.

The Delaware and Florida programs would almost certainly have reduced welfare receipt had it not been for their work incentive policies.17 Another waiver study found that a Min-nesota program that included work requirements and an expanded earnings disregard — but no time limit — increased welfare receipt.18 The fact that the Delaware and Florida programs had no effect suggests that some program features most likely time limits and/or sanctions induced people to leave welfare more quickly while the incentives encouraged them to stay on welfare longer, with the end result being a wash.19

A study that used data from the Florida FTP evaluation reached exactly that conclusion.20 This study used the impacts for people with no children under age 16 to isolate the effects of all components of the program other than the time limit. These individuals would have been required to leave welfare within two years regardless of their research group, so, in effect, there was no special time limit for the program group members. As expected, the study found that the other components of FTP increased welfare receipt and that the time limit decreased welfare receipt, especially for recipients with young children.

There is some evidence that the anticipatory effects of time limits may depend on the way the limits are implemented and how staff resolve the inherent conflict between time limits and earnings disregards. The authors of the Virginia study noted that, in the one county that generated decreases in welfare receipt before the time limit (Petersburg), staff strongly urged working recipients to leave welfare in order to save or bank their months. This was not the case in the other counties. In Connecticut’s program, which substantially increased welfare receipt in the pre-time-limit period, a banking message would not be credible given the generosity and structure of the disregard: Working recipients would give up $543 per month if they opted not to receive benefits. In Florida FTP, which generated no early impacts on welfare receipt, staff were quite likely to encourage recipients to use their months on welfare to obtain education or training, rather than urging them to bank their months.

Non-random assignment studies. As discussed in the first section of this chapter, many studies have used data on state welfare reform policies, economic conditions, and welfare caseloads to estimate the impact of the reforms on welfare receipt. Most of these caseload studies focused on the effects of waivers, but a few extended the analysis beyond 1996. Although the findings vary, most studies concluded that both welfare reform and the economic expansion contributed to the caseload decline. A study by the Council of Economic Advisors concluded that welfare reform explained about one-third of the caseload decline between 1996 and 1998 and a smaller proportion of the decline during the earlier waiver period.22

The fact that most of the econometric studies concluded that welfare reform reduced caseloads appears to contradict the results of the random assignment studies discussed earlier. There are several possible explanations for this discrepancy. First, it is possible that welfare reform affected caseloads primarily by reducing applications for assistance, an outcome that is not measured in the random assignment studies. Second, it may be that a large part of the impact of welfare reform was attributable to changes in community perceptions of welfare; as discussed earlier, it is impossible to fully isolate control group members from this effect. Third, welfare reform may have had relatively small impacts on caseloads in the states and counties where the random assignment studies were conducted. Finally, it is possible that the caseload studies overestimated the impact of welfare reform (as already stated, the random assignment studies probably understate those impacts).

Only a few of the econometric studies attempted to sort out the effect of individual components of welfare reform policies, including time limits. Most found that time limits had little or no impact on caseloads.23 However, a recent study by Grogger, using a different methodology, reached the opposite conclusion. Grogger used data from the Current Population Survey to test a theoretical model that predicts that families with the youngest children should be more responsive to time limits. In addition, unlike some of the other studies, in specifying the state policies Grogger used the date when the time limit was imposed rather than the date when families began to reach it (he was looking specifically at anticipatory effects rather than effects caused by benefit termination). Grogger concluded that time limits significantly reduced welfare receipt and that the impacts were strongest for families with young children. He estimated that time limits may have accounted for 16 percent to 18 percent of the decline in welfare use among female-headed families.24

Effects After Families Reach Time Limits

Regardless of whether people respond prior to reaching time limits, they may respond when their benefits are canceled — by going to work, taking steps to reduce expenses, or in other ways. One way to examine whether this happens is simply to follow people whose grants are canceled at a time limit. Chapter 6 discusses the results of several post-time-limit surveys that used this approach, and some of them found, for example, that employment rates grew slightly after people’s grants were canceled. However, if a study finds that some people go to work after their benefits are canceled, there is no way to know how many of them would have gone to work even if they had been allowed to stay on welfare; employment rates for any group of welfare recipients tend to increase over time.

Unfortunately, there is also no direct way to use results from a random assignment study to measure the impacts of benefit termination, because there is no way to know which members of the control group would have had their cases closed at the time limit had they been subject to one.25 In the absence of direct evidence, it is most useful to examine the pattern of overall program impacts during the period before and after families begin reaching the time limit. Conducting a similar analysis for subgroups of sample members who were particularly likely to reach the time limit may provide additional evidence.

Effects on Employment, Welfare, and Income

Figure 5.1 illustrates the Connecticut program’s effects on cash assistance receipt (top panel), employment (middle panel), and income (lower panel). Program group members started reaching the 21-month time limit in quarter 7 (as indicated by the vertical line). As expected, the top panel shows that, when families started reaching the time limit and having their benefits canceled, the impact on welfare receipt abruptly changed from positive to negative. In other words, the program increased welfare receipt before the time limit (for reasons discussed earlier) and reduced it afterwards.

The middle panel of Figure 5.1 shows a very different pattern for employment. In this case, the impact was relatively constant throughout the follow-up period, with no sudden change when families started to reach the time limit. In other words, there is no evidence that recipients responded to benefit termination by going to work. This is also not surprising because, as discussed in Chapter 4, most of the families whose benefits were canceled at the time limit in Connecticut were already employed. Although not shown in the figure, the program’s impact on earnings grew somewhat around quarter 7 — suggesting that some people may have increased their hours of employment after their benefits were canceled — but that impact then declined shortly thereafter.26

Welfare Time Limits

Figure 5.1

Connecticut's Jobs First Program: Quarterly AFDC/TANF Receipt, Employment, and Total Income
[D]

It is also worth noting that the effects on employment in Connecticut persisted throughout the four-year follow-up period; most evaluations of welfare-to-work programs without time limits found that effects diminished over time. The time limit may have something to do with Connecticut’s longer-lasting impacts; for example, perhaps former recipients tried harder to retain their jobs because they believed that returning to welfare was not an option for them.

The income effects (bottom panel of Figure 5.1) display still another pattern.27 In the pre-time-limit period, when the program increased both work and welfare, the program group had substantially higher income than the control group. After the time limit, the two groups had about the same income.28 This does not mean that people who reached the time limit lost no income (in fact, their income dropped sharply when their cash grants were closed) but, rather, that the program group, as whole, was no better or worse off than the control group. In fact, most of the people whose cases were closed at the time essentially lost the expanded earnings disregard (recipients could not receive an extension if they had income above the welfare payment standard). Thus, the income effects for the post-time-limit period look very much like the results of many welfare-to-work programs that did not include expanded disregards: Relative to the control group, the program group gained about as much in earnings as it lost in welfare, and its members ended up with about the same amount of income.

Figure 5.2 shows the same set of outcomes for the Florida FTP program (the vertical lines indicate the timing of the 24-month and 36-month time limits). The patterns are similar, although the impacts are less dramatic. In this case, there is a slight jump in the employment impact around quarter 8, when people started reaching the 24-month time limit — the FTP program granted few extensions — but that impact declined shortly thereafter (there was no such jump around quarter 12, when people began reaching the 36-month limit). As in Connecticut, the income effects changed from positive to neutral late in the follow-up period after many families had reached time limits.

Welfare Time Limits

Figure 5.2

Florida's Family Transition Program (FTP): Quarterly AFDC/TANF Receipt, Employment, and Total Income
[D]

Sometimes, data on average income hide the fact that some people gained income while others lost income. This could be particularly likely in the Connecticut and Florida studies, where only a fraction of program group members actually reached the time limit. In fact, both studies found some evidence that small groups of sample members may have lost income as a result of the welfare reforms. For example, in the last three months of follow-up, the Florida FTP program reduced the proportion of sample members with $1,500 to $3,000 in combined income from earnings and public assistance, and it increased the proportion with income below $1,500. Similarly, the Connecticut study found that, for the subgroup facing the most barriers to employment, the program slightly increased the proportion of sample members with income below $1,500. It is important to note, however, that both of these results only consider income measured in administrative records; no similar pattern was evident when income was measured using surveys (which, for example, include income obtained by other household members).

Finally, although not shown in the figures, the studies also provide evidence about whether these two time-limit programs affected Food Stamp receipt and payment amounts in the post-time-limit period. One might hypothesize that a reduction in cash assistance would lead to greater reliance on Food Stamps (although it might also be the case that some people are confused and believe that Food Stamps are also time-limited). In fact, neither program had significant effects on Food Stamp payments in the latter part of the follow-up period, although the Connecticut program decreased the number of people who received Food Stamps.

The Delaware and Virginia studies also report limited impact results from the post-time-limit period; there are few indications that the imposition of time limits caused a jump in employment impacts. In Delaware, employment impacts disappeared during year 2 and did not reemerge when recipients began to encounter the 24-month work requirement. Welfare impacts persisted after the second year, probably driven in large part by the high sanctioning rate discussed earlier. Employment impacts also declined in two of the three Virginia study sites, although this may be because the control group was phased into the welfare reform program.

Effects on Material Well-Being and Other Outcomes

Both the Connecticut and Florida studies administered extensive surveys to the program and control groups well after people started reaching the time limits (the survey took place 36 months after study entry in Connecticut and at the 48-month point in Florida). Both surveys included many measures of material well-being, hardship, household composition, and other outcomes. As with the other results, the survey results cannot be used to isolate the impacts of time limits. However, if the time limits generated substantially negative (or positive) impacts for the people who reached them, it seems likely that this would show up in either the overall results or the results for subgroups that were particularly likely to reach the time limits.

As shown in Table 5.5, neither of the programs generated consistent effects on material well-being or hardship for the full study sample; the same was true for key subgroups (subgroups results are not shown). The Connecticut program had both positive and negative impacts, while the Florida program had a few small positive effects. Similarly, the programs generated no impacts on marriage or fertility and few effects on household composition.

Interestingly, both the programs increased child support receipt. Although certainly plausible — custodial parents may have tried harder to pursue support in the absence of welfare — these results should be considered with caution. At the time of the surveys, program group members were more likely to be off welfare and thus, perhaps, more likely to be aware of how much support was being collected: When custodial parents receive cash assistance, support payments are mostly retained by the state as reimbursement for welfare costs.

Effects on Children

As noted earlier, the Connecticut and Florida FTP studies both collected extensive survey data on the well-being of respondent’s children. Most of these data were reported by parents, but the Connecticut study also included a small teacher survey. Both studies found few effects for elementary-school-age children, the age group for whom the most complete data were collected.

Both programs appear to have generated some negative effects for adolescent children (the Connecticut program generated both positive and negative effects). Once again, however, there is little evidence that these effects were driven by the time limits. Such effects have appeared in other studies of programs that did not include time limits, including programs that increased family income.29

The Delaware study used administrative records to examine effects on child neglect, abuse, and foster care placements. The welfare reform program increased the fraction of families with an incident of child neglect, and these effects were concentrated among the most disadvantaged sample members. The program had no impact on other kinds of maltreatment or on foster care placements. These effects occurred both in the pre-time-limit period, when a large proportion of the program group experienced full-family sanctions, and after families began reaching the 24-month work trigger.30

Finally, a recent study used state-level child welfare data to examine the association between welfare reform and child maltreatment; the methodology was similar to that used in the caseload studies discussed earlier. The study found an association between short time limits and increases in measured child maltreatment and the number of children in out-of-home care.31 The Connecticut and Florida FTP evaluations did not analyze child welfare data. However, survey data in both studies showed no effect on the proportion of sample members with a minor child living outside their household.

Welfare Time Limits

Table 5.5

Impacts on Selected Measures for Connecticut's Jobs First Program and Florida's Family Transition Program
	Program Group (%)	Control Group (%)	Difference	Program Group (%)	Control Group (%)	Difference
	Connecticut			Florida FTP
Lives with other adults	44.9	42.4	2.5	53.4	53.4	0
Married, lives with spouse	9.1	10.8	-1.6	17.2	19.1	-1.9
Gave birth since Random Assignment	20.7	20.7	0.1	23.9	22.7	1.2
Receives child support	25.7	22.7	3 *	29.5	21.9	7.6 ***
Food Insecurea	38.7	40.2	-1.5	34.1	35.8	-1.8
Owns a car	40.9	36.7	4.2 **	59.1	60.2	-1.1
Has debt	64.6	60.1	4.6 **	67.4	67.1	0.3
No health insurance	13.9	18.4	-4.4 ***	39.3	38.4	0.9
In prior year:
Phone disconnected	26.3	27.3	-1	33.5	31.5	2
Utilities shut off	18.5	21.9	-3.4 **	15	15.6	-0.6
Ever homeless	2.6	1.5	1.1 *	3.7	4.9	-1.1
Ever evicted	6.4	7.1	-0.6	6.5	6.3	0.1
Neighborhood problemsb
None	35.5	29.4	6 ***	32.9	33.7	-0.8
1-3	39.8	45.8	-6 ***	49.9	45.3	4.6 *
4 or more	24.7	24.7	0	17.2	21	-3.8 *
Housing problemsc
None	63.4	60.5	2.9	64.1	60.8	3.3
1	18.9	21.4	-2.5	21.8	20.8	1
2 or more	17.7	18.1	-0.4	14.1	18.4	-4.3 **
At end of month, usually:
Some money left over	14.3	17.1	-2.8 *	22.3	20.5	1.8
Just enough	42	41.1	0.9	46.7	42.5	4.3 *
Not enough money	43.7	41.8	1.9	30.9	37	-6 ***

SOURCES: Published survey data (Connecticut: Bloom et al., 2002; Florida FTP: Bloom et al., 2000).

NOTES: The data were collected three years after random assignment in Connecticut and four years after random assignment in Florida.

aThe USDA-recommended six-item food security scale was used to measure food security. The items in the scale include questions about food consumed and the kinds of things people resort to when money allocated for food is exhausted. The scale ranges from 1-6, and two or more affirmatives indicate food insecurity.(back)

bNeighborhood problems include the following: unemployment; drug users or pushers; crime, assault, or burglaries; run-down buildings and yards; and noise, odors, or heavy traffic.(back)

cHousing problems include the following: leaky roof or ceiling; broken plumbing; broken windows; electrical problems; roaches/insects; heating system problems; and broken appliances.(back)

1See Moffitt and Pavetti (2000) for a discussion of the theoretical framework for considering the potential effects of time limits.(back)

2In fact, in all the random assignment studies, some control group members reported in surveys that they believed they were subject to time limits. For example, in the Connecticut Jobs First evaluation, 23 percent of control group members reported that they were subject to a time limit; the corresponding figures were 29 percent in the Florida FTP evaluation and 66 percent in the Delaware ABC evaluation.(back)

3The Connecticut and Florida evaluations asked program group members whether they agreed with a series of statements about how the time limit had affected their behavior. About 40 percent of respondents in Florida either agreed a little (15 percent) or agreed a lot (25 percent) with the statement “Because of the time limit, I decided not to apply for welfare at a time when I could have applied.” About 35 percent agreed with the same statement in Connecticut.(back)

4In Delaware, the analysis focuses on individuals randomly assigned from October 1995 to Septem-ber 1996 (most were randomly assigned by March 1996) and presents 2.5 years of follow-up for each person. However, the control group was phased into the welfare reform program beginning in March 1997. Thus, for the most part, results for the first year of follow-up fully capture the impacts of the welfare reform, while results for the second year and beyond do not. In Virginia, all sample members were randomly assigned in July 1995, and data are available through December 1998 (42 months). However, the welfare reform program began at a different time in each of the three main study counties (October 1995 in Lynchburg, April 1996 in Prince William, and January 1997 in Petersburg), and the state began phasing the control group into the welfare reform program in October 1997. As a result, the available post-welfare reform follow-up ranges from two years in Petersburg to a little more than three years in Lynchburg, and the last 15 months of data do not fully capture the impact of the welfare reform. (back)

5Another random assignment study, in Vermont, was designed to isolate the added impact of a time-triggered work requirement that was initially referred to as a time limit.(back)

6The extent to which people will respond in anticipation of time limits depends on their discount rates and liquidity constraints — that is, the relative value that people place on short-term versus long-term gains and their perception of the alternatives to welfare. For example, if current or potential recipients believe that they have few alternatives to welfare, they will be less likely to bank months. See Moffitt and Pavetti, 2000.(back)

7Results for the Arizona study are not included because survey data showed that few program group members were aware of the time limit and that a roughly equal proportion of control group members thought that they were subject to the limit. Thus, the study does not appear to provide a fair test of the anticipatory effects of a time limit. The Texas results are discussed below.(back)

8The authors of the Virginia study believe that employment impacts may be understated in Prince William, the one site that did not generate statistically significant gains. This is because many county residents work for the federal government and such jobs are not covered in the unemployment insurance wage records used in the analysis.(back)

9The Texas study, which was designed to isolate the impact of a time limit, did not find any early impacts on employment. However, the implementation study notes that many caseworkers did not actively discuss the time limit and that staff had difficulty maintaining the distinction between the research groups. Both the program and the control groups eventually became subject to a 60-month time limit.(back)

10In fact, it is difficult to make direct comparisons between the waiver studies discussed in this chapter and earlier studies of welfare-to-work programs. In the earlier studies, the control groups typically were not required to participate in any employment-related activities. In the waiver studies, the control groups were subject to the state policies that existed before the waiver programs began. In most states, those preexisting policies included at least some employment-related requirements. In effect, the waiver evaluations measure the impact of the 1990s reforms over and above the impacts of earlier reforms.(back)

11In some studies, the employment impacts changed as program group members drew nearer to the time limit, but there is no clear pattern in these results. In Delaware and Indiana, the employment impacts were smaller at the end of year 2 than at the end of year 1; in Florida FTP, the impacts grew somewhat larger during that period; in Connecticut, they remained roughly constant over time; and in Virginia, the patterns varied by county.(back)

12 See Blank (2001) for a summary of these studies.(back)

13These data are only available for the first year in Delaware.(back)

14As noted earlier, results for the Arizona and Texas projects are not included in the tables. Neither program generated impacts on cash assistance receipt in the pre-time-limit period.(back)

15During this period, the Indiana program used a “fixed grant” policy: Normal AFDC disregards were applied when a recipient went to work, but the grant was then frozen to provide an incentive for advancement. Indiana later implemented an expanded disregard.(back)

16Initially, Indiana’s time limit counted calendar months rather than months of benefit receipt. As a result, there was no way for a recipient to stop the clock by leaving welfare. Also, one might assume that the incentive to bank months would be weaker with a reduction time limit than with a termination time limit.(back)

17The Florida program disregarded $200 plus half of any remaining earnings in calculating recipients’ monthly grants. The Delaware program used “fill-the-gap” budgeting, another policy that allows people to earn more without losing their full welfare grant. In Virginia, recipients could keep their entire grant as long as their total income from TANF and earnings did not exceed the federal poverty level. There is, of course, no way to know whether the employment impacts would have been smaller without the work incentives.(back)

18Knox, Miller, and Gennetian, 2000.(back)

19Interestingly, substantial impacts on welfare receipt emerged in year 2 in Delaware, although the study authors attribute these impacts to sanctions rather than to the time limit. Nearly one-fifth of the program group experienced a full-family sanction in year 2 alone, and other families were probably induced to exit before a full-family sanction was actually imposed. The Florida FTP program did not use full-family sanctions during the early years of the study period. It generated decreases in cash assistance payments during year 2 but had no impacts on receipt rates until after families began reaching the time limit.(back)

20Grogger and Michalopoulos, 2001.(back)

21The Vermont study mentioned previously found that a time-triggered work requirement that took effect after 30 months of welfare receipt generated increases in employment and reductions in welfare receipt even before anyone was required to work. In other words, people appear to have responded in anticipation of the work requirement (which was referred to as a time limit). See Hendra and Michalopoulos, 1999.

22Council of Economic Advisors, 1999.(back)

23For example, Council of Economic Advisors (1999) found that “through 1998, time limits had not significantly altered national caseloads.” A similar result was found by Hofferth, Stanhope, and Harris (2001), in a study that used data from the Panel Study of Income Dynamics (PSID) to examine the effect of welfare waivers on work exits.(back)

24Grogger, 2000.(back)

25One could compare program group members whose benefits were terminated with control group members who received enough months of benefits to reach the time limit. However, a different group of control group members might have stayed on welfare if they had been subject to the welfare reform. In addition, this method is particularly problematic in states where many people receive exemptions or extensions; it is impossible to predict which control group members would actually have had their benefits cut off.(back)

26The earnings impact was $149 in quarter 7, $198 in quarter 8, and $219 in quarter 9. However, the impact then declined and was no longer statistically significant by quarter 12. In fact, there is no way to tell whether earnings growth is caused by higher hourly wages, more hours of work per week, or more weeks of work in a quarter. However, it seems unlikely that many recipients could have responded to benefit termination by quickly finding a higher-paying job.(back)

27This is not a full measure of household income. It includes only the study sample member’s cash assistance, Food Stamps, and UI-covered earnings.(back)

28The impact on income persisted for a few months after families began to reach the time limit. This is probably related to the temporary increase in the earnings impact discussed earlier. In effect, the larger earnings impact temporarily offset the welfare decrease.(back)

29 Gennetian et al., forthcoming, 2002.(back)

30 Fein and Lee, 2000;(back)

31 Paxson and Waldfogel. 2001.(back)

Table of Contents | Previous | Next