PUBLICATIONS
MDRC
List Publications By:


July 2001
Sustained Employment and Earnings Growth
New Experimental Evidence on Financial Work Incentives and Pre-Employment Services

Charles Michalopoulos

This paper appears in Low Wage Workers in the New Economy,which was edited by R. Cazis and M. S. Miller and published by the Urban Institute Press in 2001.

We know how to get low-income people to go to work:  build a strong and growing economy filled with jobs, make work pay through generous tax credits and welfare programs that allow working people to keep more of their benefits, and implement programs with employment and training services and time-limited welfare benefits to encourage people to work. However, we know little about the types of policies that will help people stay employed and increase their earnings over time. This paper seeks to partially fill the gap by pulling together recent evidence on how pre-employment services and financial work incentives can promote sustained employment and earnings growth.

The paper describes results from 13 programs begun since the early 1990s that share several important characteristics. First, each tested a policy designed to help or encourage single-parent welfare recipients to work. Second, each program now has enough information to assess whether the programs promoted sustained employment and promoted growth in hourly wages or quarterly earnings. Finally, each of the programs was studied by the Manpower Demonstration Research Corporation (MDRC) using a rigorous experimental research design that many people think gives the most reliable information about the effects of new policies. [1] In these studies, people were assigned at random to either a program group which was required to participate in an employment and training program or was offered a financial work incentive, or a control group which was not.

In other ways, the programs are quite diverse. They operated in a number of places: Atlanta, Georgia; Columbus, Ohio; Detroit, Michigan; Grand Rapids, Michigan; Portland, Oregon; Riverside, California; seven counties in Minnesota; and the Canadian provinces of British Columbia and New Brunswick. The programs vary in origin; most were part of state welfare-to-work programs funded under the Job Opportunities and Basic Skills Training (JOBS) program of the Family Support Act of 1988 (FSA), but one was a Canadian federal demonstration to test the effects of supplementing the earnings of long-term welfare recipients, and one was begun as a test of a change to the old AFDC program. Most important, the programs used different methods to help or encourage parents to find work. Two relied solely on financial work incentives that supplemented the earnings of people who went to work, ten used employment and training services such as job search assistance or adult basic education, and one combined financial work incentives with employment and training services.

The studies described in this paper provide useful information on policies that promote sustained employment, but they are not perfect. First, they included only welfare recipients and therefore cannot indicate how their strategies would affect a broader group of low-skill workers. Second, their primary objectives were to get people to go to work, and none of the programs used post-employment strategies to help people stay employed or advance. [2] Third, they did not collect information on the types of jobs people held, so they cannot directly tell us whether people who went to work because of these programs advanced in their jobs. In addition, only one of the studies has information on hourly wages over time, so that little can be said about these programs’ effects on growth in hourly wages. Finally, researchers studying the programs used a short-term measure of sustained employment, and it is impossible to know at this time whether the programs will help people to stay employed for long periods of time.

Despite these drawbacks, the studies can provide useful information on sustained employment and growth in earnings. Their key lessons include the following:

  • Programs with financial work incentives can promote sustained employment. Three programs supplemented the income of people who went to work. All three programs increased the number of people who worked. In addition, all three programs encouraged most people who went to work to stay employed for a year or longer. This makes sense. By providing families with extra income, the programs provided a reason to keep working and provided financial resources to weather temporary crises such as child care or transportation problems.

  • Programs that emphasize going to work immediately can promote sustained employment, but not all programs are equally effective. Four programs used pre-employment services such as job club to help people find jobs. Of these programs, two were more effective than the others at promoting sustained employment. The two less effective programs emphasized job search and work experience almost exclusively. The two more effective programs used a broader mix of job search and adult basic education. In addition, the most effective program operated in a strong economy, and its staff urged people to wait for “good” jobs that paid more than the minimum wage, were full-time, and offered opportunities for advancement.

  • Programs that emphasize building skills through adult basic education can promote sustained employment, but most of the programs studied had small effects. Six programs required people to enroll in adult basic education or vocational training to increase their employability. The programs generally had modest effects on employment overall and on sustained employment. There is some evidence, however, that requiring all people to enroll in basic or vocational education is as effective at promoting sustained employment as requiring all people to look for work initially. In two sites, programs were run side-by-side, one requiring most people to look for work initially, and one enrolling most people initially in adult basic education or vocational training. In these sites, the two approaches increased the number of people who went work and stayed employed by about the same amount.

  • Sustained full-time work may be the key to increasing hourly wages. One program supplemented the earnings of people who worked full-time (30 hours or more per week) but did not reward part-time work. People who went to work because of the incentive therefore worked full-time, and many of them were able to sustain their full-time employment. In this program, wages were more likely to increase for people who were offered the incentive than for people who were not offered the incentive.

  • Pre-employment services focused on getting people to work can result in earnings gains over time, but growth in earnings may be more closely linked to sustained employment. Programs that used pre-employment services to encourage immediate work increased the number of people whose earnings increased over time. Programs that had the largest effects on sustained employment, however, were also the most likely to result in earnings that increased over time. 

The story is complex and somewhat speculative.  There is not just one way of increasing retention, earnings, or wages. Financial work incentives appear consistently effective, but employment and training services are also effective in some settings.  Details on these points are presented in the sections below.

I.  Description of Programs

The programs studied in this paper include ten programs that were evaluated as part of the National Evaluation of Welfare-to-Work Strategies  (NEWWS); two versions of the Minnesota Family Investment Program (MFIP); and Canada’s Self-Sufficiency Project (SSP). [3] Brief descriptions of the programs are provided below. An appendix provides more details on data sources and the way that sustained employment and earnings growth were defined.

  • Labor Force Attachment (LFA) programs in Atlanta, Grand Rapids, and Riverside. [4] These programs required most participants to look immediately for work, usually through a job club that lasted from one to three weeks.  People who completed job search without finding a job were often then enrolled in adult basic education, vocational training, or work experience.

  • Education-focused programs in Atlanta, Grand Rapids, Riverside, Columbus and Detroit. [5] Six programs studied under NEWWS emphasized education: “human capital development” (HCD) programs in Atlanta, Grand Rapids, and Riverside; two programs in Columbus that tested different forms of case management; and one program in Detroit. In each of these programs, most participants were initially placed into education and training programs, particularly adult basic education and vocational training

  • Portland, Oregon JOBS Program. This was an employment-focused program that used job search for people who were considered ready to work but allowed people who were thought to need more skills to enroll initially in short-term adult basic education or vocational training before looking for work. As a result, only about one-third of participants in the program were required to look for work immediately, one-third were allowed to participate in education or training, and one-third were not assigned to any activity within three months of entering the program because they had already left welfare or begun working.  As in the three LFA programs, staff in Portland emphasized to clients that the goal of the program was to get a job. Unlike the LFA programs, however, the Portland program encouraged participants to wait until they found “good” jobs that paid more than the minimum wage, were full-time, and offered opportunities for advancement.

  • Minnesota Family Investment Program (MFIP). [6] Begun in 1994 to test whether financial incentives would encourage welfare recipients to work, MFIP allowed working welfare recipients to keep more of their welfare benefit than they could under AFDC.  For example, a mother of two who worked 20 hours per week and earned $6 per hour received almost $250 per month more in income under MFIP than under AFDC. In addition, MFIP required people who received welfare for 24 or more months over a three-year period to participate in employment and training services. MFIP’s services required most people to look for work and encouraged them to take jobs quickly, especially in comparison with the employment services available to the control group through AFDC. To understand the effects of MFIP’s incentives alone, some individuals were assigned to a program (called MFIP Incentive Only) that offered them the financial incentive but did not require them to participate in employment and training services.

  • Canada’s Self-Sufficiency Project (SSP). SSP offered a three-year earnings supplement to selected single-parent long-term welfare recipients in British Columbia and New Brunswick. The earnings supplement was a monthly cash payment available to single parents who had been on welfare for at least one year and who left welfare for full-time work (30 hours or more per week) within a year of entering the program. The supplement was paid on top of earnings for up to three continuous years, as long as the person continued to work full-time and remained off welfare. While collecting the supplement, an eligible single parent received an immediate payoff from work; in most cases, her total income before taxes was about twice her earnings.

            The period covered by these studies runs from the middle of 1991 until the end of 1998. People were first randomly assigned to program and control groups in Riverside in June 1991, and random assignment for the Riverside sample covered in this paper concluded in June 1993. Since results for Riverside include four years of information for each person, the period covered ends in the middle of 1997 for the last people randomly assigned. The most recent information comes from Portland, where random assignment took place between February 1993 and December 1994. Because of the four-year follow-up period in Portland, information for the last person randomly assigned in Portland comes from the end of 1998.

II. Evidence on Sustained Employment

Lesson 1:      Programs with financial work incentives can increase sustained employment.

            Two of the programs — SSP and MFIP Incentive Only — did nothing but try to make work pay by allowing welfare recipients to keep more of their welfare benefits when they went to work (MFIP) or by providing them with an earnings supplement if they went to work full-time (SSP). A third program — the full MFIP program — combined the program’s financial work incentives with mandatory employment-focused services. All three programs increased employment, and all three programs increased sustained employment.

            Evidence on the effects of financial incentives is presented in Table 1. As in all of the tables in this paper, results are based on experimental comparisons.  That is, average outcomes for the entire program group were compared with average outcomes for the entire control group. When the table refers to the proportion of people who did something, it consequently refers to the proportion of all people who were assigned to a program group or a control group; it does not refer merely to the people who ever worked.

The first three rows of Table 1 show the effects of SSP on full-time employment and on whether SSP increased sustained full-time employment. [7] The first row of the table indicates that 42.5 percent of the program group worked full time early enough in the follow-up period that we could determine whether they stayed employed for a year or longer.  During the same period, only 27.3 percent of the control group worked full time.  The difference in outcomes between the two research groups indicates the effect of the program.  In this case, it indicates that SSP increased the proportion of people who ever worked full time by 15.2 percentage points, an increase of 55.6 percent over what the control group did without the supplement offer.


Table 1: Effects of Programs with Financial Work Incentives on Sustained Employment

                       
       

Program

Control

Effect

Percentage

Employment outcome

Group (%)

Group (%)

(Difference)

Change (%)

SSP

                   

Ever worked full time

42.5

 

27.3

 

15.2

***

55.6

 
 

Left full-time work quickly

21.6

 

17.0

 

4.6

***

27.4

 
 

Stayed employed full time for a year or more

20.9

 

10.4

 

10.6

***

101.8

 

MFIP Incentives Only

               

Ever worked

44.4

 

39.2

 

5.2

 

13.3

 
 

Left work quickly

12.2

 

13.5

 

-1.3

 

-9.6

 
 

Stayed employed for a year or more

32.2

 

25.7

 

6.5

***

25.3

 

MFIP

               

Ever worked

50.5

 

39.2

 

11.4

***

29.1

 
 

Left work quickly

16.3

 

13.5

 

2.8

 

20.7

 
 

Stayed employed for a year or more

34.2

 

25.6

 

8.6

**

33.6

 

Source: Calculations from 18- and 36-month follow-up survey data in SSP and 36-month follow-up survey data in MFIP.

Notes:    Two-tailed t-tests were applied to differences between the outcomes for the program and control groups. Statistical significance level are indicated as: * = 10 percent; ** = 5 percent; *** = 1 percent.

         Impact may not appear to be the difference between the program group and control group outcomes due to rounding. 


People who ever worked full time can be divided into two groups:  those who stopped working full time quickly, and those who did not. The next two rows of the table report two composite outcomes: (1) the proportion of the program and control groups that found full-time jobs but stayed employed full time for less than a year, and (2) the proportion that found full-time jobs and stayed employed full time for a year or more

How should these measures be interpreted?  At one extreme, all people encouraged by SSP to work full time might have stopped working quickly (after less than a year).  In that case, the effect of the program on the proportion who worked full time for a year or more would be zero, and its effect on the proportion who worked full time for less than a year would be as large as its effect on full-time employment overall (15.2 percentage points). 

At the other extreme, all people encouraged by SSP to work full time might have done so for a year or more. In that case, the effect of the program on full-time employment that lasted a year or more would be the same as its effect on full-time employment (15.2 percentage points), and the effect on full-time employment that lasted less than a year would be zero.

The actual results show that neither of the extremes occurred.  Most, but not all, of the initial full-time employment generated in SSP did last at least a year. In particular, more than twice as many people in the program group as in the control group found full-time jobs and stayed employed full-time for a year or longer — 20.9 percent compared to 10.4 percent.  That is, SSP increased sustained full-time employment by 10.6 percentage points. 

The second panel of Table 1 shows similar results for people who were offered MFIP’s financial incentives but who were not required to participate in its employment services. [8] While 39.2 percent of the control group worked at some point, 44.4 percent of the program group did, for an increase of 5.2 percentage points. Like SSP, MFIP’s incentives primarily increased sustained employment: the incentives increased employment of a year or more by 6.5 percentage points, even more than it increased employment overall. 

The third panel of Table 1 shows similar results for people in MFIP who were not only offered the program’s financial incentives but also required to participate in its employment services. Combining incentives with mandatory employment services appears to be somewhat more effective than offering incentives alone.  The full MFIP program increased employment overall by more than 10 percentage points and increased sustained employment by nearly 10 percentage points.

The three programs that tried to make work pay produced consistent results.  All three increased employment, and all three increased sustained employment much more than they increased temporary employment. This makes sense.  Incentives in MFIP and SSP were available every month, giving people a reason to keep their jobs or to find new jobs when they lost work, and giving them financial resources to weather crises such as problems with child care or transportation.

Lesson 2:      Programs that emphasize going to work immediately can promote sustained employment, but not all programs are equally effective.

            Table 2 shows results on sustained employment for four JOBS programs that were focused on getting people to go to work — LFA programs in Atlanta, Grand Rapids, and Riverside, and the program in Portland. As described earlier, the LFA programs required almost all participants to initially enroll in job search, most commonly in job clubs lasting from one to three weeks. As a result, nearly 90 percent of participants in the Atlanta and Grand Rapids programs and about 70 percent of participants in the Riverside program initially looked for work. In Portland, in contrast, only people considered job-ready were required to look for work, while those considered most in need of basic skills were allowed to enroll in adult basic education or vocational training.

            According to the results in Table 2, the four employment-focused programs had quite varied results, both in terms of how much they encouraged people to find work, and how much they increased sustained employment. The most successful program at increasing employment overall was Riverside LFA, which increased employment by 10.6 percentage points. At the other extreme, Atlanta’s LFA program increased employment by only 3.5 percentage points.  In between were Grand Rapids LFA (5.5 percentage points) and Portland (7.0 percentage points). 

            The programs also differed substantially in whether they resulted in sustained employment.  In Portland and Atlanta LFA, the increase in sustained employment was as big as or bigger than the increase in employment overall.  For example, Portland increased employment overall by 7.0 percentage points and increased sustained employment by 6.7 percentage points.  Atlanta LFA increased sustained employment by 5.9 percentage points even though it increased employment overall by only 3.5 percentage points.  In contrast, Grand Rapids LFA and Riverside LFA increased primarily short-term employment.


Table 2:  Effects of Employment-Focused Welfare-to-Work Programs on Sustained Employment

                         
         

Program

Control

Difference

Percentage

Employment outcome

 

Group (%)

Group (%)

(Effect)

Change (%)

Atlanta LFA

                 

Ever worked

 

74.6

 

71.1

 

3.5

**

4.9

 
 

Left work quickly

 

36.8

 

39.3

 

-2.5

 

-6.3

 
 

Stayed employed for a year or more

 

37.8

 

31.9

 

5.9

***

18.6

 

Grand Rapids LFA

                 

Ever worked

 

85.1

 

79.6

 

5.5

***

6.9

 
 

Left work quickly

 

51.3

 

47.8

 

3.4

*

7.1

 
 

Stayed employed for a year or more

 

33.9

 

31.8

 

2.1

 

6.6

 

Riverside LFA

                 

Ever worked

 

66.6

 

55.9

 

10.6

***

19.0

 
 

Left work quickly

 

35.4

 

28.7

 

6.7

***

23.3

 
 

Stayed employed for a year or more

 

31.1

 

27.2

 

4.0

***

14.6

 

Portland

                 

Ever worked

 

80.3

 

73.4

 

7.0

***

9.5

 
 

Left work quickly

 

37.6

 

37.4

 

0.3

 

0.8

 
 

Stayed employed for a year or more

 

42.7

 

36.0

 

6.7

***

18.5

 

Source: Calculations from employment reported to state unemployment insurance systems.

Notes:    Two-tailed t-tests were applied to differences between the outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; *** = 1 percent.

         Impact may not appear to be the difference between the program group and control group outcomes due to rounding. 


           It is not clear why the programs had such different effects on sustained employment. Three possible explanations are the mix of education and job search used in the different programs, attitudes of staff about the effectiveness of the LFA strategy, and economic conditions.

            Atlanta LFA and Portland used a broader mix of job search and adult basic education than did Riverside LFA and Grand Rapids LFA. As described earlier, this mix was an explicit part of Portland’s program. Although it required nearly everyone to initially look for work, Atlanta LFA made substantial use of adult basic education for people who looked for work without finding a job.  In contrast, Riverside LFA required many people who failed to find work to continue looking for work, and Grand Rapids LFA placed much of this group into unpaid work experience. This does not imply that relying primarily on adult basic education is a key to sustained employment, but rather that the mix of job search and adult basic education may be important, perhaps because programs that use both job search and education can target each strategy at people who would benefit from them.  Lesson 3 will discuss results for several programs that required most people to enroll in education programs.

            Atlanta LFA and Portland also differed from Riverside LFA in the services that case managers thought would best help welfare recipients move to work. In both Portland and the Atlanta LFA program, twice as many staff preferred human capital development to quick job entry as a strategy for moving clients to work. In Riverside’s LFA program, in contrast, nearly all case managers preferred labor force attachment to human capital development as a means of moving people to work.

            A third possible explanation for the larger increases in sustained employment in Atlanta and Portland is the state of the economy. Between 1993 and 1998, employment grew in all four sites, but the unemployment rate was by far the highest in Riverside, particularly in the early part of the program. [9] When Riverside’s program moved people to work quickly, it was in a poor economy that might have provided short-term, temporary, or undesirable jobs. The economy was strong in Grand Rapids during this period, however, suggesting that a strong economy is not enough to ensure the employment services will generate increases in sustained employment. 

            The fact that Riverside’s program increased short-term employment more than it increased sustained employment should not necessarily be viewed as negative. The primary goal of the program was to help people go to work, and Riverside’s program was the most effective of the four employment-focused programs in accomplishing this goal. It does suggest, however, that a program that tries to get people to go to work as quickly as possible may need other features such as post-employment services to help people stay at work and advance in their careers (although the few experimental evaluations of post-employment strategies have not been encouraging).

Lesson 3:      Programs that encourage most people to build skills through adult basic education can also increase retention, but most of these programs had small effects.

             The third major approach to encouraging welfare recipients to work (in addition to financial incentives and job search) is by increasing their skills and thereby increasing their attractiveness to employers and their ability to earn a living wage. Table 3 presents results on sustained employment for six programs that required most people to enroll initially in adult basic education or vocational training. Overall, these programs did not promote much sustained employment. Nevertheless, focusing on basic education appears to be as effective as focusing on job search in encouraging sustained employment.


Table 3:  Effects of Education-Focused Welfare-to-Work Programs on Sustained Employment

                         
         

Program

Control

Difference

Percentage

Employment outcome

 

Group (%)

Group (%)

(Effect)

Change (%)

Atlanta HCD

                 

Ever worked

 

74.2

 

71.1

 

3.0

**

4.3

 
 

Left work quickly

 

36.8

 

39.3

 

-2.5

 

-6.3

 
 

Stayed employed for a year or more

 

37.4

 

31.9

 

5.5

***

17.3

 

Grand Rapids HCD

                 

Ever worked

 

82.9

 

79.6

 

3.3

**

4.1

 
 

Left work quickly

 

49.0

 

47.8

 

1.2

 

2.4

 
 

Stayed employed for a year or more

 

33.9

 

31.8

 

2.1

 

6.6

 

Riverside HCD

                 

Ever worked

 

55.7

 

49.2

 

6.5

***

13.2

 
 

Left work quickly

 

31.7

 

27.6

 

4.2

**

15.1

 
 

Stayed employed for a year or more

 

24.0

 

21.7

 

2.3

 

10.7

 

Columbus Integrated

                 

Ever worked

 

81.8

 

78.2

 

3.6

***

4.6

 
 

Left work quickly

 

38.1

 

38.3

 

-0.1

 

-0.3

 
 

Stayed employed for a year or more

 

43.6

 

39.9

 

3.7

**

9.3

 

Columbus Traditional

                 

Ever worked

 

80.8

 

78.2

 

2.7

**

3.4

 

 

Left work quickly

 

38.7

 

38.3

 

0.4

 

1.1

 
 

Stayed employed for a year or more

 

42.2

 

39.9

 

2.2

 

5.6

 

Detroit

                 

Ever worked

 

75.8

 

72.4

 

3.4

**

4.7

 
 

Left work quickly

 

46.6

 

45.3

 

1.3

 

2.9

 
 

Stayed employed for a year or more

 

29.1

 

27.1

 

2.1

 

7.6

 
                         

Source: Calculations from employment reported to state unemployment insurance systems.

Notes:    Two-tailed t-tests were applied to differences between the outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; *** = 1 percent.

         Impact may not appear to be the difference between the program group and control group outcomes due to rounding.


The focus of these programs was generally short-term adult basic education and vocational training. In the three HCD programs, for example, between 36 percent and 57 percent of participants were initially required to enroll in adult basic education.  In Atlanta and Grand Rapids, another  20 to 30 percent of participants were enrolled initially in vocational training. (Only people in need of basic education were assigned to the Riverside HCD program. As a result, few participated in vocational training and a majority were assigned to adult basic education.) The average participant was engaged in program activities for about 8 months in the 24 months after random assignment. [10] However, nearly 30 percent of participants were enrolled for more than 12 months during this two-year period.

In general, these education-focused programs had fairly small effects on employment, ranging from 2.7 percentage points in the Columbus Traditional program to 6.5 percentage points in the Riverside HCD program.  They also had fairly small effects on sustained employment. Only the Atlanta HCD program and the Columbus Integrated program significantly increased the number of people who went to work and stayed there for a year or longer.

It would be a mistake to conclude that adult basic education is less effective than job search at promoting sustained employment, because most of the education-focused programs operated in different sites with different people from the employment-focused programs shown in Table 2. Fortunately, results from Atlanta and Grand Rapids present a rare opportunity to compare the job search and adult basic education approaches. In both sites, people were assigned at random to the LFA and HCD programs. [11]  As a result, any differences in the effects of the two programs can reliably be attributed to differences in the programs.  While there were a number of differences in implementing the two types of programs, the primary difference was in the self-sufficiency approach: in the LFA programs, most people were asked to look for work initially; in the HCD programs, most people were asked to enroll in adult basic education or vocational training.

When results for the Atlanta and Grand Rapids LFA programs in Table 2 are compared with results for the HCD programs in Table 3, the LFA and HCD approaches appear equally effective (or ineffective) at encouraging sustained employment.  The Atlanta LFA program increased sustained employment by 5.9 percentage points, but the Atlanta HCD program increased sustained employment by 5.5 percentage points.  Likewise, in Grand Rapids neither the LFA program nor the HCD program significantly increased sustained employment.  Thus, something about the site — for example, the state of the economy, the people enrolled in the programs, or other local policies — appears to be more responsible than the self-sufficiency approach for increases in sustained employment.

III. Evidence on Growth in Hourly Wages and Quarterly Earnings

It has not been clear from prior evidence whether low-skilled workers earn higher hourly wages over time. In looking at five years of information for a group of women who had left welfare, Maria Cancian and colleagues at the University of Wisconsin-Madison (1999) found that the group’s average hourly wages gradually increased over time.  In the year after they left welfare, the median worker in their sample earned $6.36 per hour. Four years later, the median person earned $6.73 per hour.  Although hourly wages increased on average, the increase was slow; the rate of increase was only about one percent per year.

 Cancian and her colleagues found more promising news about annual earnings. In the year after they left welfare, the average person in their sample earned $7,668.  By the fifth year after leaving welfare, they earned $10,942 on average, an increase of more than 40 percent over four years. Even among people who earned the least, earnings increased by about 60 percent over the four years, although this group continued to earn very little. Since hourly wages increased very slowly over time, the large increases in earnings mean that people were working more hours or working more often over time.

Two recent studies provide more hope that welfare recipients can increase their hourly wages.  The key may be steady, full-time work. In a recent paper, Tricia Gladden and Christopher Taber (1999) suggest that wages increase as much for low-skilled workers as for high-skilled workers when they increase their work experience by similar amounts. Since high-skills workers typically work full-time all the time, this suggests that low-skills workers would also see substantial increases in their wages if they could work full time all the time.  A second recent paper by Mary Corcoran and Susanna Loeb (1999) found that former welfare recipients who worked full time had substantially higher wage growth than those who worked part time.

Although the results of these papers are interesting, they are not based on rigorous, random-assignment evaluations. The remainder of this section presents evidence on how SSP affected growth in hourly wages and how employment-focused programs from the NEWWS evaluation affected growth in quarterly earnings. [12] Although these measures do not directly indicate whether people were advancing in their careers, growing wages and earnings may be indicators of advancement.

Lesson 4:      Full-time work may be the key to increasing hourly wages over time.

Of the studies summarized in this paper, only SSP collected information on hourly wages over time. As is described earlier, SSP encouraged a substantial number of people to work full-time and stay there. Table 4 indicates that SSP also resulted in growing hourly wages for a substantial number of people. [13]

The first row of results in Table 4 indicates that hourly wages decreased for 5.7 percent of the control group and 6.7 percent of the program group between the end of the first year after random assignment until the end of the third year after random assignment.  In other words, few people had wages that decreased over time, and there was little difference in this outcome between the two research groups even though SSP had a substantial overall effect on employment. 

This suggests that SSP’s main effect was to encourage people to take jobs in which their wages increased.  This is confirmed in the next row of the table.  While 13.4 percent of the control group took jobs in which their wages increased, 21.1 percent of the program group did.  Thus, SSP increased the number of people in jobs with growing wages by 7.6 percentage points, or more than 50 percent over the proportion in the control group.  (Note, however, that wage growth could not be calculated for the majority of people in SSP, either because they never worked or because they worked too sporadically to provide a meaningful measure of wage growth.)


Table 4: Effects of SSP on Wage Growth Between End of Year 1 and End of Year 3 After Random Assignment

                         
         

Program

Control

Difference

Percentage

Employment outcome

 

Group (%)

Group (%)

(Effect)

Change (%)

                       
 

Employed and hourly wage decreased

 

6.7

 

5.7

 

1.1

 

19.3

 
 

Employed and hourly wage increased

 

21.0

 

13.4

 

7.6

***

56.5

 

Source: Calculations from 18- and 36-month follow-up survey data in SSP.

Notes:    Two-tailed t-tests were applied to differences between the outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; *** = 1 percent.

         Impacts may not appear to be the differences between the program group and control group outcomes due to rounding.

          Results show whether wages increased or decreased from the end of the first year after random assignment to the end of the third year after random assignment.


Results from SSP are more encouraging than results described by Cancian et al (1999) but are consistent with the work of Gladden and Taber (1999) and Corcoran and Loeb (1999) that suggested that working full time and working regularly were keys to growing wages. This makes sense. Someone who works full time is spending more time gaining valuable skills through on-the-job training than someone who works part time, perhaps because employers are more willing to invest in full-time workers. Likewise, someone who is working in most months will be able to gain on-the-job skills faster than someone who is working sporadically, while someone who is in and out of work may lose skills while looking for new employment.

Lesson 5:      Pre-employment services focused on getting people to work quickly can result in earnings gains over time, but growth in earnings may be linked to sustained employment.

Table 5 shows whether the four employment-focused programs (Portland and the LFA programs in Atlanta, Grand Rapids, and Riverside) resulted in quarterly earnings that increased over time. [14] Each of the four programs caused more people to experience earnings growth. However, variation across the programs in their ability to increase people’s earnings was similar to variation in their ability to promote sustained employment. This may suggest that the two outcomes are linked.

The first two rows of the table show the effects for the Atlanta LFA program on whether or not earnings increased over time. The first row of the table indicates that 32.9 percent of the program group worked at some point but did not have their earnings increase over time.  This group is actually composed of three smaller groups.  Some people went to work and had their earnings decrease over time; some people went to work and had constant earnings over time; and some people worked so sporadically that it was impossible to reliably determine whether their earnings increased or decreased over time.  During the same period, only 31.7 percent of the control group worked but did not have their earnings increase.  The difference in outcomes between the two research groups indicates the effect of the program.  In this case, it indicates that the Atlanta LFA program had little effect on the proportion of people who went to work and then did not have their earnings increase.


Table 5: Effects of Programs with Employment Services on Earnings Growth Over Four Years

                         
         

Program

Control

Difference

Percentage

Employment outcome

 

Group (%)

Group (%)

(Effect)

Change (%)

Atlanta LFA

                 
 

Employed and earnings did not increase

 

32.9

 

31.6

 

1.2

**

3.8

 
 

Employed and earnings increased

 

32.0

 

28.3

 

3.7

**

13.1

 

Grand Rapids LFA

                 
 

Employed and earnings did not increase

 

39.3

 

35.2

 

4.2

**

11.9

 
 

Employed and earnings increased

 

38.0

 

34.2

 

3.8

**

11.1

 

Riverside LFA

                 
 

Employed and earnings did not increase

 

35.0

 

25.1

 

9.9

***

39.4

 
 

Employed and earnings increased

 

24.4

 

20.0

 

4.4

***

22.0

 

Portland

                 
 

Employed and earnings did not increase

 

33.5

 

30.1

 

3.3

***

11.0

 
 

Employed and earnings increased

 

36.9

 

30.1

 

6.8

***

22.6

 

Source: Calculations from state unemployment insurance systems.
Notes:   Two-tailed t-tests were applied to differences between the outcomes for the program and control groups. Statistical significance levels are indicated as: * = 10 percent; ** = 5 percent; *** = 1 percent.

Impacts may not appear to be the differences between the program group and control group outcomes due to rounding.

Earnings growth was measured from the first to last employment spell, or from the beginning to end of employment if a person had only one employment spell. Because earnings in the first or last quarter of a spell are more likely to reflect partial quarters of employment, earnings were included in the analysis only for spells that lasted at least three quarters, and earnings from the first and last quarter of a spell were not included in the analysis.


The second row of Table 5 indicates that the Atlanta LFA program also increased the number of people with earnings that grew over time. While 32.0 percent of the program group worked and had their earnings increase, 28.3 percent of the control group worked and had their earnings increase.  Thus, Atlanta’s LFA program increased the proportion of people who went to work and then had their earnings increase by 3.7 percentage points.

Each of the other three programs also increased the number of people with growing earnings. Like Atlanta LFA, the LFA programs in Grand Rapids and Riverside increased the proportion of the sample who went to work and had increased earnings by about 4 percentage points.  The Portland program had an even larger effect on the number of people with increasing earnings.

In some ways, results on earnings growth are similar to results on sustained employment. The LFA programs in both Grand Rapids and Riverside increased sustained employment less than they increased short-term employment and increased the number of people with increasing earnings less than the number of people with stagnant or falling earnings. The programs in Portland and Atlanta, on the other hand, both primarily promoted sustained employment and primarily increased the number of people with earnings that increased over time. These results may provide further evidence that sustained employment is important in helping people increase their earnings over time. Alternatively, some factors in Portland and Atlanta — such as a use of both adult basic education and job search, an education-friendly staff, and strong economies — may have simultaneously promoted sustained employment and growing earnings.

IV. Conclusions and Implications

             It is risky to draw conclusions from so few programs.  Nevertheless, some patterns across the programs suggest some speculative recommendations about how pre-employment programs might promote sustained employment and wage or earnings growth.

  • Offer and market financial work incentives that increase the income of working families. Most of us work for pay; few of us volunteer.  It makes sense, then, that making work pay for low-income families through earnings supplements, welfare earnings disregards, or earned income tax credits would encourage people to stay employed. In addition to providing incentives, they provide extra income that might help low-income families weather short-term crises, such as car problems, that might otherwise keep them from working. Are financial work incentives a credible option today? Not only are they credible, but they are available to all working poor families through the federal Earned Income Credit. A number of states have similar credits. In addition, more than 40 states today allow working welfare recipients to keep more of their welfare benefits than they did prior to welfare reform. Even if new financial incentives cannot be offered, it is important to make sure that low-wage workers know about and are using the ones that they can receive. 
  • Encourage full-time work. There is growing evidence that full-time work may be a key to both employment retention and advancement. There may be a number of reasons for this. Employers are less likely to hire a full-time worker to fill temporary needs. Full-time work is more likely than part-time work to come with fringe benefits such as health insurance, and those benefits provide an additional incentive to continue working. Full-time, steady work provides people with more work experience that can be marketed to future employers. Of course, full-time work also has potential drawbacks. Some parents with low skills might not be able to find full-time work, thus limiting the number of families who benefit from programs that help only full-time workers. Full-time work will increase the need for child care, possibly cutting into take-home pay and possibly affecting children’s psychological and social development. In the New Hope project (Bos et al., 1999), a program that offered an array of earnings supplements, medical insurance, and child care subsidies to low-income parents who worked 30 hours or more per week, there was no evidence that children were hurt when their parents went to work, and there was evidence that some children benefited. Thus, programs that encourage full-time work may also need to invest more in child care, either by increasing child care subsidies or developing after-school programs.
  • Job search alone is not enough. Requiring people to look for work and helping them look for work is effective in getting people to work. Keeping them at work requires more, however. Of the job search programs described in this paper, one combined job search with financial incentives, two allowed many people to build basic skills through adult basic education or vocational training, and one encouraged people to wait for jobs that have fringe benefits or pay more than the minimum wage. Since job search can be so effective in getting people to work, combining job search with post-employment services or education may also be a good way to encourage retention and advancement. It is still not clear what types of services would work, however.
  • Basic education alone is not enough. Requiring all people to enroll in adult basic education or vocational training is no more effective at promoting sustained employment than requiring all people to look for work. Again, a broad mix of the two approaches may be most effective. This is not to say that education more generally will not promote retention and advancement. None of the programs discussed in this paper used long-term education or community college, for example, as ways to build skills that might help people get better jobs. These strategies remain possibilities that should be examined in the future.
  • Something more is needed. Although some of the strategies discussed in this paper appear to promote sustained employment and wage growth, many people never work, and many people who do work are not able to sustain their employment. In other words, the strategies have been effective to a point, but much more is needed. Many states and localities have begun to address problems such as substance abuse, domestic violence, and depression. In many places, post-employment services are also being tried to give people advice on how to get promoted, help them identify educational opportunities that might help them advance, and so on. At this time, we have no credible information on how much these approaches will help. Fortunately, the U.S. Department of Health and Human Services recently launched the Employment Retention and Advancement initiative to use random assignment evaluations to test the effectiveness of a variety of approaches. In several years, we may have much better information on what works to promote retention and advancement.

 Privacy PolicySite Map | ©2009 MDRC®

Appendix

A.  Sites and Samples

The sites and programs varied substantially in the generosity of their welfare systems and the robustness of the local economy, but people involved in the programs had similar histories and demographic characteristics. These factors may affect whether programs result in sustained employment or growing wages. For example, in a strong economy it is easier to find high-quality jobs that are more likely to be stable and to provide opportunities for advancement. It is likewise easier to find a job that pays better than welfare in low-grant states than in high-grant states, so people in low-grant states who go to work might be more likely to take low-wage or unstable jobs. In each case, however, it would be easier for both the control group and program group to find stable employment.  It is therefore not clear whether these factors would increase the effects of the program on stable employment or decrease their effects.

Reflecting the strong state of the national economy over the last decade, most sites had moderate unemployment rates of 4 to 6 percent when their programs began.  A notable exception was Riverside, which had an unemployment rate of 11.5 percent in 1993.  Unemployment was also quite high in the two Canadian provinces, at about 9 percent in the Vancouver area and more than 10 percent in New Brunswick during most of the period covered in this paper.

The seven sites fall into three rough categories according to the generosity of welfare benefits in their states and provinces.  The most generous welfare benefits are in California and the Canadian provinces. In 1993, a single parent with two children in California was eligible for cash assistance of $624 per month from AFDC. In British Columbia, the family would have been eligible for a benefit of more than $800 from 1992 through 1998, while a similar family in New Brunswick was eligible for basic assistance of nearly $600 per month. [15] Three states – Michigan, Oregon, and Minnesota – provided moderately generous benefits between $450 and $550 per month around the same time period. The final two states — Georgia and Ohio provided less than $350 per month in welfare benefits for a family of three in 1993.

            In many ways, the samples were quite similar across the programs. Because the studies focused on single-parent families, the vast majority of sample members were female (more than 90 percent in each site).  About half of the people in each site had neither graduated from high school nor earned a GED by the time of random assignment.  The average age of sample members in each site was about 30, and the average family in each site had about two children.  Even in these samples of welfare recipients and welfare applicants, the vast majority (more than 90 percent in some sites) had worked at some time in the past. In all programs studied in this paper, most or all people were receiving welfare when they were randomly assigned. In MFIP, all people had received welfare for 24 of the 36 months prior to random assignment. In SSP, all were receiving welfare when they were randomly assigned and all had received welfare in 11 of the 12 prior months.  In Atlanta, Grand Rapids, Portland, and Riverside, a majority of sample members had received welfare for at least two years prior to random assignment.

B.  How Sustained Employment and Earnings Growth Were Measured

            Results described in this paper come from three recent or forthcoming reports:  Michalopoulos et al. (2000) on SSP, Miller et al. (2000) on MFIP, and Freedman (Forthcoming) on programs in Atlanta, Columbus, Detroit, Grand Rapids, Portland, and Riverside. No new analyses have been done for this paper; therefore, decisions made in the reports limit what can be presented in this paper.

            For programs in Atlanta, Columbus, Detroit, Grand Rapids, Portland, and Riverside, follow-up information on employment and earnings was taken from reports made by employers to each state's unemployment insurance (UI) system. For MFIP, information was taken from a survey administered about three years after random assignment, while for SSP, information was taken from surveys conducted about 18 months and 36 months after random assignment.

            In examining the effects of the programs on sustained employment, the three reports made different decisions about how to define who was employed but made similar decisions about how to define sustained employment.  In SSP, sustained employment was measured for people who worked full-time within the first 18 months after entering the evaluation of the program. SSP is alone in limiting its analysis of sustained employment to full-time employment, although that decision makes sense in light of the fact that SSP’s incentives could be received only by people who worked full-time.  In MFIP, sustained employment was measured for people who worked at all during the first year after entering the evaluation.  The shorter period was chosen by the MFIP team because it represented the time of the program’s largest effects on employment. In the programs in Atlanta, Columbus, Detroit, Grand Rapids, Portland, and Riverside, sustained employment was measured for people who worked within three years of entering the program’s evaluation. The longer time makes sense for these programs. Education-focused programs in most of these sites were not expected to increase employment immediately, since people were expected to undertake education to benefit from the programs.

            The three reports used similar definitions of sustained employment. In each study, a person was considered to have sustained employment if she worked for an entire year after first finding work.  Since results for MFIP and SSP are based on monthly employment, a person retained her employment if she worked in twelve consecutive months.  In contrast, results for Atlanta, Columbus, Detroit, Grand Rapids, Portland, and Riverside are based on quarterly UI-reported earnings.  A person was consequently considered to have retained her employment if she worked in four consecutive quarters. Some people work in a quarter but do not work in each month in the quarter; therefore, it is somewhat easier for a person to appear to have sustained employment in Atlanta, Columbus, Detroit, Grand Rapids, Portland, and Riverside than in MFIP or SSP.

            The method used to define wage or earnings growth differed across the reports.  In SSP, individuals reported information about their hourly wage, and wage growth was calculated only for people who were working at the end of the first year after entering the evaluation and at the end of the third year after entering the evaluation.  The sample was limited in this way to ensure that wage growth was calculated over a fairly long period of time and to ensure that it was calculated only for people who were most likely to have worked most of the time.  In Atlanta, Grand Rapids, Portland, and Riverside, information on wages was not available at different points in time, so the analysis of earnings growth relied on quarterly earnings reported to the UI system.  Thus, earnings growth in these sites refers to whether quarterly earnings increased or decreased between the first and last spells of employment, or the beginning and end of employment if there was only one spell, regardless of whether that change stemmed from a change in the number of hours worked in a quarter or a change in the hourly wage.  Wage and earnings growth were not analyzed by the MFIP researchers.


References

Bos, Hans, Aletha Huston, Robert Granger, Greg Duncan, Tom Brock, Vonnie McLoyd. 1999. New Hope for People with Low Incomes: Two-Year Results of a Program to Reduce Poverty and Reform Welfare. New York: Manpower Demonstration Research Corporation.

Cancian, Maria, Robert Haveman, Thomas Kaplan, Daniel Meyer, and Barbara Wolfe. 1999. “Work, Earnings, and Well-Being after Welfare:  What Do We Know?” Focus. 20(2):22-25.

Corcoran, Mary, and Susanna Loeb. 1999. “Will Wages Grow With Experience for Welfare Mothers? Focus. 20(2):20-21.

Gladden, Tricia, and Christopher Taber. 1999. “Wage Progression Among Less-Skilled Workers.” Chicago: Joint Center for Poverty Research Working Paper 72.

Freedman, Stephen. Forthcoming. National Evaluation of Welfare to Work Strategies: Four-Year Effects of Ten Programs on Sustained Employment and Earnings Growth. Washington: U.S. Department of Health and Human Services, Office of the Assistant Secretary for Planning and Evaluation and Administration for Children and Families, and U.S. Department of Education, prepared by Manpower Demonstration Research Corporation.

Miller, Cynthia, Virginia Knox, Lisa A. Gennetian, Martey Dodoo, Jo Anna Hunter, and Cindy Redcross. 2000. MFIP Adult Report: The Minnesota Family Investment Program. New York: Manpower Demonstration Research Corporation.

Michalopoulos, Charles, David Card, Lisa A. Gennetian, Kristen Harknett, and Philip K. Robins. 2000. The Self-Sufficiency Project at 36 Months: Effects of a Financial Work Incentive on Employment and Income. Ottawa: Social Research and Demonstration Corporation.

Rangarajan, Anu, and Tim Novak. 1999. The Struggle to Sustain Employment: The Effectiveness of the Postemployment Services Demonstration. Princeton: Mathematica Policy Research.

Rangarajan, Anu, Peter Schochet, and Dexter Chu. 1998. Employment Experiences of Welfare Recipients Who Find Jobs: Is Targeting Possible? Princeton: Mathematica Policy Research.

Quets, Gail, Philip K. Robins, Elsie C. Pan, Charles Michalopoulos, and David Card. 1999. Does SSP Plus Increase Employment? The Effect of Adding Services to the Self-Sufficiency Project’s Financial Incentives. Ottawa: Social Research and Demonstration Corporation.

Strawn, Julie, and Karin Martinson. 2000. Steady Work and Better Jobs: How to Help Low-Income Parents Sustain Employment and Advance in the Workforce. New York: Manpower Demonstration Research Corporation.


Notes:

[1] Other studies have done research on sustained employment and earnings growth but have not used the rigorous random assignment method. For a summary of this research, see the recent MDRC how-to guide on sustained employment (Strawn and Martinson, 2000).

[2] To learn more about experimental evaluations of post-employment services, see the reports on the Postemployment Services Demonstration (Rangarajan and Novak, 1999) and the Self-Sufficiency Project Plus (Quets et al., 1999), two projects that used post-employment services to try to increase employment among ex-welfare recipients but that failed to do so. Results on sustained employment and wage or earnings growth were not available for these two evaluations, so they are not discussed in this paper.

[3] Results described in this paper come from three recent or forthcoming reports:  Michalopoulos et al., 2000, on the Canadian program; Miller et al., 2000, on the programs in Minnesota; and Freedman, Forthcoming, on programs in Atlanta, Columbus, Detroit, Grand Rapids, Riverside, and Portland. See those reports for more information about the programs or results described in this paper.

[4] The Riverside programs discussed in this paper are not the Riverside GAIN program studied by MDRC beginning in 1988.  That program had some of the largest effects on employment seen in a random assignment evaluation of a welfare-to-work program. Like the program in Portland, Riverside GAIN emphasized employment but allowed people in need of basic education to enroll in adult basic education before looking for work.  Results on sustained employment were not calculated for the Riverside GAIN program and are therefore not presented here. The program and its results are described in Riccio, Friedlander, and Freedman, 1994.

[5] Oklahoma City also ran an education-focused program that was studied in NEWWS.  However, too little follow-up information was collected to allow researchers studying the program to show its effects on sustained employment or earnings growth.

[6] MFIP is also the name of Minnesota’s TANF program which is a substantially modified version of the pilot MFIP program described here. In comparison to the pilot version of MFIP, Minnesota’s TANF program has time limits on receipt of welfare, less generous financial work incentives, more stringent requirements for participating in employment and training services, and a greater emphasis on job search.

[7] As described in the appendix, SSP researchers looked only at sustained full-time employment because full-time employment was the goal of the program. Because some people in SSP were employed part-time but never worked full-time, employment rates reported for SSP in this paper are somewhat lower than in the other studies.

[8] Results for MFIP are limited to long-term welfare recipients in urban counties since this is the only group for which sustained employment was calculated by MFIP researchers.

[9] In 1993, for example, the unemployment rate in Riverside was 11.5 percent; in Portland, it was only 5.5 percent. While many jobs were added in Riverside during this period, the unemployment rate remained quite high at 6.9 percent in 1998.

[10] According to Hamilton et al. (1997), average months of participation among people who ever participated in the HCD programs was 7.5 months in Riverside, 8.3 months in Grand Rapids, and 9.4 months in Atlanta. Results on duration of participation are not yet publicly available for the programs in Columbus and Detroit.

[11] Riverside also ran HCD and LFA programs. Only people in need of basic education were assigned to the HCD program, however.  Because results are not available for members of the LFA group in need of basic education, results for the LFA program cannot be compared with results for the HCD program. 

[12] Evidence on earnings growth is also available for the education-focused programs studied in NEWWS. Only one of the six programs significantly increased the number of people with growing earnings, however, so results are not shown in this paper. Results on wage growth and earnings growth are not available for MFIP.

[13] Wage growth can be meaningfully measured only for people working at the beginning and end of a fairly long period of time. In SSP, wage growth was therefore calculated only for people who were working at the end of the first year and at the end of the third year after entering the evaluation.

[14] Hourly wages were not available over time in these studies. Results therefore indicate growth in quarterly earnings rather than growth in hourly wages.

[15] Dollar amounts have been converted to U.S. dollars at a rate of 75 cents U,S. per Canadian dollar, which is approximately the exchange rate at the beginning of the SSP evaluation.




 Privacy PolicySite Map | ©2009 MDRC®