| This report presents the latest findings on the effectiveness
of California's Greater Avenues for Independence (GAIN) Program,
a statewide initiative aimed at increasing the employment
and self-sufficiency of recipients of Aid to Families with
Dependent Children (AFDC), the nation's major cash welfare
program. Based on three years or more of follow-up data for
33,000 people who entered GAIN between early 1988 and mid-1990,
the study examines the program's effects in six counties on
employment, earnings, welfare receipt, and other outcomes,
as well as findings from a comprehensive benefit-cost analysis.
The results are of broad relevance to welfare reform because
California has the country's biggest AFDC caseload and GAIN
is the largest and one of the most ambitious programs operating
under the federal Job Opportunities and Basic Skills Training
(JOBS) Program, created by the Family Support Act of 1988.
Operating as California's JOBS program since July 1989, GAIN
currently accounts for almost 13 percent of federal spending
on JOBS. GAIN is overseen by California's Department of Social
Services (CDSS) and administered by the 58 counties. This
report is part of a multi-year evaluation conducted for CDSS
by the Manpower Demonstration Research Corporation (MDRC).
The Findings in Brief
Each of the 33,000 sample
members was assigned at random to either an experimental group
(who were subject to GAIN's participation requirements) or
a control group (who were precluded from the program but could
seek other services in the community on their own). The two
groups were tracked over time, and the differences between
them (e.g., in earnings and welfare receipt) constitute the
effects or "impacts" of GAIN - the difference the
program made.
Single Parents (AFDC-FGs)
Overall. Over
the entire three-year follow-up period, and across the six
counties, GAIN produced increasing earnings impacts for single
parents (AFDC-FGs), a group mostly with children age 6 or
older when they enrolled in the study. In year 3, average
earnings for the experimental group were $636 higher than
the control group's average (a 25 percent gain); for the full
three years, they were $1,414 higher (a 22 percent gain).
(Earnings for each group were averaged over all members
of each group, including those who did not work as well as
those who did.) Moreover, some data point to sustained or
still larger earnings impacts in the future. GAIN also continued
to produce welfare savings in year 3 at the same level as
in year 2. In year 3 and across the six counties, experimental
received an average of $331 less in AFDC payments (an 8 percent
reduction) compared to the control group average; the reduction
was $961 (6 percent) for the entire three-year period. Longer-term
trends suggest a gradual tapering off of these welfare effects
in the future.
GAIN also had an effect on
the rate of AFDC case closures, but it was not large. Across
all six counties, over half of the experimental group was
on AFDC in the last quarter of the three-year follow-up period
(53 percent, or only 3 percentage points lower than the rate
for controls).
County-Specific.
GAIN's impacts on single parents varied across the six counties
in the study. Riverside County, which had unusually large
first- and second-year earnings gains and welfare savings,
again produced large effects in year 3. Over the three-year
period, Riverside increased the experimental group's earnings
by an average of $3,113, a 49 percent gain over the control
group average. It reduced welfare payments by $1,983, a 15
percent reduction compared to the control group. These impacts
were the biggest for any of the six counties, and are greater
than those found in previous large-scale experimental studies
of state welfare-to-work programs. Riverside also produced
large earnings gains and welfare savings for a special sample
- single parents with children under the age of 6.
GAIN's three-year impacts
on earnings were moderate to large in three of the other five
counties: $1,492 in Alameda (a 30 percent increase above the
county's control group average), $1,474 in Butte (a 21 percent
increase), and $1,772 in San Diego (a 22 percent increase).
Tulare produced a moderate impact ($518) in year 3, but its
average effect for the full follow-up period was under $400,
as was true in Los Angeles. Four of the remaining five counties
(all but Tulare) achieved welfare savings for AFDC-FGs for
the three-year period, ranging from an average of $782 per
experimental in Alameda (a 4 percent reduction compared to
the control group average) to $1,136 in San Diego (an 8 percent
reduction).
Benefits and Costs.
GAIN was a relatively expensive program compared to the simpler
and primarily job search welfare-to-work programs of the 1980s.
For the single-parent sample in all six counties combined,
over a five-year period, county welfare departments were estimated
to have spent an average of $2,899 per experimental, about
60 percent of which was for case management functions. In
addition, schools and other non-welfare agencies spent $1,515
per experimental to provide education and training instruction
as part of the GAIN program, bringing the total cost of GAIN
to $4,415 per experimental. Another important cost number
is the net cost per experimental, which measures the
government's net expenditures after adding the cost of education
and training activities experimentals entered on their own
after leaving GAIN, and then subtracting the cost of services
that members of the control group received on their own. The
net cost over five years was $3,422 per experimental for the
six counties combined, but varied widely by county, from under
$2,000 per experimental in Riverside and San Diego to over
$5,500 in Alameda and Los Angeles. The higher costs in the
latter two counties, which enrolled only long-term welfare
recipients into their GAIN programs, reflect, to an important
extent, a greater net increase in the use of education and
training activities in Alameda and Los Angeles compared to
the patterns in other counties.
Net cost estimates are key
because they are used in the benefit-cost study to determine
whether the program costs or saves taxpayers money. That analysis
also assesses whether people in the experimental group are
made financially better off by the program. (The benefit-cost
analysis does not take into account non-monetary gains or
losses.)
When measured earnings gains
are compared to welfare reductions and other losses over five
years, welfare recipients in five of the six counties (Alameda,
Butte, Riverside, San Diego, and Tulare) were, on average,
better off financially as a consequence of the GAIN program.
Net benefits ranged from $948 per experimental in San Diego
to $1,900 per experimental in Riverside, for an overall average
of $923 per experimental.
The GAIN program in two counties
- Riverside and San Diego - resulted in government budgets
coming out ahead. A third county - Butte - produced a "break-even"
effect, while the results were negative in the remaining three
counties. From the government budget perspective, a positive
result means that, on average, for every extra dollar the
government invested per experimental (above and beyond the
public cost of education and training controls received on
their own initiative), it got more than a dollar back in the
form of reduced costs for AFDC and other transfer programs
and increased tax payments arising from experimentals' increased
employment. This return was exceptionally large in Riverside
- $2.84 per every net $1 invested. The return was $1.40 per
$1 in San Diego, and $1.02 per $1 in Butte, but less than
a dollar ($.76) per $1 for all six counties together. It is
worth mentioning that return per net dollar invested is a
standard of success by which few social programs are assessed.
Heads of Two-Parent Families
(AFDC-Us)
GAIN also produced earnings
gains and welfare savings for the heads of two-parent families
(AFDC-Us), who make up about 18 percent of all AFDC cases
in California. Although the longer-term trends were not as
impressive as they were for single parents, GAIN's earnings
effects over the full three-year follow-up period were moderate
to large in three counties (Butte, Los Angeles, and Riverside,
although they were declining over time in Riverside). They
were especially large in Butte, reaching $3,295 per experimental.
The same three counties also produced moderate to large welfare
savings, as did San Diego. GAIN's benefit-cost results for
AFDC-Us show a large positive effect from the perspective
of welfare recipients solely in Butte, and a modest positive
return on the government's investment in Butte ($1.22 per
net $1 invested) and Riverside ($1.61 per net $1 invested).
In sum, the results of this
evaluation show that the GAIN program can work, especially
for single parents on welfare, who account for about 82 percent
of California's welfare caseload. For that group, both welfare
recipients and the government budget came out ahead in two
counties as a result of GAIN, with one county (Riverside)
producing the most impressive results yet observed for a large-scale
welfare-to-work program. Of the remaining four counties, three
made welfare recipients better off, but without producing
net budgetary savings (although the government essentially
"broke even" in one). An important open question
is whether some of the implementation approaches of the better-performing
counties, especially those of Riverside, can be adapted by
other localities and produce similarly impressive results.
The GAIN Program
Model
A key feature of GAIN, which
distinguishes it from most other welfare-to-work and JOBS
programs, is the way it uses educational and basic skills
levels to sort registrants into one of two service streams.
Those who do not have a high school diploma (or a General
Educational Development certificate - a GED) or fail to achieve
predetermined scores on both parts of a math and reading test
or are not proficient in English are deemed by GAIN to be
"in need of basic education." These individuals
can choose to attend a basic education class - Adult Basic
Education (ABE), GED preparation, or English as a Second Language
(ESL) instruction - or a job search activity first, but if
they choose job search and fail to obtain employment, they
must then enter basic education. Registrants judged "not
in need of basic education" - those who pass both parts
of the math and reading test and possess a high school diploma
(or a GED) - usually must participate in job search first.
Registrants already enrolled in education and training programs
when they enter GAIN may continue in those activities if the
activities meet certain criteria (e.g., they must prepare
registrants for occupations in need of workers in the local
labor market, and registrants must be able to complete the
training within two years after enrolling in GAIN). Participants
in any of these three sequences who do not find employment
after completing their initial activities undergo an employability
assessment designed to help them choose their next activity,
e.g., skills training, vocationally oriented post-secondary
education, on-the-job training, or unpaid work experience.
Any GAIN registrant, who, without "good cause,"
fails to participate in GAIN's orientation and services may
incur a "sanction," i.e., a reduction of the welfare
grant. (The grant level in California is one of the nation's
highest.)
The GAIN Evaluation
The six counties selected
to participate in the study of GAIN's impacts capture a wide
variety of local conditions and population characteristics
account for more than one-third of the state's GAIN caseload
and more than one-half of its AFDC caseload. Three counties
are in southern California: Los Angeles, with about
one-third of the state's caseload and a welfare population
larger than all but a few states'; San Diego, with
the state's second-largest caseload; and Riverside,
a county encompassing both urban and rural areas. Two counties
are in northern California: Alameda, an urban county
that includes the City of Oakland, and, further north, the
county of Butte, which had the smallest population
of the six counties. Tulare is located in the largely
agricultural, rural Central Valley. (Table
11, at the end of this summary, presents a brief profile
of each county.)
It is important to stress
that this report's descriptions of the counties' strategies
for implementing GAIN are based on information collected no
later than mid-1991, and prior to that in most cases. This
is the relevant information for describing the research sample's
actual experiences in GAIN. However, some of the information
does not portray the counties current modes of operating
GAIN. All of the counties have continued to revise their implementation
strategies as they have acquired more experience in operating
this very complex welfare-to-work initiative, and in response
to changes in funding and other circumstances.
The findings on GAIN's implementation,
effectiveness, and benefits and costs come from a study of
33,000 applicants for and recipients of AFDC whose participation
in GAIN was mandatory, i.e., a condition for receiving their
full welfare grant. This group included single heads of families
(AFDC-FGs, who are usually mothers) mostly with children age
6 or older, and all heads of two-parent families (AFDC-Us,
typically fathers). (It is important to note that almost one-third
of Alameda's sample consisted of single parents with children
younger than age 6.)
During the period in which
members of the research sample enrolled in GAIN and thus became
part of the study (March 1988 to June 1990), four of the six
counties had sufficient resources to enroll all registrants
in their caseloads who were mandatory for GAIN under the pre-JOBS
rules. The other counties - Alameda and Los Angeles - focused
exclusively on long-term recipients, in conformity with GAIN's
rules in cases where resources did not permit serving all
those required to participate.
To determine the effects of
GAIN, mandatory registrants who attended an orientation to
the program were randomly assigned to either an experimental
group (who were subject to GAIN's participation mandate) or
a control group (who were precluded from GAIN but could seek
other services in the community). Random assignment assured
that the two groups did not differ systematically on measured
and unmeasured background characteristics when they entered
the study, and that any differences in their subsequent labor
market and welfare experiences could be attributed with confidence
to the GAIN program. The two groups' employment rates, average
earnings, average AFDC payments, and other outcomes were compared
over the course of the follow-up period, and the differences
between them are referred to as the estimated "impacts"
of GAIN. The data used in this study came from a variety of
sources, including automated employment, earnings, and welfare
records for the full 33,000-person sample, a registrant survey
administered two to three years after orientation to a subsample
of experimentals and controls in five counties (excluding
Butte because of the evaluation's limited survey budget),
and program participation and fiscal information obtained
from the counties and various state agencies.
Findings on
Program Implementation
Not surprisingly, given California's
state-supervised but county-operated welfare system, and the
absence of evidence when GAIN started as to what strategies
would work best, the six counties varied in how they sought
to prepare registrants for employment. Viewing almost any
job as a positive first step, with advancement to come by
acquiring a work history and learning skills on the job, Riverside's
staff placed much more emphasis on moving registrants into
the labor market quickly than did the staff in any other county.
Most distinctive was Riverside's attempt to communicate a
strong "message" to all registrants (even those
in education and training activities), at all stages of the
program, that employment was central, that it should be sought
expeditiously, and that opportunities to obtain low-paying
jobs should not be turned down. The county's management underscored
this message by establishing job placement standards as one
of several criteria for assessing staff performance, while
at the same time attempting to secure the participation of
all mandatory registrants. In addition, the county instituted
a strong job development component to assist recipients in
gaining access to job opportunities.
Alameda illustrates a very
different approach. Its GAIN managers and staff believed strongly
in "human capital" development - the use of education
and training as a path to getting jobs that offer a better
chance to get off or stay off welfare. Within the overall
constraints imposed by the GAIN model's service sequences,
Alameda's staff encouraged registrants to be selective about
the jobs they accepted and to take advantage of GAIN's education
and training to prepare for higher-paying jobs. Butte, Los
Angeles, San Diego, and Tulare took approaches falling between
those of Riverside and Alameda, but closer to Alameda's than
to Riverside's.
Over 90 percent of experimentals
said on the registrant survey that they believed it was "likely"
or "very likely" that their AFDC grants would be
reduced if they were assigned to a GAIN activity but did not
go. Casefile records showed that up to about 6 percent of
experimentals (in Los Angeles and Riverside) were sanctioned
within the first 11 months after GAIN orientation, although
self-reported information from the survey and interviews with
GAIN staff suggest that the rates rose over time in all the
counties. Evidence also suggests that case managers in Los
Angeles and Riverside were quickest to invoke the "threat"
of sanctioning in response to noncompliance. About half to
three-quarters of survey respondents believed the participation
mandate to be "fair" and "a good idea,"
and only about one-quarter of respondents in both the experimental
and control groups agreed with the statement, "Making
welfare mothers work if they don't want to is bad for their
children."
Impacts
on Participation in Employment-Related Activities for AFDC-FGs
An important measure of the
GAIN intervention, a major determinant of its net costs, and
a potentially key influence on its impacts is the extent to
which experimentals had different participation patterns than
controls.
To determine GAIN's effect
on experimentals' use of employment-related activities, the
evaluation compared experimentals' rates and duration of participation
in all such activities (including GAIN and post-GAIN participation)
with the amount of participation in non-GAIN activities by
the control group. The difference in the amount of participation
represents the "impact" of GAIN, which tells how
much experimentals' participation changed compared
to what it would have been in the absence of GAIN.
Few controls (4 percent) participated
in job search activities, which, in comparison to opportunities
for education and training, are not widely available in the
community. Moreover, few (8 percent) participated in basic
education classes (for ABE, GED, and ESL instruction). Although
more widely available, basic education may have been of less
interest to controls than occupational skills training (nor
was it generally needed by those who already had a high school
diploma or GED). Only a handful of controls took part in unpaid
work experience and on-the-job training (OJT) assignments.
In contrast, a full 23 percent participated in vocational
training or post-secondary education.
Given that the GAIN model
requires most participants to enter upfront job search or
basic education as their initial GAIN activity, it is not
surprising that GAIN's largest impacts were on the use of
these two activities. Across all six counties, 29 percent
of experimentals participated in job search compared to only
4 percent of controls, for a difference of 25 percentage points.
Similarly, GAIN increased experimentals' participation in
ABE, GED, and ESL activities (taken together) by 28 percentage
points. The program had little overall impact (3.3 percentage
points) on the percentage who participated in vocational training
or post-secondary education, although, as discussed later
in this summary, it did in some counties (especially Alameda)
for registrants determined not to need basic education. Few
experimentals took part in unpaid work experience (PREP) or
OJT. (More recently, the use of PREP has increased in several
counties.)
Impacts on
Employment, Earnings, and Welfare Outcomes for AFDC-FGs
Impacts on Earnings
and Welfare Payments
-
GAIN increased the
average earnings of experimentals by 25 percent in the
third year after orientation, continuing its trend of
progressively stronger earnings effects over time. It
reduced experimentals' average AFDC payments by 8 percent,
a result that reflected a leveling off of GAIN's impacts
on this measure.
The average earnings for all
experimentals and all controls were calculated for
the full sample, including people who did not work (and whose
earnings were counted as zero). Averaged across the six counties,
with each county given equal weight, earnings for AFDC-FGs
in the third year (as shown in official automated earnings
records) were $3,159 per experimental group member and $2,523
per control group member. This yields an earnings gain, or
impact, of $636 per experimental (or 25 percent of the average
control group member's earnings), as shown in the "all
counties" section of Table
1. (This, again, is an average that includes sample members
who did not work at all; those who worked benefited more than
this $636 suggests.) Welfare savings were $331 per experimental
in year 3 (i.e., AFDC payments were 8 percent lower than the
average payments of $4,163 for controls). As indicated by
the asterisks for the "all counties" rows in Table
1, these results were statistically significant, meaning
that one can have greater confidence that they were due to
the program rather than to statistical chance. The earnings
impacts compare favorably with the three-year results for
simpler (mostly job search) programs studied previously, and
the AFDC impacts compare very favorably.
Over the entire three-year
follow-up period, GAIN's earnings impacts grew progressively
larger. Averaged across the six counties, with each county
given equal weight, the program's impact on earnings nearly
doubled between the first and second years of follow-up and
rose by another 24 percent between the second and third follow-up
years, reaching $1,414 per experimental for the entire period.
(See Table 1 and
Figure 1A.) An analysis
of GAIN's effects for an early cohort of sample members (i.e.,
those who entered the study early on and for whom more quarters
of follow-up are available) points toward sustained or still
larger earning impacts after the third year.
GAIN's effects on AFDC payments
leveled off in year 3, but totaled $961 for the full three-year
period. (See Table 1
and Figure 1B.)
After having grown by about 23 percent between years 1 and
2, they were about the same in year 3 as in year 2. Longer-term
trends for the early cohort suggest a gradual tapering off
of these welfare effects in the future.
-
GAIN's impacts varied
by county. One county (Riverside) had large earnings gains
and welfare savings in all three follow-up years. Three
counties (Alameda, Butte, and San Diego) had more moderate
earnings gains and welfare savings. Of the two remaining
counties, one (Los Angeles) achieved welfare savings but
with little effect on earnings gains, while the other
(Tulare) produced earnings gains but with little effect
on welfare payments.
Riverside, which had unusually
large first- and second-year earnings gains and welfare savings,
also produced large third-year effects on AFDC-FGs. (See Table
1 and Figure 1.)
Over the entire three-year period, the experimental group's
earnings in Riverside were $3,113 higher, on average, than
the control group's earnings, an increase of 49 percent. Their
welfare payments were $1,983 lower, a 15 percent reduction
compared to the control group. These impacts were the largest
in any of the six counties, and are larger than those found
in previous large-scale experimental studies of state welfare-to-work
programs. They are notable as much for their consistency as
for their magnitude: Riverside had statistically significant
earnings gains for many key subgroups of the single-parent
research sample, and these gains were almost always accompanied
by welfare savings. Such a consistent pattern was not found
in any other county.
Among the other five counties,
three (Alameda, Butte, and San Diego) had middle-level three-year
earnings impacts ($1,474 to $1,772 per experimental, or 21
to 30 percent above the control group average). Also of note
was the $518 earnings impact in year 3 for Tulare, where positive
and statistically significant effects were observed for the
first time. Four of these five counties (Tulare was the exception)
achieved moderate welfare savings (ranging from $782 per experimental
over the three years to $1,136, or a 4 to 8 percent reduction).
(The three-year earnings and welfare impacts in Butte were
not statistically significant, possibly owing to the small
control group sample size there.)
In Los Angeles, the finding
that GAIN produced welfare savings but had little effect on
earnings ($260, and not statistically significant) may have
resulted from GAIN's producing an increase in the rate of
employment, but in jobs that were low-paying, of short duration,
or both. The welfare savings may also partly reflect the influence
of financial sanctions (grant reductions) for noncompliance
with GAIN's participation mandate and any effect the mandate
may have had in increasing or hastening case closures among
experimentals who were working "off the books."
It is also worth noting that although the earnings impact
in Los Angeles was small overall, this was not true in all
five of the county's GAIN offices. The two offices located
outside of central-city areas produced three-year earnings
impacts exceeding $2,100 per experimental (an effect that
was statistically significant in one office). None of the
other offices, all of which were in central-city locations,
produced an earnings gain.
In Riverside, each of the
four local offices operating GAIN in the four economically
diverse regions of that county produced large and statistically
significant earnings gains and welfare savings. San Diego
also had consistently positive results (though not always
statistically significant) across most of its local offices,
but Tulare did not. (Alameda and Butte each had only one GAIN
office.)
All in all, the evidence of
impacts across the six counties shows that GAIN can produce
earnings gains, welfare savings, or both within a three-year
period, even when it is operated in very different ways and
under different circumstances. This is an encouraging finding
because local conditions will always vary across counties
and because some variation in key implementation practices
is inevitable.
Impacts
on Employment Rates, Earnings Levels, and Job Quality
For the six counties combined,
automated official records show that 40 percent of experimentals
had worked at some time during year 3 compared to 34 percent
of controls, resulting in a statistically significant difference
of 6 percentage points (see
Table 2). A similar impact is found when the proportions
of experimentals and controls ever employed over the entire
three-year period are compared (57 percent versus 51 percent,
respectively). GAIN's impact on the rate of employment was
largest in Riverside, where it exceeded 9 percentage points
in year 3 and almost 14 percentage points over the full follow-up
period. Despite this accomplishment, the data in Table
2 imply that, across the six counties, about two-thirds
of experimentals and controls did not work during year
3, and almost half never worked during the entire three-year
period. In response to a question on the registrant survey
in four counties (Alameda, Riverside, San Diego, and Tulare),
about 60 percent of experimentals who were not working at
the time of the interview said that they were not looking
for work. Of that group, 28 percent cited their own ill health
or disability as the most important reason, 4 percent cited
the ill health or disability of their children, and 22 percent
said that they were in a school or training program. Only
4 percent said that the main reason they were not looking
for work was that they could not afford or arrange for child
care (perhaps in part because the study sample was composed
largely of women with no preschool-age children), although
10 percent said that their major reason was that they wanted
to stay home with their children.
Of those who had never worked
during the survey follow-up period, only 34 percent said that
they had heard of the Earned Income Tax Credit (EITC), a federal
tax credit for low-income workers intended to enhance the
financial payoff from working. Of those who had worked, 54
percent said that they had heard of it.
In Riverside and, to a lesser
extent in San Diego, GAIN appears to have produced earnings
impacts because experimentals had higher employment rates
and more quarters of employment, but the jobs they held paid
about as much, on average, as the jobs held by controls. In
Alameda and Butte, in contrast, approximately half the earnings
gains were associated with increased earnings per quarter
of employment for experimentals, implying that, on average,
experimentals who worked held better jobs than controls who
worked.
These differences across the
counties are also reflected in the characteristics of the
most recent jobs reported on the registrant survey by experimentals
and controls who had worked at some time during the two- to
three-year follow-up period. In Riverside, similar proportions
of employed experimentals and employed controls
(64 percent) had worked full-time (i.e., 30 hours a week or
more) in their most recent job, and average weekly wages were
somewhat lower for all workers in the experimental group ($191
per week) than for all workers in the control group ($206).
In contrast, employed experimentals in Alameda got jobs providing
more hours of work per week than the jobs obtained by employed
controls (e.g., 59 percent versus 55 percent, respectively,
were full-time), and higher weekly wages for those working
($209 versus $167).
It is also of interest that
approximately 28 percent of employed experimentals in the
four counties had jobs providing health care coverage. Among
controls, the rate was 25 percent.
Table
3 shows that, for all six counties combined, about 20
percent of experimentals earned at least $5,000 in
year 3 compared to 16 percent of controls, for an impact of
almost 4 percentage points; 12 percent of experimentals, compared
to 9 percent of controls, earned at least $10,000 - an amount
of money that exceeds the poverty line for a single parent
with one child.
Another way to view earnings
levels is to consider what proportion of workers, rather
than all experimentals or all controls, earned above certain
thresholds. Although experimental-control differences on such
a measure are not true estimates of GAIN's impacts (since
the background characteristics of those who found jobs in
each group may not have been equivalent), they illustrate
that many of those who did find work had more substantial
earnings. For example, for all six counties combined, about
31 percent of all employed experimentals earned above
$10,000 in year 3. Among employed controls, the rate
was 27 percent.
To approximate GAIN's effects
on poverty, the analysis compared sample members' total year
3 earnings, AFDC payments, and Food Stamps with the official
poverty line, taking into account the size of each sample
member's family at the time of GAIN orientation. (In 1992,
the poverty line for a single parent with one child was $9,190.)
The income measure used here is different from the Census
Bureau's official poverty measure in that Food Stamps are
not counted in the official measure, while other family income
not measured in the GAIN evaluation is counted. The analysis
suggests that GAIN helped move some families out of poverty:
20 percent of the experimentals across the six counties, compared
to 17 percent of the controls, had a combined income above
the poverty line. In other words, experimentals' poverty rate
was reduced by 3 percentage points. This impact reached almost
5 percentage points in Butte and Tulare.
Impacts on Case Closures
Table
2 shows the proportion of sample members who had received
any AFDC payments in the last quarter of each follow-up year.
The proportion of experimentals on AFDC had dropped to 53
percent (for all six counties combined) by the end of the
three-year period. However, only a portion of this change
can be attributed to GAIN, since the control group experienced
a similar decline. Nonetheless, the counties collectively
produced a reduction of 3 percentage points in the proportion
of experimentals receiving welfare by the end of year 3, ranging
from under 2 percentage points in Butte and Tulare to over
5 percentage points in Riverside.
Table
3 (bottom panel) shows the proportion of people who had
both worked and received no AFDC payments during the
last quarter of the follow-up period. This combined status
comes closer than any other measure in this study to representing
the achievement of "self-sufficiency through employment."
By this criterion, about 19 percent of experimentals (for
all six counties combined) achieved self-sufficiency by working
compared to 16 percent of controls, for a small (statistically
significant) impact of almost 3 percentage points. The impact
on this measure was highest in Riverside and Alameda, where
it exceeded 4 percentage points. (During this same quarter,
another 10 percent of experimentals both worked and received
welfare.)
Welfare recipients who leave
AFDC often return to the rolls. Across the six study counties,
27 percent of experimentals who left AFDC for at least one
full quarter during the first half of the follow-up period
(i.e., from quarters 2 through 7) returned to AFDC before
the three years were out. (This rate ranged from 22 percent
in Los Angeles to 30 percent in Tulare.) Nonetheless, three
counties increased the likelihood that experimentals would
get off welfare and remain off the rolls. For example, 39
percent of all experimentals in Riverside, compared to 35
percent of all controls, had left AFDC during the first half
of the three-year follow-up period and did not return during
the rest of that period. This 4 percentage point difference
was statistically significant and accounts for more than half
of Riverside's impact of nearly 8 percentage points on the
total percentage of experimentals who left AFDC within
the first half of the follow-up period. Los Angeles and San
Diego each had an impact of 3 percentage points (statistically
significant) on the likelihood of exiting AFDC and remaining
off welfare through the end of the follow-up period, but little
effect was detected in the other three counties (Alameda,
Butte, and Tulare).
Impact
Findings for Selected AFDC-FG Subgroups
A central question for GAIN
is whether particular subgroups of welfare recipients are
or are not affected by the services the program offers and
by its participation mandate. All GAIN registrants were classified
into two groups for whom the GAIN program model established
different service sequences: those determined "not in
need of basic education" and those deemed "in need
of basic education." Overall, GAIN produced earnings
gains and welfare savings for both of these subgroups among
AFDC-FGs.
Three counties (Alameda, Riverside,
and San Diego) produced large earnings gains - ranging from
about $3,000 to $4,000 - for registrants determined not
to need basic education, as shown in the top panel of Table
4. Two of these counties (Riverside and San Diego) also
produced large welfare savings, while the third (Alameda)
did not. (The pattern in Alameda could have occurred if its
earnings impact was concentrated among individuals who, during
the follow-up period, would have left welfare and worked even
in the absence of GAIN, but in lower-paying jobs.) In contrast,
Los Angeles achieved large welfare savings for this subgroup,
but more modest (and not statistically significant) earnings
gains.
Alameda's success (noted above)
in raising the quality of jobs suggests that the use of job
search to explore career options, combined with subsequent
participation in vocational training and post-secondary education,
may have played a role in producing Alameda's earnings impact.
As the top panel of Table
5 shows, Alameda raised experimentals' participation in
training and post-secondary education 16 percentage points,
on average, above the control group rate - a participation
impact that was higher than in the other counties; it also
had the largest impact on the duration of participation in
these activities. Moreover, Alameda increased the proportion
of experimentals in the not-in-need-of-basic-education subgroup
who received a trade certificate by almost 6 percentage points
(not statistically significant) and receipt of a Bachelor's
degree by 3 percentage points. In contrast, Riverside did
not increase participation in training and post-secondary
education, nor did it increase the receipt of education credentials,
implying that its earnings impacts for this subgroup came
about from other sources - possibly through a combination
of factors, including the large impact on participation in
job search activities (48 percentage points, as shown in the
top panel of Table 5)
and other program features that made Riverside distinctive.
(See the section above on implementation findings.) San Diego's
experience appears to have been closer to Riverside's in that
it did not have a large impact on the use of vocational education
and training.
For registrants who were
determined to need basic education, increasing experimentals'
use of ABE, GED, and ESL classes (relative to the use of those
classes by controls) may have contributed to positive earnings
impacts, for Butte, Riverside, and Tulare all had a positive
impact on the rate of participation in those activities (see
the bottom panel of Table
5 for the Riverside and Tulare impacts). All three counties
(Tulare to a lesser extent) also produced statistically significant
earnings increases, as shown in the middle panel of
Table 4. In addition, two of them (Butte and Riverside)
produced welfare savings. At the same time, the experience
of the other three counties indicates that even a large impact
on the use of basic education may not result in earnings gains.
For example, Alameda had a 56 percentage point impact on the
in-need-of-basic-education subgroup's rate of participation
in basic education, yet its three-year impact on this group's
earnings was relatively small.
If an impact on the use of
basic education contributes to an impact on earnings, the
mechanism by which this occurs may sometimes involve factors
other than simply an increase in basic skills or credentials.
For example, it is noteworthy that Riverside achieved its
earnings gain for this subgroup without having had an impact
on the proportion of experimentals who obtained a GED and
without having an impact on literacy skills. Furthermore,
impacts on GED attainment were found in Alameda (an 8 percentage
point impact), while impacts on the literacy test were concentrated
in San Diego - two counties that did not produce a
statistically significant increase in earnings for this subgroup.
It is possible that in Riverside
(and perhaps elsewhere) basic education may have increased
skills not measured by the literacy test used in this evaluation,
or increased participants' interest in - or self-confidence
about - working. Perhaps these kinds of influences, when combined
with other aspects of Riverside's implementation of GAIN (including
its strong employment message and its substantial impact of
31 percentage points on the rate of participation in job search
for the in-need-of-basic-education subgroup, as shown in the
bottom panel of Table
5), help to explain why Riverside achieved an impressive
earnings impact for this subgroup without improving measured
educational gains.
GAIN produced earnings and welfare savings
for a variety of other subgroups, including (in some counties)
registrants who had received AFDC for more than two years
prior to entering the program, showing GAIN's potential to
reach a difficult-to-serve population.
Among long-term recipients,
the total three-year earnings impact was moderate to large
(and statistically significant) in three counties (Alameda,
Butte, and Riverside), ranging from $1,492 to $3,538, as shown
in the bottom panel of Table
4. Three-year welfare savings of $782 to $2,184 were found
across five counties (and were statistically significant in
four of them). It is noteworthy that Riverside produced the
largest earnings gains and the largest welfare savings
for long-term AFDC-FG recipients. It also produced statistically
significant impacts on these outcomes when "long-term"
is defined more strictly to mean recipients who received AFDC
continuously for at least the six years prior to orientation.
The evaluation examined GAIN's
impacts on a variety of other subgroups and found evidence
of earnings gains and welfare savings, although not consistently
in all counties. Across racial and ethnic groups, the largest
impacts were found among whites and blacks. For blacks in
Alameda (who constituted almost 70 percent of that county's
sample), there was a relatively large year 3 earnings impact
of $1,020. These results in Alameda are especially interesting
because that county's sample was drawn entirely from relatively
long-term recipients and an inner-city area (Oakland). For
Hispanics in the three counties that had large samples of
Hispanics (Los Angeles, Riverside, and San Diego), only Riverside
produced a statistically significant earnings impact in year
3 ($920), but none of the three produced statistically significant
welfare savings for this group.
In some counties, GAIN also
achieved impacts for individuals facing conditions commonly
thought to reflect important barriers to employment. As previously
discussed, the program produced earnings gains and welfare
savings for subgroups with long welfare histories (as it did
for those who were welfare applicants or shorter-term recipients
when registering for GAIN). It also achieved impacts for those
with little employment experience prior to entering GAIN and
for those with two or more children. At the same time, however,
it had weak earnings effects for a "most disadvantaged"
subgroup, defined as sample members with multiple barriers:
more than two years' previous receipt of AFDC and no
employment in the year preceding GAIN orientation and
no high school diploma. Larger earnings impacts for this group
may be particularly difficult to achieve because of those
multiple barriers, although Riverside's success in doing so
shows GAIN's potential to reach even them.
Impact
Findings for Single Parents with Children Younger than Age 6 in Three
Counties
Under the JOBS legislation,
starting in July 1989, GAIN's participation mandate was extended
to single parents with children 3 to 5 years old at the time
of orientation. Although this group was not part of the main
research sample for the evaluation (except in Alameda), employment,
earnings, and welfare data were collected for a supplementary
sample of such individuals in Riverside and Tulare. This sample
was somewhat younger, on average, than the main sample, but
fewer than a quarter of them were under age 25.
Over the entire three-year
follow-up period, Riverside produced large average increases
in earnings ($3,511) and reductions in AFDC payments ($2,558)
for this group, just as it had for its main sample. Similarly,
Alameda showed a sizable earnings impact for this sample ($2,220),
as it had for its main sample, although the effect was not
statistically significant (perhaps because of a small sample
size). However, Alameda did not substantially reduce AFDC
payments for this sample (it had a somewhat larger effect
for the main sample). Tulare produced no earnings gains or
welfare savings for this group (although it achieved earnings
gains in year 3 for the main sample).
The Riverside Case
Management Experiment
A special study was conducted
in Riverside to test whether assigning registrants to staff
with smaller caseloads, and allowing staff to monitor them
more closely and work with them more intensively, would produce
larger impacts on earnings and AFDC. Using random assignment
procedures, experimentals and case managers were divided into
two groups: an "enhanced" group and a "regular"
group. The average registrant-to-staff ratio in the enhanced
group (53 to 1) was about half as large as the ratio for the
regular group (97 to 1).
Both the enhanced and regular
experimental groups obtained large gains in earnings and large
reductions in AFDC, but, contrary to what had been expected,
these impacts were not greater for the enhanced group. These
findings suggest that there may be little advantage to operating
a GAIN program - at least one like Riverside's - with caseloads
substantially below 100 registrants per case manager, and
that keeping them in the moderate range of about 100 to 1
may be one way of containing program costs without jeopardizing
program effectiveness.
Findings
on Program Costs for AFDC-FGs
This study calculated several
different types of cost estimates, including: the county
welfare department's average expenditure per experimental;
the total GAIN cost per experimental, which adds to
the welfare department cost the average expenditures by schools
and training providers for services provided to GAIN participants
as part of the GAIN program; and the net cost (or net
investment) per experimental. Net cost per experimental is
the total public expenditure on employment-related activities
per experimental - for post-GAIN activities as well as the
total GAIN cost - minus the public cost of (non-GAIN) services
to controls. Net cost is the cost measure used in the benefit-cost
analysis, discussed later in this summary. All cost estimates
cover a time horizon of five years after orientation (in order
to capture long-term participation in GAIN activities and
to be consistent with the benefit-cost analysis), and are
expressed in 1993 dollars.
Table
6A summarizes the average county welfare department expenditure
for each of the six counties. Four of the six (Butte, Riverside,
San Diego, and Tulare) spent between $2,000 and $2,700, while
the remaining two counties (Alameda and Los Angeles) spent
about $4,000 or more. Across the six counties, about 60 percent
of these expenditures were on activities that could be classified
as case management (including conducting orientations, appraisals,
and assessments; assigning registrants to activities; arranging
for support service payments; responding to noncompliance;
etc.). Among the other welfare department expenditures were
the costs of conducting (or subcontracting the operation of)
job club sessions and supervising individual job search activities,
paying schools to provide extra monitoring and attendance
data (to help the welfare department measure compliance with
GAIN's participation mandate), and paying for child care and
other support services (e.g., for transportation and such
ancillary items as books, tools, and uniforms).
The total cost of GAIN per
experimental is the sum of the GAIN-related expenditures of
the county welfare department and other agencies. Non-welfare
agencies - adult schools, community colleges, and other organizations
- provided the education and training for GAIN registrants
who were assigned to basic education classes, vocational training,
and post-secondary education to meet their participation obligation,
or who were participating in approved self-initiated activities
begun prior to entering GAIN. Thus, the expenditures made
by the non-welfare agencies to serve GAIN registrants are
considered to be GAIN-related costs, even though they were
not controlled directly by the county welfare departments.
For all six counties combined, these expenditures averaged
$1,515 (Table 6A).
Adding these GAIN expenditures to those made by county welfare
departments ($2,899) yields the total GAIN cost of $4,415
per experimental.
The pie charts in Table
6 show how this six-county total cost per experimental
was distributed across the key components of GAIN. The first
chart (Table 6C)
illustrates that the cost to the welfare department of processing
registrants through the orientation and appraisal stages of
the program (including following up on those who failed to
attend their scheduled orientation sessions), plus the cost
of assessments, accounted for about 17 percent of the $2,899
average welfare department GAIN cost, while expenditures on
registrants assigned to job search activities and basic education
(ABE, GED, or ESL activities) each accounted for about one-quarter
of those expenditures. (Again, this includes the cost of the
case managers' effort to monitor attendance and progress,
arrange support services, follow up on nonattenders, etc.,
for these two activities.) Another 8 percent was spent on
child care, and 12 percent was spent on other support services
(transportation and ancillary services). It is important to
note that, across all six counties, the average cost of GAIN
child care per experimental would have been higher if the
research sample had been composed mostly of parents with younger
children, a group that has a greater need for child care.
For those with schoolage children, GAIN activities were often
scheduled to take place while the children themselves were
in school. Also, those whose youngest child was a teenager
(up to about one-quarter of the research sample in some counties)
would not have been eligible for GAIN-funded child care.
The second pie chart (Table
6D) shows the distribution of total GAIN costs,
i.e., after adding in the expenditures by other agencies providing
the education and training received by GAIN participants while
they were enrolled in GAIN. It shows that of the total average
GAIN cost ($4,415), three-quarters is accounted for by expenditures
on registrants assigned to job search activities (16 percent),
basic education activities (31 percent), and vocational training
and post-secondary education (27 percent).
-
The total cost of GAIN
varied widely by county, ranging from under $4,000 per
experimental in four counties (Butte, Riverside, San Diego,
and Tulare) to almost $6,000 or more in two counties (Alameda
and Los Angeles).
Four counties - Butte, Riverside,
San Diego, and Tulare - had an average total GAIN cost (including
welfare department and non-welfare agency expenditures) in
the range of about $3,000 to $4,000, while Los Angeles spent
almost $6,000 per experimental and Alameda, more than $6,600.
GAIN costs were lowest in Riverside ($2,963) owing, to an
important extent, to Riverside experimentals' quicker departures
from the GAIN program and their shorter length of participation,
on average, in education and training activities in that county
compared to experimentals in other counties. The unusually
high costs in Alameda and Los Angeles (both of which served
only long-term welfare recipients) are attributable to a combination
of factors, including their experimentals' relatively long
lengths of stay in GAIN and heavy use of education and training
activities. In Los Angeles, this high usage was mostly in
basic education activities, while in Alameda it extended to
vocational training and post-secondary education as well.
Longer participation in activities also produces greater expenditures
for support services.
Net costs are key to determining
whether GAIN has been a cost-effective investment from the
perspective of government budgets. They represent the difference
between the five-year average total cost per experimental
(including public expenditures on experimentals who participated
in non-GAIN employment and training activities after leaving
GAIN) and the average cost per control for non-GAIN services.
The government's net cost per experimental for the six counties
combined is thus obtained by subtracting the total cost per
control ($1,472) from the total cost per experimental for
GAIN and non-GAIN activities ($4,895), which yields
$3,422 (after rounding). This number is presented in the last
column of Table 6B.
These costs were largest where the cost of GAIN itself was
highest - in Los Angeles ($5,789) and Alameda ($5,597) - and
lowest in Riverside ($1,597) and San Diego ($1,912).
Benefit-Cost Findings
for AFDC-FGs
The benefit-cost analysis
addresses three questions: Are welfare recipients financially
better or worse off as a result of the GAIN program? Is the
government's net investment in services for the experimental
group offset by subsequent budget savings? Does society as
a whole come out ahead or behind as a result of the program?
The analysis takes into consideration GAIN's effects on earnings,
AFDC payments, Food Stamps, and Unemployment Insurance payments,
fringe benefits, taxes, Medi-Cal (i.e., Medicaid) payments,
administrative costs for AFDC and other transfer programs,
and the net cost of employment-related services. It does not
formally incorporate intangible positive or negative effects
of the program, such as the increased sense of pride or feelings
of stress or loss of time with their families that registrants
might have felt in substituting work for welfare, or any enhancement
of their self-esteem from obtaining a GED or other education
credential through the GAIN program. The analysis also assumes
that no displacement of other workers occurred as a result
of employment gains for experimentals, because the displacement
effects could not be measured.
The benefit-cost estimates
presented in this summary cover the five years after GAIN
orientation, a time frame similar to the one used in most
previous MDRC evaluations of welfare-to-work programs. (Because
a full five years of follow-up data were not available for
earnings, welfare payments, and other outcomes, the overall
benefit estimates include some projected values, up to two
years for some sample members but less than that for most.)
It should be noted, however, that this probably is a conservative
estimate, since five years is not likely to be long enough
to capture the total effects of GAIN.
-
In five of the six
counties, experimentals, on average, were better off financially
as a result of the GAIN program.
As shown by the impact analysis,
GAIN increased the earnings of experimentals in most counties.
The measured and projected earnings gains and their associated
fringe benefits constitute the primary financial gain from
the standpoint of experimentals (referred to in benefit-cost
analyses as the "welfare sample perspective"). However,
these gains were offset to some extent by reduced AFDC payments
and other transfer payments.
Nonetheless, GAIN experimentals
- with the exception of those in Los Angeles - experienced
a net financial gain as a result of the program, averaging
$923 per experimental for the six counties combined over the
five-year period, as shown in Figure
2A and Table 7.
(The average net gain equals $1,420 when Los Angeles is excluded.)
In Los Angeles, experimentals' losses in transfer payments
(especially AFDC payments) exceeded their measured
earnings increases, leaving them with a net loss overall of
$1,561. (Any effect GAIN may have had on "off the books"
earnings is not considered in this analysis.) In all other
counties, experimentals realized an average net gain of between
$948 in San Diego and $1,900 in Riverside. It is noteworthy,
however, that in Tulare this positive result was achieved
with a smaller earnings increase and a smaller reduction in
AFDC payments compared to the other counties. In contrast,
Riverside's results, compared to all of the other counties,
reflect both a large increase in earnings and a large reduction
in welfare payments - in other words, a greater substitution
of work for welfare.
From the "government
budget perspective," the gains of the program include
reduced AFDC payments, reductions in other transfer payments,
reductions in transfer program administrative costs, and the
increased taxes paid by experimentals. The net expenditures
for GAIN and non-GAIN services constitute the net costs to
the government. Overall, the results for this perspective
- which sets a tough standard for programs to meet - are mixed,
as Figure 2B and
Table 7 show. Average
costs incurred by the government exceeded savings per experimental
by $3,054 in Alameda, $3,442 in Los Angeles, and $2,261 in
Tulare. There was a moderate net gain (i.e., savings
and increased tax revenues exceeded net costs) in San Diego
($767), and a quite large net gain in Riverside ($2,936).
In Butte, GAIN resulted in the government budget breaking
even (with a slight net gain of $54). The losses in Alameda
and Los Angeles to an important extent reflect the comparatively
high net expenditures on employment-related services per experimental,
especially for education and training activities. On average,
across the six counties, the GAIN program incurred a net loss
to the government budget of $833 within a five-year time horizon.
One can also consider the
cost-effectiveness of the GAIN program from the standpoint
of the government budget by estimating the value of budgetary
savings and tax increases per dollar of investment (i.e.,
per dollar of net costs). This measure is called return
to budget per net dollar invested. An average gain of
more than $1 means that the program brings in more than a
dollar's worth of additional revenues and savings for each
additional dollar spent on employment-related services to
experimentals; an average return that is less than $1 implies
a net loss for the government.
Riverside's program produced
$2.84 in increased revenues and savings for every net dollar
spent on experimentals, a substantial return to the budget.
(If Riverside had operated GAIN solely with the higher caseload
sizes assigned to staff in the "regular" case management
group, its return to the government budget would have been
higher than $2.84.) The program in San Diego and (to a slight
extent) Butte also returned more than $1 in revenues and savings
($1.40 and $1.02, respectively). Alameda, Los Angeles, and
Tulare returned less than $.50 per dollar of net costs; and
the six counties combined returned $.76, on average.
-
Overall, three counties
(Butte, Riverside, and San Diego) achieved a net gain
from the societal perspective.
The net financial gain or
loss to "society as a whole" is approximated by
summing the results from the welfare sample and government
budget perspectives. As Table
7 shows, Butte, San Diego, and especially Riverside achieved
a net financial gain from the societal perspective, and were
the only counties to do so. In Alameda and Tulare, the government
incurred a net loss but welfare recipients gained - a kind
of trade-off that policymakers may or may not find acceptable.
For experimentals determined
not to need basic education, Alameda (which served
longer-term welfare recipients) stands out as having produced
the largest net gain for the welfare sample ($5,328 per experimental).
At the same time, Alameda's average net cost per experimental
in this subgroup was unusually high ($7,161, compared to less
than $1,100 in Riverside and San Diego), in part because of
its high net increase in experimentals' use of vocational
training and post-secondary education. These expenditures,
in combination with the absence of substantial reductions
in AFDC payments, resulted in a substantial net loss for the
government budget ($6,041 per experimental), as shown in Table
7. Riverside and San Diego illustrate an alternative pattern:
Although they placed much less emphasis on vocational training
and post-secondary education, they too achieved a net gain
for the welfare sample (in the range of $3,000 per experimental),
although it was considerably smaller than in Alameda. Because
their expenditures were lower, these two counties also produced
a net gain for the government budget: by $3,576 in Riverside
and $2,610 in San Diego (a return of $4.36 and $3.95, respectively,
per net dollar invested).
For experimentals who were
determined to need basic education, GAIN resulted in a net
gain from the welfare sample perspective in only two counties
(Riverside and Tulare) and for the government budget in two
counties (Butte and Riverside). Riverside was the only county
of the six to produce a net gain for both of the basic education
subgroups from both the welfare sample and government budget
perspectives.
Summary of Impact and
Benefit-Cost Findings for AFDC-Us (Heads of Two-Parent Families)
-
GAIN produced earnings
gains for the heads of two-parent families (AFDC-Us) that
were about the same in year 3 as in year 2, and welfare
savings that were somewhat lower. Butte had the most impressive
earnings impacts, which were large and sustained over
time.
Averaging across five counties
(omitting Alameda because of a small sample size) yields three-year
earnings gains of $1,111 per AFDC-U experimental group member
(a 12 percent increase over the control group average) and
three-year AFDC impacts of $1,168 (a saving of 6 percent relative
to the average AFDC payments to controls). (See Table
8.)
The results varied considerably
by county. GAIN increased earnings in the three-year follow-up
period in three of the five counties - Butte, Los Angeles,
and Riverside. However, only in Butte did earnings impacts
increase from year 1 to year 2; they then held steady from
year 2 to year 3, reaching a total of $3,295 per experimental
over the entire three-year period.
Reductions in AFDC payments
were found in four counties - Butte, Los Angeles, Riverside,
and San Diego - although they were not statistically significant
in Butte (possibly because of a small sample size). Riverside's
welfare impacts were the largest: a saving of $2,064 per experimental
over the three years, or 14 percent of the average payments
to controls. Butte, Los Angeles, and San Diego were in a middle
range, while Tulare produced no AFDC impacts. It appeared
unlikely there would be much addition to total AFDC impacts
after year 3 except in Butte.
-
GAIN had a positive
impact on AFDC-U experimentals' rate of employment in
year 3 in three counties (Butte, Los Angeles, and Riverside).
However, it did not reduce the proportion on welfare.
Table
9 indicates that across the five counties included in
the AFDC-U analysis, nearly 45 percent of experimentals had
ever been employed in year 3, compared to 40 percent of controls,
a difference of almost 5 percentage points. This impact was
concentrated in Butte, Los Angeles, and Riverside. Although
Los Angeles had the largest impact (10 percentage points on
this measure), this effect did not translate into a correspondingly
large earnings gain, perhaps because the jobs were short-term,
low-paying, or both.
Table
9 also shows that GAIN had little overall effect on the
proportion of AFDC-Us receiving AFDC in the last quarter of
follow-up, although Butte did show a reduction (not statistically
significant) of almost 5 percentage points. In fact, the proportion
of both groups receiving welfare at the end of year 3 was
high in most counties, exceeding 50 percent (and reaching
78 percent in Los Angeles). These levels are comparable to
those found for the AFDC-FGs, which was not expected because
AFDC-Us are typically considered to be more "job-ready"
and shorter-term users of welfare. These patterns may partly
reflect the fact that the AFDC-U samples in several counties
included a relatively high proportion of registrants who were
not proficient in English. This was especially so in Los Angeles, |
