Search Type Title Author/Investigator Search Advanced Search

Table of Contents | Previous | Next

Chapter 4: The Effects of the Texas ERA Program on Employment, Public Assistance, and Income

This chapter uses administrative records to examine whether the Texas Employment Retention and Advancement (ERA) program resulted in better job-finding, employment retention, and advancement outcomes than Choices, the state’s standard welfare-to-work program for recipients of Temporary Assistance for Needy Families (TANF). Administrative records are also used to determine whether the additional services and incentives offered by ERA relative to Choices have had any effect on public assistance receipt or total income. As noted previously, control group members were not eligible for ERA services but could receive services through Choices and through other programs and agencies in their area.

Using data from the ERA 12-Month Survey, this chapter examines whether ERA increased the percentage of sample members who found jobs that had better characteristics, such as higher wages and more fringe benefits. Findings are presented for the full report sample, for an early cohort, and for subgroups. As in Chapter 3, all the findings are presented separately by site: Corpus Christi, Fort Worth, and Houston.1 For the full report sample, two years of follow-up are available for outcomes created from unemployment insurance (UI) wage data, and six quarters of follow-up are available on measures of TANF, food stamp receipt, and total measured income.

Key Findings

• The Texas ERA program did not produce consistent or large effects on employment outcomes during the first two years of the study. During the two-year follow-up period, the ERA programs in Corpus Christi and Fort Worth led to increases above the control group’s average on some measures of employment and employment retention. However, differences on most outcomes are small and not statistically significant. In Houston, the program had no effect on employment and earnings.

• ERA group members in Corpus Christi received more in combined income from earnings, public assistance, and the monthly stipend (discussed below) than control group members received from earnings and public assistance alone. There were no effects on combined income in Fort Worth or Houston.

• In Corpus Christi and Fort Worth, the two-year impacts were concentrated among sample members with recent employment history prior to random assignment –– people who were more likely to work after random assignment and (among ERA group members) more likely to receive the stipend. ERA did not raise employment or earnings above control group levels for sample members with no recent employment.

The Expected Impacts of the ERA Program

A program like ERA can increase job-finding, employment retention, and advancement in a number of ways. The ERA stipend, which was available after four months of employment for those working a minimum of 30 hours per week,2 would be expected to lead to improved retention and advancement outcomes and to higher total income. The $200 per month stipend was a source of income that was available only to the ERA group, and it could increase total income as much as $2,400 for a one-year period. According to the Texas Workforce Commission (TWC), the goal of the stipend was to provide a work support that would enable welfare leavers to resolve problems that might prevent them from maintaining employment, such as difficulty covering child care, transportation, and other job-related expenses. Thus, if this goal were achieved, ERA group members should have better employment retention outcomes.3 Also, by “making work pay,” the stipend should provide an incentive for TANF recipients to work. The stipend would motivate an individual to look harder for a job or, even, to accept a job that one might otherwise forgo. That said, the maximum welfare grant in Texas was rather low,4 so the incentive to work in Texas (regardless of the ERA stipend) was already stronger than in higher-grant states.

The 30-hour minimum work requirement might encourage some ERA group members to upgrade from part-time to full-time employment if the stipend were marketed early and often prior to employment.5

ERA’s preemployment services might not generate impacts because, as discussed in Chapter 2, the ERA group is being compared with a control group that was engaged in a well-established welfare-to-work program (Choices) that strengthened throughout the study period. Thus, the impact analysis measures the value added of ERA over a substantial and well-developed system of services and supports with a strong employment focus. While the participation analysis found important differences between the ERA and control groups, these participation differences are not large compared with similar programs analyzed in the past6 –– which raises the possibility of small effects on employment, retention, and advancement. However, the fact that ERA is being compared with a strong control group program should be helpful for states that already have strong welfare-to-work programs and that are deciding whether it is worthwhile to invest in Texas’s mix of postemployment services and a stipend.

ERA’s postemployment services were designed to help former welfare recipients stay employed and increase their wages and benefits over time. Such postemployment services as employer site visits and monitoring and supportive services should promote improved retention and advancement outcomes. These improvements would likely translate into a reduction in welfare use and recidivism. Retention would also be expected to be promoted through more frequent contact with ERA staff (and with the attendant work supports, problem solving, and reemployment services). Increases in retention might be expected to emerge later in the first year. Impacts on advancement, however, often require promotions or job-changing, which takes time. Impacts on advancement may not be detected until Year 2 or later.

As discussed in Chapter 1, random assignment in Texas took place at the time of TANF application or recertification. Thus, the study sample includes individuals whose applications were denied and who never received TANF, the stipend, or ERA employment services –– which may weaken the effects of the program. Also, to a lesser extent, the inclusion of exempt individuals might weaken the impacts, since their participation in ERA was voluntary.7

Data Sources and Samples

Unemployment insurance (UI) wage data, public assistance payment records, and ERA program tracking data are the primary data sources for creating outcomes of employment, earnings, TANF, food stamps, and stipend receipt and for estimating impacts on these outcomes. Quarterly earnings records are available for three years prior to random assignment and two years after random assignment for a total of 4,288 sample members (2,137 in the ERA group and 2,151 in the control group), who were randomly assigned from October 2000 through June 2002.8 This represents three-quarters of the eventual sample that will be analyzed in Texas.9 For this same cohort, monthly public assistance records are available for two years prior to random assignment and only six quarters after random assignment.10 In this report, “Year 1” refers to the first through fourth quarters following the quarter of random assignment. “Quarter 1” refers to the quarter of random assignment. Because Quarter 1 contains some earnings, TANF payments, and food stamp payments from the months and weeks immediately preceding random assignment, it is excluded from the summary measures of the first year of follow-up.11

The UI wage data are a reliable source for estimating employment and earning impacts because UI wage records are stored in computerized systems shortly after the completion of a quarter and most employers are required to submit them. UI records do, however, miss wages not reported to the UI system in Texas,12 and they do not measure job characteristics. For these reasons, data from the ERA 12-Month Survey are also used. However, UI wage records are reliable for jobs that are covered by the UI system.

For the ERA 12-Month Survey, MDRC selected a random sample of adults in single-parent families who spoke either English or Spanish and who were randomly assigned from January through June 2002 in Corpus Christi and Houston and from September through December 2002 in Fort Worth. As noted in Chapter 1, the survey was administered to 775 sample members across the three sites, approximately 12 months following random assignment, and it achieved response rates of 82 percent in Corpus Christi, 75 percent in Fort Worth, and 80 percent in Houston. (See Appendix F for further details on the survey response analysis.) The survey also has limitations. Individuals may recall incorrectly or may misreport some of the outcomes. Because the survey sample is smaller than the administrative records sample, results from the survey are less reliable. It is therefore more difficult to detect statistically significant impacts among the survey samples. Furthermore, because the survey samples were drawn from only six months of the random assignment period in Corpus Christi and Houston — and from only three months in Fort Worth — the survey findings may not be generalizable to the entire report sample. In particular, the Fort Worth survey sample and report sample do not overlap at all.

Impacts for the Full Report Sample

Table 4.1 summarizes the impacts of ERA in each of the three sites in which the program was evaluated.13 Impacts are presented for two years of follow-up on measures of employment and earnings that were created from UI wage records. The set of three columns at the left of the table shows the outcomes and impacts in Corpus Christi. The first column shows the average value for each outcome for the ERA group, and the second column shows the average value for each outcome for the control group. The control group outcomes represent the benchmarks against which the ERA program is being compared. Because of random assignment, the control group outcomes represent what would have been expected for ERA group members if only Choices and not ERA had been implemented. The third column in the set shows the effects, or impacts, of the ERA program in Corpus Christi. Impacts are calculated as the difference in average outcomes between the ERA group and the control group.14 The presence or absence of “stars” (asterisks) in the impact column indicates whether or not a difference is statistically significant.15 Since random assignment ensures that there are no systematic differences between the ERA and control groups, other than exposure to the program being studied, any statistically significant differences in outcomes after random assignment can be attributed to the ERA program. The sets of columns in the middle and at the right of Table 4.1 follow the same format but show the impacts of ERA in Fort Worth and Houston, respectively.

The Employment Retention and Advancement Project

Table 4.1

Years 1-2, Impacts on UI-Covered Employment and Earnings

Texas

Outcome		Corpus Christi			Fort Worth			Houston
		ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)
Years 1-2	Ever employed (%)	82.4	84.7	-2.3	80.7	76.4	4.4 *	72.5	72.1	0.5
	Average quarterly employment (%)	53.3	50.5	2.8	48.5	46.5	2.0	42.6	42.1	0.5
	Employed 4 consecutive quarters (%)	48.9	45.0	3.9	44.6	41.5	3.1	37.2	36.3	0.9
	Earnings ($)	8,599	8,088	512	9,802	9,206	595	8,269	8,299	-29
	Earned over $20,000 (%)	13.8	11.6	2.2	17.6	14.6	3.0	14.8	14.6	0.3
Year 1	Ever employed (%)	73.5	74.0	-0.5	69.1	67.3	1.8	64.1	63.6	0.5
	Average quarterly employment (%)	53.3	49.8	3.5 *	48.8	47.3	1.5	42.5	43.4	-0.8
	Employed 4 consecutive quarters (%)	30.7	26.4	4.3 *	26.3	25.6	0.7	21.4	22.8	-1.4
	Earnings ($)	3,940	3,593	347	4,443	4,283	160	3,790	3,863	-73
	Earned over $10,000 (%)	11.0	11.1	-0.1	16.5	14.8	1.6	12.6	13.6	-1.0
Year 2	Ever employed (%)	71.1	70.3	0.8	68.8	62.7	6.1 **	59.5	57.7	1.8
	Average quarterly employment (%)	53.3	51.1	2.1	48.2	45.8	2.4	42.8	40.9	1.9
	Employed 4 consecutive quarters (%)	33.8	33.0	0.9	27.3	27.9	-0.6	25.4	24.3	1.1
	Earnings ($)	4,659	4,495	164	5,359	4,923	435	4,480	4,436	44
	Earned over $10,000 (%)	17.5	16.6	0.9	22.0	19.6	2.3	17.1	17.0	0.2
Last quarter of Year 2	Ever employed (%)	54.2	49.1	5.1 *	48.6	44.4	4.2	40.1	41.5	-1.4
	Total earnings ($)	1,210	1,146	64	1,452	1,212	240 **	1,088	1,175	-88
	Earned $2,500 or more (%)	20.9	19.6	1.3	24.4	20.2	4.2 *	18.0	20.7	-2.7
	Earned between $500 and $2,499 (%)	23.0	22.8	0.1	16.5	16.8	-0.3	16.0	14.7	1.3
	Earned between $1-$499 (%)	10.3	6.7	3.7 **	7.7	7.4	0.3	6.1	6.1	-0.1
Sample size (total = 4,288)		654	652		578	586		905	913

SOURCES: MDRC calculations from UI records from the State of Texas. NOTES: See Appendix B. This table includes only employment and earnings in jobs covered by the Texas unemployment insurance (UI) program. It does not include employment outside Texas or in jobs not covered by UI (for example, "off-the-books" jobs, some agricultural jobs, and federal goverment jobs).

Benchmark Employment, Retention, and Earnings Outcomes

According to data from the baseline information form, nearly 94 percent of study participants were unemployed at the time of random assignment. Despite this high percentage, most control group members in Texas worked at some point after random assignment. As shown in Table 4.1, in Corpus Christi, nearly 85 percent of control group members worked in a UI-covered job sometime in the first two years after random assignment. Control group members were somewhat less likely to work in Fort Worth (76 percent) and Houston (72 percent).

One of the main goals of ERA was to promote employment retention. A review of employment retention outcomes for the control group provides compelling evidence of the need for an intervention like ERA. While some employment was common, few control group members were stably employed. Across the sites, between 42 percent and 50 percent of the control group worked in UI-covered jobs in any given quarter.16 Among control group members, 20 percent who were employed in Quarter 2 did not work in Quarter 3 (not shown). Only about 35 percent of control group members who were working in Quarter 2 worked in each of the quarters through the last quarter of Year 2. Though many Texas sample members eventually found jobs, overall employment rates in Texas were no higher in Year 2 than Year 1.

Earnings among control group members were low for all three sites and throughout the follow-up period. This is partly attributable to low employment rates. (Table 4.1 and all tables in the report include zeroes for those who were not working, unless otherwise specified.) However, i n any given quarter, employed control group members in Corpus Christi earned an average of only $2,000 (not shown).

Two-Year Impacts on Employment, Employment Retention, and Earnings

The first section of Table 4.1 shows the impacts over the first two years after random assignment. This table, like most of the tables in this report, shows little evidence that ERA made a difference in the employment and earnings outcomes of sample members.

The ERA program in Corpus Christi did not produce any statistically significant effects on employment, employment retention, or earnings over the two-year follow-up period. ERA had no effect on the percentage ever employed. The program did not increase the percentage of ERA group members who were employed above the control group’s benchmark of 84.7 percent. The lack of an effect on employment is likely attributable to the fact that, as implemented, the ERA program was similar to the Choices program during the preemployment phase. This shows that the stipend did not induce individuals to work who would not have worked anyway. During Years 1 and 2, the ERA group in Corpus Christi did not earn significantly more than the control group’s average of $8,088.

Table 4.1 shows that, during Year 1, ERA produced increases in measures of employment retention in Corpus Christi. ERA increased average quarterly employment by nearly 4 percentage points above the control group’s average of approximately 50 percent. ERA increased the proportion of program group members who were employed in four consecutive quarters — a key measure of retention — by about 4 percentage points above the control group’s average of 26 percent. Despite the increases in employment stability, the program did not increase earnings. Given the variability in earnings, it is occasionally the case that effects on employment rates are statistically significant while effects on earnings are not.17

In Year 2, the impacts on measures of employment retention were no longer statistically significant. This contradicts the expectation that the impacts would be stronger later in the follow-up period and that they would vary based on stipend receipt. In Corpus Christi, only about 16 percent of the ERA group received a stipend in Year 1 (not shown). In Year 2, approximately 24 percent of the ERA group received a stipend, but the impacts on employment retention measures were no longer significant.

The impacts on quarterly employment and earnings are shown in Appendix Table E.5. In Corpus Christi, there are several quarters in which employment differences, although positive, are not statistically significant. Table 4.1 shows, however, that impacts on quarterly employment emerged late in Year 2 in Corpus Christi. In Quarter 9, ERA increased the percentage employed in Corpus Christi by 5.1 percentage points; however, much of this effect on employment was in jobs with very low earnings (less than $500 per quarter). These low earnings may be attributable to part-time jobs, low wages, or short-term employment.18 As a result, ERA group members did not earn more, on average, than control group members. It is difficult to determine whether this late effect is a temporary phenomenon, but the fact that none of the impacts in the previous quarters are close to significant suggests that this impact may not persist.19

In Fort Worth, nearly 81 percent of ERA group members were ever employed over the two-year follow-up period, which is 4 percentage points higher than the control group’s average. This impact is consistent with the participation results, discussed in Chapter 3, which show that ERA increased participation in group job search activities in Fort Worth (which was not the case in Corpus Christi). This impact may also suggest that, unlike in Corpus Christi, the stipend did have a modest effect on employment in Fort Worth. There were no other significant impacts on employment or earnings over the two-year follow-up period in Fort Worth. Over that period, average earnings for the two research groups were approximately $9,500.

In Year 1, there were no impacts in Fort Worth on any of the measures of employment and earnings shown in Table 4.1. The weak results in Year 1 may be due to the startup problems in Fort Worth (noted in Chapter 2). In contrast to the pattern of impacts in Corpus Christi, the impacts in Fort Worth were stronger in Year 2, when nearly 69 percent of ERA group members were employed –– 6 percentage points more than control group members. This impact emerged because the employment rate among ERA group members was the same in Year 2 as Year 1 while the percentage of control group members ever employed in Year 2 declined. However, there were no effects on measures of job retention or earnings.

By the end of Year 2, the impacts in Fort Worth were stronger than they were earlier in the follow-up period. Though ERA no longer had a statistically significant impact on the employment rate, the program produced a statistically significant $240 increase in earnings in the last quarter of Year 2. ERA increased the percentage of sample members who had earnings of $2,500 or more –– an important measure of advancement. While these results are promising, it is impossible to know whether they represent a statistical anomaly or a real trend. Future reports will provide more definitive results.

In Houston, ERA had no effect on measures of employment, retention, or advancement. Over the two-year follow-up period, ERA group members in Houston earned $8,269, which is nearly the same as control group members earned. The two groups’ rates of employment and employment retention were nearly identical over the follow-up period. In any given quarter, approximately 42 percent of ERA and control group members were employed, which is somewhat lower than in Corpus Christi and Fort Worth. The consistent lack of impacts for several measures, subgroups, and cohorts suggests that it is unlikely that full-sample impacts will emerge in Houston.

Impacts on Stipend Receipt, Public Assistance, and Income

As discussed, TANF and food stamp data are available only through January 2004, and so two fewer quarters of follow-up are available from these sources. For this reason, the measures in the upper panel of Table 4.2 cover six quarters after random assignment (one and a half years) rather than two years.20 Table 4.2 shows that, over the follow-up period, ERA group members in Corpus Christi received more in combined income from earnings, public assistance, and the stipend than control group members received from earnings and public assistance alone. Receipt of the stipend accounts for most of the difference. There were no effects on combined income in Fort Worth or Houston.

Approximately 82 percent of control group members in Texas ever received TANF in the first six quarters. This is important, since ERA’s pre- and postemployment services –– and the stipend –– were available only to those who received TANF. Although the percentage ever receiving welfare was high, the rate declined rapidly during the follow-up period.

Across the three sites, between 92 percent and 97 percent of control group members received food stamps at some time in the follow-up period. While TANF receipt rates declined rapidly, the food stamp receipt rates were more stable: Approximately 70 percent of sample members were still receiving food stamps in Quarter 7. Thus, for this sample, food stamps were a much more important source of income than TANF payments.

Total measured income includes income from earnings, TANF payments, food stamp payments, and stipends (for ERA group members only). Despite the unstable employment and low earnings of control group members in Texas, earnings were the primary source of income. On average, approximately 53 percent of total income was derived from earnings; 34 percent came from food stamps; and TANF provided only 13 percent. Total income among control group members over the six-quarter period varied from $11,247 in Corpus Christi to $12,227 in Fort Worth.21

The Employment Retention and Advancement Project

Table 4.2

Impacts on Public Assistance and Measured Income

Texas

Outcome		Corpus Christi			Fort Worth			Houston
		ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)
First 6 quarters after random assignment	Earnings ($)	6,197	5,772	425	6,978	6,782	195	6,066	6,030	36
	Ever received TANF (%)	83.9	82.1	1.9	83.7	81.8	1.8	87.2	85.3	1.9
	Amount of TANF received ($)	1,363	1,391	-28	1,555	1,579	-24	1,729	1,630	98 *
	Ever received food stamps (%)	96.1	96.7	-0.6	94.1	92.4	1.7	92.6	93.4	-0.8
	Amount of food stamps received ($)	3,991	4,085	-94	3,984	3,863	120	4,105	4,053	52
	Amount of stipend received ($)	299	0	299 ***	241	0	241 ***	105	0	106 ***
	Total measured income a ($)	11,850	11,247	604 *	12,758	12,227	530	12,005	11,713	292
Second quarter of Year 2	Earnings ($)	1,148	1,120	28	1,302	1,310	-8	1,133	1,102	31
	Ever received TANF (%)	30.4	35.2	-4.8 *	39.0	40.1	-1.1	44.6	41.7	2.9
	Amount of TANF received ($)	139	154	-15	180	198	-18	215	201	14
	Ever received food stamps (%)	72.8	73.9	-1.1	69.6	68.4	1.2	71.4	71.2	0.2
	Amount of food stamps received ($)	631	655	-24	634	611	22	656	676	-20
	Amount of stipend received ($)	66	0	66 ***	54	1	54 ***	34	0	34 ***
	Total measured income a ($)	1,992	1,929	64	2,170	2,120	51	2,032	1,978	54
Sample size (total = 4,288)		654	652		578	586		905	913

SOURCES: MDRC calculations from UI, TANF, and food stamps administrative records from the State of Texas. NOTES: See Appendix B. This table includes only employment and earnings in jobs covered by the Texas unemployment insurance (UI) program. It does not include employment outside Texas or in jobs not covered by UI (for example, "off-the-books" jobs, some agricultural jobs, and federal goverment jobs). a This measure represents the sum of UI earnings, TANF, food stamps, and stipends.

ERA did not affect the number of people who received any TANF or food stamps in Corpus Christi. However, because only ERA group members were eligible for the stipend, total income from all four sources was $604 higher among ERA group members (a 5 percent increase over the control group level). Over the two-year follow-up period, the average amount received from the stipend was approximately $300 (not shown). During this time period, 27 percent of the ERA group received a stipend.22 This implies that those who received the stipend earned more than $1,100, on average, over the follow-up period ($300/0.27 = $1,111).

By Quarter 2 of Year 2 (see the lower panel of Table 4.2), ERA produced a statistically significant reduction in welfare receipt in Corpus Christi. Reducing recidivism was a key goal of the program. By that quarter, ERA group members were nearly 5 percentage points less likely to receive welfare. Since it was the first quarter in which an impact was evident, it is too early to say whether the ERA program in Corpus Christi will continue to reduce welfare receipt.23 ERA no longer produced a significant impact on total income in Corpus Christi by the second quarter of Year 2.

In Fort Worth, ERA had no effect on TANF, food stamp payments, or total income over the first six quarters. By Quarter 2 of Year 2, the ERA program in Fort Worth was still not producing an effect on these outcomes. The lack of an impact on income in Fort Worth is partly due to the lower take-up of the ERA stipend in that site. Approximately 19 percent of ERA group members in Fort Worth had received a stipend by Quarter 7.

In Houston, ERA generated a small increase in the amount of TANF received over the six-quarter follow-up period. This impact was the result of a series of small but statistically insignificant increases in TANF over several quarters. However, because of the lack of an impact on earnings and the relatively low amount received from the stipend in Houston, there was no impact on total measured income.24

Figure 4.1 shows the percentage who were employed and receiving the stipend and the percentage who were employed without receiving the stipend in each quarter of the follow-up period in Corpus Christi.25 The control group’s bars show the employment rates that would have been expected in the absence of ERA. To the extent that the shaded portion of the ERA group’s bar overlaps the control group’s bar, the stipend was paid to those who would have worked anyway. (In the study of incentive programs, this is known as “windfall.”) To the extent that the shaded region exceeds the control group’s employment bar, the stipends encouraged new employment. Although the ERA group’s bars are mostly a little higher than the control group’s, the figure shows that stipends do not appear to have added much additional employment. While Figure 4.1 shows that the stipend did not achieve its employment goals, the stipend did generate an increase in income in Corpus Christi (Table 4.2).

[D]

Impacts on Job Characteristics

Table 4.3 summarizes the impacts of the Texas ERA program on several measures of job characteristics, based on the ERA 12-Month Survey.26 It shows that, for the most part, ERA had no effect on the characteristics of participants’ jobs.

In addition to the cautions raised about the representativeness of the survey cohorts (see “Data Sources and Samples,” above), two new cautions emerged from reviewing the findings that are based on administrative records. First, the ERA 12-Month Survey covers only Year 1, which is a period when only 16 percent of the ERA group had received a stipend in Corpus Christi (for example). Second, Table 4.1 shows that both ERA and control group members who worked had higher average earnings in Year 2 than in Year 1. For these reasons –– and the fact that the survey is drawn from fairly narrow cohorts –– the Year 1 results may not be representative.

Most control group members were working in low-wage jobs. As suggested in the administrative records analysis, wages were especially low in Corpus Christi, where 60 percent of employed control group respondents worked at jobs that paid less than $7 per hour. In all three sites, relatively few worked in jobs with employer-provided benefits, such as sick days, dental benefits, and health insurance.

In Corpus Christi, few effects are large enough to be statistically significant. Among the few impacts that do reach significance, perhaps the most encouraging are the effects on working in jobs that require important skills. ERA group respondents in Corpus Christi were more likely than those in the control group to work in jobs that required computer skills and that required arithmetic.27

In Fort Worth, the survey sample size is below 100 per research group, so the findings are less reliable. Although an impact is found in the administrative records, no effect on job placement is evident in the survey. Not much can be made of this, however, because the cohorts do not overlap. In Fort Worth as in Corpus Christi, ERA increased the percentage of sample members who worked with computers.

The Employment Retention and Advancement Project

Table 4.3

Impacts on Characteristics of Current Job

Texas

Outcome			Corpus Christi			Fort Worth			Houston
			ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)
Employment status (%)	Ever employed since random assignment		76.3	70.8	5.5	72.1	74.6	-2.5	64.0	62.6	1.4
	No longer employed		23.6	21.9	1.7	28.8	35.9	-7.1	22.7	27.8	-5.1
	Currently employed		52.7	48.2	4.5	43.3	38.7	4.6	41.3	34.8	6.5
	Current working status	Full time	37.6	33.6	4.0	36.3	31.9	4.4	32.6	23.9	8.6
		Part time	15.1	14.6	0.4	7.0	6.8	0.2	8.7	10.8	-2.1
	Currently employed at a "good job a(%)		9.6	10.3	-0.7	15.5	9.1	6.3	15.5	9.4	6.1
Hours	Average hours per week		17.2	16.6	0.6	16.0	14.3	1.7	14.2	11.7	2.5
	Total hours per week (%)	Less than 30	15.1	14.6	0.4	7.0	6.8	0.2	8.7	10.8	-2.1
		30-34	12.1	6.7	5.4	3.1	7.5	-4.4	6.2	1.9	4.3 *
		35-44	20.9	19.8	1.1	27.5	17.4	10.2	22.4	19.3	3.0
		45 or more	4.6	7.1	-2.5	5.7	7.1	-1.4	4.0	2.7	1.3
	Average hourly wage (%)	less than $5.00	5.3	9.1	-3.8	6.0	8.9	-2.9	5.3	4.8	0.6
		$5.00 - $6.99	24.5	19.8	4.7	11.1	10.2	0.9	8.2	10.7	-2.5
		$7.00 - $8.99	14.8	11.5	3.4	13.4	11.1	2.2	17.8	10.4	7.4 *
		$9.00 or more	8.1	7.8	0.2	12.8	8.6	4.3	9.9	8.9	1.0
Earnings	Average weekly earnings ($)		121	113	8	132	100	32	104	98	6
	Total earnings per week (%)	Less than $200	22.0	20.2	1.8	12.0	10.4	1.6	13.3	11.6	1.6
		$201-$300	21.0	15.7	5.4	14.0	16.8	-2.9	15.0	9.2	5.8
		$301-$500	7.1	10.8	-3.7	12.0	10.4	1.7	13.3	10.2	3.1
		$500 or more	2.6	1.6	1.0	5.3	1.2	4.2	-0.3	3.7	-4.0 ***
Benefits (%)	Employer-provided benefits at current job	Sick days with full pay	12.7	12.1	0.6	12.6	7.7	4.8	12.6	14.4	-1.8
		Paid vacation	18.8	11.1	7.8 *	11.9	11.5	0.5	16.1	14.9	1.3
		Paid holidays other than Christmas and New Year	16.2	15.6	0.6	10.1	8.0	2.1	16.6	11.0	5.6
		Dental benefits	12.6	11.6	1.0	11.8	8.5	3.4	13.6	9.9	3.7
		A retirement plan	12.4	11.1	1.3	6.7	9.2	-2.5	10.9	8.6	2.3
		A health plan or medical insurance	15.4	13.6	1.8	14.0	9.5	4.6	16.4	11.9	4.5
Schedule b (%)	Regular		30.1	27.9	2.1	21.6	21.0	0.6	24.9	18.1	6.8
	Split		-0.1	1.4	-1.5	0.8	2.4	-1.6	0.6	0.1	0.4
	Irregular		2.0	3.5	-1.5	1.0	4.2	-3.2	0.7	2.0	-1.3
	Evening shift		5.9	7.2	-1.3	1.8	4.5	-2.7	2.8	6.0	-3.1
	Night shift		1.7	3.1	-1.4	5.6	0.9	4.7 *	2.0	2.7	-0.7
	Rotating shift		10.6	1.4	9.2 ***	7.5	0.1	7.3 **	7.7	4.4	3.4
	Other schedule		1.0	1.0	0.0	1.2	-0.1	1.2	2.2	-0.2	2.4 **
	Odd job		1.4	2.7	-1.3	3.9	5.7	-1.8	0.3	1.7	-1.4
Jobs skills index c			30.5	29.0	1.6	29.2	26.1	3.1 **	28.6	27.9	0.7
	Percentage reporting that job requires each at least monthly	Requires reading and writing skills	37.9	33.2	4.7	31.1	25.5	5.6	27.0	26.2	0.7
		Works with computers	25.0	16.6	8.4 *	19.2	6.6	12.7 ***	16.8	13.5	3.3
		Does arithmatic	30.6	21.4	9.3 *	23.1	14.4	8.7	22.9	22.9	0.1
		Requires customer contact	47.8	42.7	5.1	38.3	33.1	5.2	38.7	31.9	6.8
Sample size (total = 775)			141	149		92	96		150	147

SOURCE: MDRC calculations from ERA 12-Month Survey. NOTES: See Appendix D. a This definition of a good job is adapted from Johnson and Corcoran (2003). A "good job" is a job in which a respondent works 35 or more hours per week and either (1) pays $7.00 or more per hour and offers health insurance or (2) pays $8.50 or more per hour. b A split shift is defined as one consisting of two distinct periods each day. An irregular schedule is defined as one that changes from day to day. A rotating shift is one that changes regularly from days to evenings to nights. c The job skills index was created by regressing the "good job" measure on 10 dummy variables that indicate whether sample members possess specific job skills. This regression generated weights that ranked each skill based on its association with working at a good job. Each sample member was given a job skills score that was created by multiplying the regression-derived weights by each of the 10 jobs skills dummy variables. The result is an index that measures the probability of working at a good job, based on the skills that are required at the current job.

In Houston, the survey impacts are more positive than would be expected based on the findings from the larger sample for which administrative records are available. For example, respondents in the ERA group were nearly 9 percentage points more likely to work in full-time jobs, which is close to being statistically significant (p-value = 0.108). This is likely due to response bias. Specifically, the impacts of ERA on employment and earnings from the administrative records are more positive among Houston’s survey respondents than among its full report sample.28

Impacts on Employment Stability

Table 4.4 shows outcomes related to job retention, based on the ERA 12-Month Survey. ERA had no significant impacts on measures of job retention in any of the Texas sites.

The control group outcomes show that between 22 percent (in Corpus Christi) and 13 percent (in Houston) of control group respondents worked more than 10 months in the first year. Approximately half of those who were employed worked for 7 or fewer months. Only about 35 percent of control group members worked for the same employer for 6 months or more. Sample members were likely to work in multiple jobs in the one-year follow-up period. For example, in Corpus Christi, 29 percent of all control group respondents worked in two or more jobs; this constitutes nearly 46 percent of those who ever worked.

In Corpus Christi, ERA produced no statistically significant effects on job retention. While the findings from the administrative records indicate that there were some effects on job retention in Corpus Christi in Year 1, these effects are not observed in the survey sample. This may be because the survey sample size in Corpus Christi is rather small, making it difficult to detect statistically significant results. In Fort Worth and Houston, there were no effects on job retention (similar to the findings from the analysis of administrative records).

Impacts in Year 3

As discussed, a smaller group of sample members who were randomly assigned through June 2001 –– the “early cohort” –– have three years of follow-up data on measures of earnings and employment. Results for the early cohort may provide a preview of the impacts for the full sample in Year 3. The sample sizes that have three years of follow-up are 668 in Corpus Christi, 710 in Fort Worth, and 673 in Houston. Impacts for this early cohort and for the report sample are shown in Appendix Figure E.9. In Corpus Christi, this small group of early enrollees in the program experienced statistically significant employment impacts in Year 3. In the other two sites, impacts did not vary across cohorts. It should be noted that although the impacts look positive in Year 3 for this early cohort in Corpus Christi, there is no assurance that the impacts for the full report sample will be positive. Indeed, for the overlapping quarters, the impacts among the early cohort are somewhat larger and more consistent than the impacts for the full report sample.

The Employment Retention and Advancement Project

Table 4.4

Impacts on Employment Retention

Texas

Outcome		Corpus Christi			Fort Worth			Houston
		ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)	ERA Group	Control Group	Difference (Impact)
Ever employed in Year 1 (%)		70.5	63.5	7.0	63.0	70.9	-7.9	60.4	58.8	1.6
Average months employed in Year 1		5.4	4.7	0.7	4.0	5.0	-1.0	4.1	4.1	0.0
Total months employed in Year 1 (%)	Less than 4	14.1	14.1	0.0	21.0	16.3	4.7	15.3	11.0	4.3
	4 to 7	17.1	16.1	1.0	16.1	20.0	-3.8	19.2	22.0	-2.8
	8 to 10	13.8	11.8	2.1	10.7	14.8	-4.1	11.7	13.3	-1.6
	More than 10	25.4	21.6	3.9	15.1	19.9	-4.7	14.3	12.6	1.6
Worked during Months 1 to 3 and worked for (%)	Less than 6 consecutive months	9.0	11.6	-2.7	8.7	8.3	0.4	8.3	9.9	-1.6
	6 or more consecutive months	34.4	28.5	5.9	25.1	26.0	-0.9	21.5	22.3	-0.8
Number of jobs in Year 1 (%)	0	29.5	36.5	-7.0	37.0	29.1	8.0	39.6	41.2	-1.6
	1	33.9	34.4	-0.5	45.8	44.7	1.1	40.6	38.8	1.8
	2 or 3	30.8	25.2	5.6	15.4	23.8	-8.5	18.3	18.8	-0.5
	4 or more	5.8	3.9	1.9	1.8	2.4	-0.6	1.5	1.2	0.3
Ever worked for one employer for 6 months or more (%)		40.5	34.8	5.7	26.6	38.1	-11.5	29.1	30.2	-1.0
Sample size (total = 775)		141	149		92	96		150	147

SOURCE: MDRC calculations from responses to the ERA 12-Month Survey. NOTES: See Appendix D.

Impacts for Subgroups Based on Employment in the Prior Year

An analysis of stipend receipt rates, discussed in Chapter 2, found that certain subgroups of the ERA group were more likely to receive the ERA stipend. In particular, groups that had more recent employment history were more likely to receive the stipend, because they were more likely to work in the post-random assignment period. Thus, it might be expected that the impacts of ERA in Texas would be stronger among subgroups that had recent employment history. It is also possible that an intervention like ERA is more effective for those who are more easily employable.

In Corpus Christi and Fort Worth, the two-year impacts were concentrated among sample members who had recent employment history prior to entering the study. For those without recent employment, there were no impacts on employment or employment retention. In Corpus Christi, ERA increased the percentage of sample members who were employed four consecutive quarters by nearly 6 percentage points among those who had worked in the year prior to entering the study. ERA did not raise employment or earnings above control group levels for sample members with no recent employment.29 The impact on employment in Fort Worth was also concentrated among those with recent employment. In Houston, there were no impacts in either subgroup.

This pattern of subgroup impacts may be attributable to two factors. First, the sample members who were employed in the year prior to entering the study were reemployed sooner and, thus, more quickly reached the postemployment phase (where the difference between ERA and Choices was largest). Second, these sample members may have been more likely to have the necessary skills and human capital to better utilize the postemployment services and supports that the ERA program provided. These results are presented in Appendix Table E.16.

---

 

Table of Contents | Previous | Next