Full text of Working Papers (Federal Reserve Bank of Chicago) : Should We Teach Old Dogs New Tricks? : The Impact of Community College Retraining on Older Displaced Workers, Working Paper 2003-25

View original document

The full text on this page is automatically extracted from the file linked above and may contain errors and inconsistencies.

Federal Reserve Bank of Chicago

Should We Teach Old Dogs New
Tricks? The Impact of Community
College Retraining on Older Displaced
Workers
Louis Jacobson, Robert J. LaLonde and
Daniel Sullivan

WP 2003-25

November 2003

Should We Teach Old Dogs New Tricks? The Impact of Community College
Retraining on Older Displaced Workers

Louis Jacobson
Westat, Inc.
Robert J. LaLonde
University of Chicago and NBER
and
Daniel Sullivan
Federal Reserve Bank of Chicago

We thank the Washington State Training and Education Coordinating Board for
providing the data and technical assistance for this project. We have benefited from
comments by Eric French, James Heckman, Alan Kruegar, Derek Neal, Soloman
Polachek, Madeline Zavodny and seminar participants at Chicago, Princeton, Cornell, the
Federal Reserve Banks of Chicago and New York and the Bureau of Labor Statistics.
Some of this research was funded by the Washington State Workforce Training and
Education Coordinating Board and by the U.S. Department of Labor Employment and
Training Administration under contract numbers 99-0-0584-75-055-01 and K-6307-7-0080-30. The views expressed in this paper are solely those of the authors and do not
necessarily represent the views of the Washington State Workforce Training and
Education Coordinating Board, the U.S. Department of Labor, the Federal Reserve Bank
of Chicago or the Federal Reserve System.

Should We Teach Old Dogs New Tricks? The Impact of Community College
Retraining on Older Displaced Workers
Abstract
This paper estimates the returns to retraining for older displaced workers--those
35 or older--by estimating the impact that community college schooling has on their
subsequent earnings. Our analysis relies on longitudinal administrative data covering
workers who were displaced from jobs in Washington State during the first half of the
1990s and who subsequently remained attached to the state’s work force. Our database
contains displaced workers' quarterly earnings records covering 14 years matched to the
records of 25 of the state's community colleges.
We find that older displaced workers participate in community college schooling
at significantly lower rates than younger displaced workers. However, among those who
participate in retraining, the per-period impact for older and younger displaced workers is
similar. We estimate that one academic year of such schooling increases the long-term
earnings by about 8 percent for older males and by about 10 percent for older females.
These per-period impacts are in line with those reported in the schooling literature.
These percentages do not necessarily imply that retraining older workers is a
sound social investment. We find that the social internal rates of return from investments
in older displaced workers' retraining are less than for younger displaced workers and
likely less than those reported for schooling of children. However, our internal rate of
return estimates are very sensitive to how we measure the opportunity cost of retraining.
If we assume that these opportunity costs are zero, the internal rate of return from
retraining older displaced workers is about 11 percent. By contrast, if we rely on our
estimates of the opportunity cost of retraining, the internal rate of return may be less than
2 percent for older men and as low as 4 percent for older women.

I. Introduction
During the last decade there has been rising interest in polices that foster human
capital investments in pre-school aged children.1 Proponents of these policies point to
evidence that such investments produce impressive social returns, even though much of
their benefits are not apparent until children reach adolescence or become young adults.
The long time periods over which these benefits accrue, the low opportunity costs of
children’s time, and the possibility that the young acquire many skills more efficiently
than the old are compelling reasons to believe that policy should be biased toward
investments in the young.
At the same time that evidence has accumulated on the benefits of investments in
the young, policy makers have directed more of their workforce development
expenditures towards older, more established workers.2 In addition to those participating
in formal government-sponsored workforce programs, significant numbers of prime-aged
workers have sought retraining on their own in the nation’s two-year community
colleges. These heavily subsidized public institutions now report that upwards of 30
percent of their students are over 30 years old (Kane and Rouse, 1999). Despite increased
public expenditures on retraining older workers, there is no body of evidence showing
that these investments have a significant payoff (Leigh 1990, Corson et. al. 1993).
Moreover, this policy change runs counter to evidence that business firms invest much
less in the skills of prime-aged and older employees than they do in the skills of younger
1

These studies include Heckman (2001), Carneiro and Heckman (2003), Barnett (1993), Barnett and
Masse (200 ), Currie and Thomas (1995), Currie, Garces, and Thomas (2001), Karoly et. al. (1998. 2002),
Olds, (1997), Reynolds et. al. (2001).
2
For a discussion of the history of employment and training polices in the U.S., see LaLonde (2003). A
significant portion of public expenditures on classroom training in these programs subsidizes classroom
training in community colleges. Over time, community colleges have shifted their emphasis toward

workers (Ashenfelter and LaLonde, 1997). Presumably, firms invest less in on-the-job
training of older workers, because its net benefits are less than for their younger
employees.
This paper provides evidence on the returns to retraining for older workers by
estimating the impact that community college schooling has on the earnings of displaced
workers who seek retraining around the time of their job losses.3 We compare these
impacts for displaced workers who are 35 and older to impacts for younger displaced
workers who enroll in similar courses or programs and to estimates reported in the
schooling literature.
Our analysis relies on a longitudinal data base containing 14 years of
administrative quarterly earnings records for displaced workers matched to the records of
25 community colleges in Washington State. These records detail when and for how
long individuals were enrolled in school, how many credits they completed and the
subject matter of the courses. Our sample contains over 65,000 displaced workers who
lost jobs during the first half of the 1990s and who remained consistently attached to
Washington State's work force throughout the fourteen year period studied. Fifteen
percent of these workers enrolled in and completed community college courses around
the time of their job loss. More than one-half of these students were 35 or over when they
enrolled in school.
We find that older displaced workers participate in community college schooling
at significantly lower rates than younger displaced workers. This evidence suggests that
providing vocational education to their students (Freeman, 1974; Grubb 1993; Kane and Rouse 1999;
Jacobson, LaLonde, and Sullivan 2003).
3
This paper adds to the literature on the returns to community college schooling by reporting impacts for
students who are on average 43 years old and therefore much older than community college students

older workers do not expect to benefit as much from retraining. However, among those
who do participate, the per-period impact of community college schooling on subsequent
earnings is comparable for older and younger displaced workers. Moreover, these gains
are comparable to those reported in the literature for schooling acquired by children and
young adults. In addition, we find that these gains persist and show no sign of
depreciating. Indeed, it is more likely that these gains will appreciate over time as the
follow-up period lengthens.
These results on community college retraining indicate that old dogs can learn
new tricks. But when we consider the differences among displaced workers' remaining
work lives and in their opportunity costs of retraining, we find that the social internal
rates of return from investments in retraining are likely smaller for older displaced
workers than for their younger counterparts. As we discuss below, whether policy
makers should teach old dogs new tricks depends to a considerable extent on how we
measure the opportunity costs of retraining.
An important caveat to our findings is that we must “cope” without the benefit of
an experimental design (Ashenfelter, 1978, p. 47). In his seminal evaluation of the
Manpower Development and Training Act (MDTA) program, Orley Ashenfelter
observed that providing evidence that econometric methods replicate the results of a
social experiment that used randomized trails was important in producing credible
estimates of the impacts of training programs. To produce this evidence, he introduced
into the training evaluation literature the practice sometimes referred to as “backcasting”-examining whether training appears to affect outcomes before it occurs (Ashenfelter

studied elsewhere. See Grubb, 1993a,b; Kane and Rouse, 1995, 1999; Leigh and Gill, 1997)

1978; LaLonde 1986; Heckman and Hotz 1989; Angrist and Newey, 1991; Heckman,
LaLonde, and Smith, 1999).
Although backcasting cannot indicate conclusively whether non-experimental
estimates replicate those from an experiment, Ashenfelter argued that this exercise “may
serve as a signal of serious problems with the maintained hypotheses (associated with the
underlying econometric model).” (Ashenfelter, 1978, p. 51). Because our Washington
State sample of displaced workers contains both a large number of trainees and
comparisons and covers many time periods both before and after training, we can
perform similar checks with our data. These checks indicate that, despite our rich
econometric specification, our model may still be misspecified. Therefore, our estimates
may be subject to bias. However, we argue that our backcasting evidence suggests that
we are no more likely to have overstated than we are to have understated the impact of
retraining. Indeed we have some evidence that our estimates of long-term impacts,
especially for older displaced workers, maybe too low.
The rest of this paper is organized as follows: We present a framework for
interpreting estimates of the impact of community college schooling in section II. In
Section III, we introduce our data and discuss the characteristics of our samples of older
and younger displaced workers. The empirical relationship between displaced workers’
age and measures of participation in community college schooling is presented in Section
IV. In Section V, we present our econometric model and introduce several alternative
specifications of the impact of community college schooling. In Section VI, we present
our main empirical results along with the results from two extensions of our model.
These extensions include consideration of the impact of different types of schooling on

earnings and the results of the backcasting specification test. Some discussion and
concluding remarks follow in Section VII, here we also present some alternative
calculations of the net-benefits and internal rates of return from investments in
community college schooling.
II. A Model of Community College Participation
Motives for enrolling in community college schooling following job loss vary.
Individuals may enroll to enhance their skills. Their decision to invest in more schooling
depends on its impact on future earnings, their rate of time preference, the time remaining
in their work lives, and the direct and indirect costs of going to school. Besides the
human capital motive, individuals also may enroll in school in order to facilitate their
search for a new job (Heckman, LaLonde, and Smith, 1999). Both exposure to new skills
and to new networks may facilitate their job search. Finally, some schooling may
constitute consumption. During a period of unemployment, the cost of acquiring such
schooling might be especially low. These differing motives affect the likely impacts of
community college schooling on individuals’ earnings.
The human capital motive suggests one framework for assessing the relationship
between a person’s age, their decision to enroll in school, and the average impact of such
schooling for those who choose to enroll (Heckman, LaLonde, and Smith, 1999). For
purposes of illustration, we characterized the decision to enroll in retraining by the
following simple formulation of the benefits and costs of schooling:

(1) δi (1 – (1/(1+r))Ni)/r - Ci > 0.

In (1), the term δi denotes the per-period impact of retraining on person i's annual
earnings. The subscript i indicates that the impact of schooling varies among individuals
in the population. We assume that these impacts are drawn from a probability distribution
F(δi ). The term (1 – (1/(1+r))Ni)/r is the present value of $1 paid annually to an individual
annually for Ni years, where Ni denotes the number of remaining years in their work life,
and r is the real interest rate. Ci denotes the costs of schooling. These costs include both
the direct costs of schooling, such as tuition, supplies and transportation, as well as the
opportunity costs of schooling connected with spending less time working or searching
for a new job. This formulation may be modified in the conventional way to account for
the possibility that the impact of schooling either depreciates or appreciates over time.
This framework implies that if the distribution of per-period impacts is the same
for all displaced workers, and the cost of participating in retraining is comparable or
larger for older persons, then older displaced workers are less likely to enroll in school
following the loss of a job. However, among those who enroll, the per-period impact of
this schooling is likely to be larger for the older trainees than it is for their younger
counterparts. Older workers are less likely to enroll in schooling because they have fewer
remaining years left in their work lives and because they likely face higher opportunity
costs of schooling due to their higher foregone earnings. However, because of these
factors, given that they enroll, the average impact of schooling must be larger in order to
offset their shorter remaining work lives and possibly their higher opportunity costs of
retraining. Therefore, even if older workers learn as efficiently as younger workers (i.e.
their draw of δi comes from the same distribution) the decision process summarized by
(1) implies that we may find that community college schooling has a greater per-period

impact on the earnings of older trainees than on the earnings of comparable younger
trainees.
The possibility that older workers do not learn as efficiently as younger persons
does not necessarily change the foregoing result. Suppose we assume that the impact of a
unit of schooling for an older worker is less than it is for a younger worker in the sense
that Fold(δi ) > Fyoung(δi ). Under these conditions, even though in a random sample of
older workers few expect large earnings gains from schooling, it is still the case that the
average impact of schooling among those who enroll in school is likely to be larger for
older than for younger trainees. If older workers also have higher costs of attending
school, this result is reinforced.
Therefore even if older individuals in the population do not learn as efficiently as
younger individuals, among workers who choose to become trainees, we expect that,
under reasonable conditions, the average impact of schooling should be larger for older
trainees than it is for younger trainees. Instead, differences in the distribution of impacts
for older and younger displaced workers should manifest themselves in differences
between older and younger displaced workers' participation rates in retraining (cf.
Heckman and Honoré, 1990). If older workers do not learn as efficiently as younger
workers, a smaller percentage should enroll in school in the first place. But this lower
enrollment rate does not imply that we should estimate that the average impact of
schooling among those who enroll is smaller for older workers than for younger workers.
The foregoing framework does not capture an important dimension of displaced
workers’ decisions to enroll in community college schooling. As we observe in the next
section, displaced workers decide not only whether to enroll but also how much training

to acquire. The above framework makes sense if community college consists of one
course. In fact, degree and certificate programs require students to complete many
courses. The problem with the framework as stated is that as long as it makes sense to
enroll in community college, it makes sense to enroll and complete as many courses.
However, we do not observe this behavior in our data. Most displaced workers who
enroll in community college courses complete only a few classes. One modification to
our framework that is consistent with this pattern of behavior is to allow for the
possibility that impacts are concave in number of credits completed or that the costs are
convex in the number of credits completed.4 We explore the former possibility in our
empirical work below.
III. Administrative Earnings and Community College Data
A. The Benefits of Administrative Data in Studies of Retraining
The value of large longitudinal databases like the one that we use in this paper has
been long recognized in the training evaluation literature. Ashenfelter’s study of the 1964
MDTA cohort began a long tradition among U.S. academic evaluators of using
longitudinal administrative data to evaluate employment and training programs
(Ashenfelter, 1975,1978; Heckman and Robb, 1985, Heckman, LaLonde, and Smith,
1999). In his study, Ashenfelter merged annual administrative earnings data from the
Social Security Administration to the administrative records from the MDTA program.

The possibility that there are fixed costs associated with attending school during any given quarter also
does not address the foregoing shortcoming of (1). The presence of fixed costs of attending school makes it
more likely that those who enroll complete many classes. Suppose that older workers face higher fixed
costs associated with their participation in school. In this case we expect that among those who enroll,
older workers complete more classes than their younger counterparts. However as we show below, this
prediction is inconsistent with the participation patterns observed in our data. Our data suggest that, all
other things equal, the fixed costs associated with acquiring retraining are relatively small and similar for
older and younger workers.

This merged data set contained relatively large samples of trainees and of comparison
group members. Each observation in his data set contained annual earnings for more than
5 years prior to and 5 years after the year of training.
Ashenfelter observed that such administrative data addressed two key problems that
arose when estimating the effect of training and informing public policy discussions of
the merits of these programs as social investments. (Ashenfelter, 1978, p. 47) First,
because public investments in these programs are usually relatively small on a per-person
basis, we expect that training should have small impacts on annual earnings and hourly
wages. The task of precisely measuring these expected small impacts are complicated
because outcomes of interest, especially earnings, exhibit high variance in the population,
even for the subset of the population likely to participate in training. Administrative data
address this problem by allowing researchers relatively inexpensive access to the
outcomes for large samples of trainees and non-trainees.
The second problem that Ashenfelter identified arises because these programs
when successful have benefits that should accumulate over a long period of time.
Estimates of the impact of training in the short-term likely provide an incomplete picture
of the impact of these programs. It may be that even successful programs appear to have
little or no effect during the first year or two after the program ends. Unfortunately, it is
usually expensive to locate and survey training participants over long periods of time.
Administrative data provide a relatively inexpensive way to follow trainees and nontrainees for relatively long periods before and after training.

B. The Washington State Administrative Data
Our sample is drawn from the records of all persons who lost jobs in Washington
State and who filed a valid claim for unemployment insurance (UI) benefits.5 To
construct our sample we used three sources of administrative data:
(1) Unemployment insurance claims records from 1990 to 1994
(2) Quarterly wage records covering 1987 to 2000
(3) Community College transcript records covering 1989 to 1996.
We matched these three sets of administrative records using individuals’ social security
numbers. From the unemployment insurance claims records, we identified the quarter of
workers’ job loss. These records also include a modest set of demographic characteristics
taken from individuals’ application for unemployment insurance benefits, including birth
year, race, gender and prior education. From the wage records, we obtain information
about workers' quarterly earnings in jobs covered by the state UI system, their job tenure
at separation, and for each calendar year their primary employer's 4-digit SIC code and
county. Individuals’ earnings when they were self-employed or when they work outside
Washington State are not reflected in these records. The community college records
contain information on the credit and noncredit courses that displaced workers enrolled
in, when and where they enrolled, and the grades that they received in courses taken for
credit.
In this study, we limited this sample of UI claimants in two important ways. First,
our sample consists only of workers who had three or more years of job tenure when they
5

These data differ from those used in an earlier paper (Jacobson, LaLonde, and Sullivan, 2003.) In this
paper we have obtained an additional five years of quarterly earnings histories. Because we study only
individuals who remain attached to Washington State's wage and salary workforce throughout the sample
period, the longer sampling frame implies that we study a smaller number of displaced workers. See

were permanently laid off. We excluded “low-tenure” displaced workers, because public
policy has been most concerned about the long-term consequences of job loss by
experienced workers (Jacobson, LaLonde, and Sullivan, 1993b).
Second, our sample consists only of workers who had a history of strong
attachment to Washington State’s workforce.6 We defined attached workers to be those
who never had more than 8 consecutive quarters between 1987 and 2000 during which
they had no earnings or were not enrolled in community college courses. We use this
definition of attachment, because many workers, including those enrolling community
college courses, never have positive UI covered earnings after they lose their jobs
(Jacobson, LaLonde, and Sullivan, 2000). Because of the numbers of individuals
involved, we believe that it is unlikely that all these individuals actually had no earnings.
When we include them back into our sample, our estimate of the impacts of community
college schooling, especially from courses teaching less quantitative subject matter
becomes smaller for both older and younger displaced workers.
The sample that we use in this paper contains more than 65,000 individual
observations.7 Approximately, 10,400 of these displaced workers enrolled in and
completed at least one community college course around the time of their job loss. At
that time, about 50 percent of these workers were 35 or older. This group constitutes our
sample of older trainees. The remaining 54,900 workers in our sample lost jobs during
the same period, but never completed any community college courses. These individuals
discussion in text and in Appendix A on how we constructed our sample.
6
See Appendix A for discussion of how we limited our samples to individuals who remained attached to
the state’s workforce.
7
The number of individual observations in the sample used in this paper is smaller than the sample used in
an earlier paper on the returns to community collect schooling, which had shorter period of follow-up data.
Because we require every sample member to remain attached to the State's work force, we lose more

constitute the comparison group in our analysis below. In our study, we can follow some
trainees' earnings for more than 10 years after they leave community college. We also
have as many as eight years of pre-training earnings histories for some workers in our
sample.
C. Characteristics of the Trainees and Comparisons
In this paper, we focus on the impacts of community college schooling on
displaced workers who were 35 or older when they lost their jobs. We compare these
impacts to their counterparts who were less than 35 when they were displaced. To
estimate these impacts, we must control for differences in their underlying attributes that
also influence their decisions to enroll in community college courses, their decisions
complete either a few or many courses once enrolled, and their subsequent employment
prospects.
Table 1 reveals that displaced workers who are retrained differ in several ways
from their counterparts in the comparison group. Among both older and younger
displaced workers, community college participants are better educated, more likely to be
white, and to be displaced from the aerospace industry than displaced workers in the
comparison group. Among the older males we see that community college participants
also are more likely displaced from the state's wood products industries.
Their higher levels of educational attainment suggest that the trainees are more
skilled than the comparisons. But, as shown in at the bottom of Panel A of Table 1, we
find that the average pre-displacement earnings of both the older and younger trainees are
similar to their respective comparison groups. This surprising result does not arise

observations when we move to a sample containing longer follow-up data. See Jacobson, LaLonde, and
Sullivan, 2003.

because the trainees tend to be younger than the comparison group members. As shown
by the first row of the table, the average age of the four groups of trainees do not differ by
more than one year from their corresponding comparison group. Thus, while trainees are
better educated than other displaced workers, they are not necessary representative of the
population of displaced workers with similar levels of education. This evidence
underscores the potential importance of controlling for individuals’ skills, including the
loss of skill associated with the loss of a job. An explanation for these participation
patterns is that those with schooling beyond high school are more familiar with the
demands of and types of courses offered by community colleges. This explanation is
consistent with anecdotal evidence that workers displaced from aerospace and wood
products were encouraged to participate in retraining (Jacobson and Sullivan, 1999).
This explanation suggests that it might be differences in information about retraining
opportunities rather than differences in skills, that influence who enrolls in community
college courses.
Another factor that may influence displaced workers' participation decision is the
prevailing condition of their labor market. Individuals whose job search prospects are
poor may choose enroll in retraining because their opportunity costs are low. As shown
towards the bottom of Panel A, our results on this point are mixed. Our two measures of
local labor market conditions, the county unemployment rate and rate of employment
growth, do not reveal any differences between trainees and comparisons.
By contrast, our measure of labor market conditions in displaced workers' prior (2
digit SIC) industry is different for trainees and comparisons. Trainees appear to be
displaced from industries that have had slower employment growth than have their

counterparts in the comparison group. This difference suggests that the trainees may be
more likely to change industries as a result of their job losses and as a result expect larger
earnings losses from displacement (Jacobson, LaLonde, and Sullivan, 1993a; Neal,
1995). We explicitly account for this possibility in our empirical work below.
D. Differences Between Older and Younger Displaced Workers’ Characteristics
The background characteristics of displaced workers indicate that older displaced
workers are more skilled than their younger counterparts. These differences suggest that
their incentives to invest in new skills also may differ. By construction, the older
displaced workers have more labor market experience. As shown by the first row of the
table, the difference in ages between older and younger displaced workers is about 14
years for both males and females. This difference suggests that the older displaced
workers have about 14 fewer years remaining in their work lives. This difference should
reduce their incentives to invest in new skills even if the per-period impact of this
retraining on their earnings is that same as it is for younger workers.8
In addition to their greater labor market experience, older displaced workers are
better educated and have accumulated more tenure with their prior employers than
younger displaced workers. Comparing columns 1 and 3 for males and columns 5 and 7
for females, we see that the percentage of older trainees with at least some college
education is about 12 and 4 percentage points, respectively, more than the corresponding
percentages for younger trainees. More striking, the percentage of older trainees that had

Using the average ages given in Table 1, and a 4 percent real discount rate, this difference of 14 years
implies that, even if the per-period impact of schooling for older and younger workers is the same, the
present discounted value of these impacts over the two groups remaining work lives is about 30 percent
higher for the younger group.

accumulated 6 or more years of tenure with their prior employer is about double the
percentage for younger trainees.
This evidence on skill is consistent with the differences in older and younger
displaced workers' pre-displacement earnings. Prior to job loss, older male trainees
earned about 30 percent more and older female trainees earned about 15 percent more
than younger trainees. To the extent that the impact of training is larger for more skilled
workers, this evidence suggests that at least for our sample the average impact of training
could be higher among older workers than their younger counterparts. Therefore, we
should keep in mind that in our sample the older displaced workers could be more
efficient learners than the younger displace workers.
Evidence from displaced workers’ baseline characteristics and quarterly earnings
indicates that job loss is more costly for older displaced workers than it is for younger
displaced workers we illustrate this point using average quarterly earnings for those
individuals who were displaced in 1991. As shown at the bottom of Panel A of Table 1,
despite earning substantially more prior to displacement, after displacement, older
displaced workers earn about the same amount as their younger counterparts. This point
is seen more clearly in Figures 1 and 2. In Figure 1, we see that 10 years after
displacement, the earnings of younger displaced workers have returned to their predisplacement levels. But in Figure 2, we see that the earnings of older displaced workers
remain well below their pre-displacement levels. Because older workers experience
larger earnings losses and workers’ decisions to enroll in retraining likely depend on the
size of these losses, in our empirical work below we allow the impact of displacement to
vary by age.

E. Participation Patterns in Community College
As expected from the human capital framework, older displaced workers
participate in retraining at lower rates than their younger counterparts.9 As shown by
Panel B, about 11 percent of older male workers enroll and complete at least one
community college course around the time of their job loss, whereas the participation rate
for younger male workers is nearly 17 percent. The gap between the participation rates
of older and younger women is similar, although both groups of displaced women
participate in retraining at higher rates than displaced males.10
The differences between older and the younger displaced workers' participation
patterns are less striking when looking at the amount of training completed among those
who completed at least one community college course. As shown by the second column
of Panel B, older male displaced workers among this group completed about 27 credits.
Under the Washington State quarter system each class is worth five credits and one
academic year consists of 45 credits. Therefore, on average the older male community
college participants completed a little less than 2/3 of an academic year of schooling. The
younger males completed only 8 percent more schooling, and their female counterparts
completed about 17 percent less schooling.11 Based on the findings reported in the
voluminous literature on the returns to schooling, we expect that if displaced workers
9

This pattern of participation by age is consistent with statistics from the Current Population Survey. In the
October 2000 supplement, nearly 12 percent of males and nearly 14 percent of females between 22 and 34
years reported they were enrolled in school. Nearly all of these persons were enrolled in post-secondary
schooling. By contrast, only about 2 percent of males and nearly 4 percent of females between 35 and 54
reported they were enrolled in school (US Census, 2001).
10
In other work, we find that even after controlling for the baseline characteristics shown in Table 1,
including pre-displacement earnings, women are about 50 percent more likely than men to enroll in
community college courses around the time they lose a job (Jacobson, LaLonde, and Sullivan, 1999). This
pattern of participation in schooling, is consistent with recent trends in college attendance among teenagers
and young adults (Jacob, 2001).
11
The differences between the number of community colleges credits completed by older and younger

community college schooling has the same impact as other schooling, this level of
participation would be associated with about a 5 to 7 percent increase in earnings (Card,
1999; Heckman, Lochner, and Todd, 2003).
As indicated by the sample standard deviations in completed credits, the variation
in the number of completed credits is similar among all four groups of community
college participants. Some of our estimates below rely on this variation in the data to
identify the impact of community college schooling. But looking across Panel B, we
observe that older trainees are less likely to go beyond the first course. About one-third of
older displaced workers who participate in retraining around the time of their job loss
complete only one course. As a result, older trainees are less likely to complete 5 or more
courses (21 or more credits) than are younger trainees. For males and females the
differences are 6 and 7 percentage points, respectively. This difference is consistent with
diminishing returns in completed credits. Because older trainees face shorter remaining
work lives, the present value of completing additional classes is less than it is for younger
trainees. Below we explicitly check whether there is evidence in these data of declining
impacts of community college credits.
Some researchers argue that community college has significant payoffs only if
participants complete programs or receive degrees (Hollenbeck, 2002). If this is really the
case, then because most participants complete only a few courses, our data indicate the
community college retraining does not benefit most displaced workers. The figures in
Panel B reveal that unless the returns to completing a large number of credits are
exceptionally large, in order for community college retraining to be beneficial on
average, it must be the case that trainees' earnings improve even if they complete only a
male and female participants are statistically significant at conventional levels of statistical significance.
17

few classes. Therefore, in our empirical work below we focus on the relationship between
community college credits and earnings rather than the relationship between completion
of community college programs and earnings.
Our administrative data reveal not only how much schooling displaced workers
complete but also in the content of their courses. As shown by Panel B, somewhat more
than one-half of the credits completed by both older and younger males are in academic
and vocational courses teaching more quantitatively oriented material or in courses in
health occupations or the trades. In our discussion below, we refer to these classes as
Group 1 courses. All other community college courses we refer to as Group 2 courses.
(See Appendix A.) Among females the pattern of completed courses is different. Only
about one-third of the completed credits are in classes teaching Group 1 subject matter.
This pattern is identical for both for older and younger female participants.
A final point about participation emerges from the temporal pattern of earnings of
the 1991 cohort of displaced workers presented in Figures 1 and 2. Displaced workers
who concentrate in Group 2 subjects are less skilled than their counterparts who
concentrate their retraining in Group 1 subjects. In the pre-displacement period, Group 2
concentrators earn less than the Group 1 concentrators or the comparison group members.
By contrast during this period the earnings of Group 1 concentrators and the comparisons
are approximately the same. This evidence indicates that the enrollment decisions of
displaced workers depend on their prior skills and that this consideration is especially
important when considering separately the returns to Group 1 and to Group 2 courses.

IV. Participation in Community College Schooling by Age
As implied by the human capital framework described in Section II, our sample
reveals that older displaced workers in Washington State were less likely to enroll and
complete community college courses than were younger displaced workers who were
under 35 years when they loss their jobs. To explore further the relation between age and
participation in our sample, we decomposed the total schooling acquired by displaced
workers into three measures of participation: (A) the probability of enrolling in
community college, (B) the probability of completing at least one course given
enrollment, and (C) the number of credits completed.12 We consider separately the
relationship between age and each of these measures of participation, using a step
function for age that allows for 8 separate age intervals. We also examine these
relationships after controlling for several individual and pre-displacement job
characteristics using OLS. These characteristics are summarized in Table 1 and include
the three measures of labor market conditions and earnings during the year prior to job
loss. We report our results from this analysis in Tables 2a and 2b, showing only our
estimates for age.
As shown by the first column of Table 2a, the number of community college
credits completed by male and female displaced workers decline nearly monotonically
with age. In the second column, we see that participation-as defined as completing one
or more courses-also declines monotonically with age. These results indicate that within
our broadly defined categories of older and younger displaced workers, participation as

Heckman and Smith (1998) use a similar decomposition to examine the determinants of training
participation in programs operated under the Job Training Partnership Act.

measured by credits completed or by attending and completing at least one credit declines
with age.
The results in the last three columns of Table 2a indicate that the reason older
male displaced workers complete less training than younger males is that they are less
likely to enroll in courses in the first place. However once they enroll in a course, they
are almost as likely to complete at least one class and given that they complete one class,
except for the very youngest and oldest age groups, they on average complete nearly the
same number of credits.13
Finally, the results in Table 2b show the relationship between age and
participation after roughly holding constant foregone earnings or the opportunity cost of
retraining. Among the characteristics we control for in this analysis are an individual’s
prior tenure and prior industry, which are likely related to the expected long-term
earnings losses associated with their displacements (Jacobson, LaLonde, and Sullivan,
1993a). These variables along with schooling, prior earnings, minority status, gender, and
region of the state also are likely predictors of post-displacement earnings.
As shown by Table 2b, the relationship between age and participation is not
altered significantly if we control for these characteristics. If we have ruled out
differences in opportunity cost of retraining by age, the age-participation relationship
might then reflect (A) the shorter remaining work lives of older workers or (B) that the
impact of schooling is less for older displaced workers.

As shown by the bottom half of Table 2a, these patterns also hold for female displaced workers. But
there are some modest differences in the results. First, among enrollees, older women are somewhat less
likely to complete courses. Second, among those who complete at least one class, women 50 and over
complete one to two fewer courses (or 5 to 10 credits) than women under 50.

The results in Table 2b might indicate that holding education constant, older
displaced workers are less.
When older and younger workers have the same prior education and earnings, as
is the case in the analysis in Table 2b, it may imply that the older displaced workers
possess an unobserved attribute that makes them less productive. Otherwise why would
older workers with more labor market experience earn the same as otherwise
observationally similar younger workers? This finding suggests that in a sample of
displaced workers, age could be correlated with this undesirable attribute. To the extent
that workers with this attribute are less effective learners (i.e. a lower value of δi), we
expect age to be associated with lower propensities to enroll in training.
V. Econometric Model
To estimate the impacts of community college schooling on the quarterly earnings
of displaced workers, we use an econometric model developed in another paper
(Jacobson, Sullivan, and LaLonde, 2003). Our framework takes advantage of the long
longitudinal earnings histories that we have available to control for some of the standard
concerns raised in the schooling and training literature about unobserved heterogeneity. It
also accounts for several of the issues that we discussed above when considering the
incentives for displaced workers to enroll in community college retraining. In particular,
our model includes a rich specification of the temporal impact of displacement on
worker’s earnings (Jacobson, Sullivan, and LaLonde, 1993a). The magnitude and
temporal pattern of these impacts should relate to the opportunity cost of retraining, and
consequently influence workers’ decisions to enroll in community colleges courses.
Thus, we expect when evaluating training and schooling interventions for displaced

workers it may be important to control for the pattern of earnings losses associated with
displacement.
In the next section of this paper, we report estimates based on the parameters in
several statistical models of the following general form:
(1) yit = τit (ci,., fi , li , zi) + Xitβ + δit (si , zi ) + αi + git + γt + εit .
According to (1), workers' quarterly earnings, yit , depend (A) on the community college
schooling that they obtained, τit (ci,., fi , li , zi), which depends on the number of credits
completed, ci, the first and last quarters individuals are enrolled in school, fi and li, and
on personal characteristics, zi; (B) on observed characteristics that vary with time, which
in this paper are age, age squared and interactions of these variables with race and
gender, and the county’s unemployment rate and employment growth; (C) on the
temporal pattern of the effects of displacement δit (si , zi ), which depends on the time
elapsed between time t and the time of displacement s, on personal characteristics, and on
county unemployment rates and employment growth at the time of job loss, and on
changes in statewide employment during the pre-displacement year in workers’ (2 digit)
industry;14 (D) on unobserved individual fixed-effects and worker-specific time
trends,15 (E) on time effects, which we specify as a vector of quarterly dummy variables

The recent program evaluation literature indicates that it is important to compare trainees and
comparisons from the same or similar labor markets (Heckman, Ichimura, Smith, and Todd, 1998;
Heckman, LaLonde, and Smith, 1999; Smith and Todd, 2003). In this paper, we allow earnings to vary
according to the quarterly unemployment rate and rate of employment growth in the county of the worker's
employer. We also allow the temporal pattern of the displacement effects to vary by the unemployment rate
and employment growth during the year prior to job loss, by the statewide change in employment in the
worker's 2 digit industry prior to their job loss. We also allow the pattern to vary by whether the worker is
from the Seattle-Tacoma SMSA, from one of the state's other smaller MSAs, or from a rural county.
15
In Jacobson, LaLonde and Sullivan, 2003 we compare estimates that control only for fixed effects with
those that also account for worker-specific time trends. We found that the estimated impact of community
college schooling is smaller when we leave out worker-specific time trends. This results because our
estimates of Group 2 are significantly affected by these trends. By contrast, our estimates of the impact of
Group 1 courses do not depend much on whether we accounted for these trends. This evidence suggests to

for each quarter covered by our data; and (F) on other time varying unobserved
characteristics characterized by an independent and identically distributed disturbance.16
The schooling effect, τit (ci,. , fi , li , zi ), includes parameters that measure how
schooling affects earnings when individuals are in school and after they leave school.
When displaced workers enroll in community college courses, we expect that schooling
may cause them to forego earnings, and that these foregone earnings losses are
proportional to number of credits completed. However, we also allow for economies of
scale in classes taken. Therefore, we specify the impact of schooling on earnings during
the schooling period as follows:
(3) τit (ci, fi , li , zi ) = ψ + κ ci /( li - fi + 1), if fi ≤ t ≤ li,
where ci /( li - fi + 1) is the average number of credits completed per quarter while
enrolled in school. In the empirical work below, we also allow the parameters in (3) to
vary according to workers' age and gender.
After displaced workers leave school, we allow the impact of retraining to vary
with time since leaving school. To capture the temporal pattern of these impacts of
schooling, we consider four specifications. First, we include a vector of more than 40
parameters, one for each post-schooling quarter, which measures the amount that
community college participation affects earnings during each quarter after leaving school.
We represent this specification as follows:
(4a) τit (ci,., fi , li , zi ) = τ0t I(t - li ), if li < t.
In (4a), I(t - li) is a dummy variable equal to one when the current period is t - li periods
after the last quarter of community college. This specification corresponds to
us that the participation process into Group 1 and Group 2 courses is different.

conventional dummy variable specification widely used in evaluations of government job
training programs (Ashenfelter 1978; Heckman, LaLonde, and Smith, 1999). In our
empirical work, we allow the vector of τ1t parameters to take on different values for old
male, old female, young male, and young female displaced workers.
Second, we capture the impact of community college through the number of
completed credits. We allow for the possibility that during any given quarter the impact
of community college credits on earnings is not proportional to the number of credits
completed. In this alternative version of (4a), the impact of credits on earnings in any
given post-schooling quarter, t – li is given by τ0t + τ1tci. To implement this idea, we
include in (1) an additional vector of variables that are the interactions between the
participation dummy variable at time t and the number of completed credits:
(4b) τit (ci,., fi , li , zi ) = τ0t I(t - li ) + τ1t I(t - li )ci, , if li < t.
The difference between 4(a) and (4b) is that in 4(b) we also use variation in the number
of completed credits to identify the impact of schooling. This approach is similar to that
used in the literature on the returns to community college schooling (Hollenbeck, 1992;
Grubb, 1993; Kane and Rouse, 1995, 1999). We allow the vector of parameters τ0t and
τ1t to take on different values for the demographic groups of displaced workers.
Both a benefit and drawback of 4(a) and (4b) is that we estimate the impacts of
Community College using up to approximately 80 parameters. On the one hand, we can
report in detail how completing community college credits affects the temporal pattern of
individuals' quarterly earnings over a ten-year period. On the other hand, it is hard for
policy purposes to convey these results succinctly. In addition, as we examine the
16

Note that in our empirical work, we report robust standard errors for all of our estimates of the per-

temporal pattern of these impacts for different groups, the cell sizes become smaller and
the precision associated with our estimates diminishes.
Therefore we also seek to summarize the short- and long-term effects of
community college retraining with just a few parameters. Some experimentation led us to
the following parsimonious specification that captures the impact of community college
credits with four parameters:
(5) τit (ci,, fi , li , zi ) = τ0 + τ1 ci + τ2 [1/(t - li )] + τ3[1/(t - li )] ci, if li < t.
In (5), the impact of completing additional community college credits in is given by {τ1
+ τ3[1/(t - li )]} ci,. During the quarter after leaving school the earnings impact of
completing additional credits is given by τ1 + τ3. Because the term [1/(t - li)] gets
smaller with the passage of time, the long-term impact of completing additional credits is
given by τ1. Completing an additional course, usually worth 5 credits, raises long-term
earnings by 5τ1 per quarter.
In (5) the parameters τ0 and τ2 measure systematic earnings differences between
displaced workers who complete at least one community college credit and their
counterparts who either do not enroll or enroll but do not complete any courses. One
interpretation of these parameters is that they are the impact of “just showing-up” and
enrolling in courses. Consistent with such an effect is the idea that exposure to the
community college environment facilitates productive job search. In this case, the longterm impact of community college schooling is given by τ0 + τ1 ci. In our empirical
work, we show how sensitive our results are to restrictions on these parameters.

period impact of community college schooling.

Specifications 4(a) - (5) assume that the impact of credits on earnings is affine.
We discussed above the possibility that the impact of credits might decline with the
number of credits completed. This possibility is consistent with the distribution of
completed credits shown in Table 1B. Therefore, our fourth specification allows for a
non-linear relation between credits and earnings. To implement this idea we divide the
number of credits completed into six intervals based roughly on the number of
individuals in the sample within each interval.17 We represent the impact of each credit
category K on earnings as follows:
τitK (ci, fi , li , zi ) = τ1K I(jK< ci < j’K) + τ3K [1/(t - li )] I(jK< ci < j’K) if li < t, and
jK< ci < j’K and K = {1, 2, …,6}.
When we estimate the parameters of this semi-parametric specification, we expect that
individuals who complete, for instance, 11 to 20 credits should earn more than their
counterparts who complete only 6 to 10 credits, who in turn should earn more than those
who complete only 1 to 5 credits.
Our analysis of the data also reveals another important extension of our model. In
equation (5), we treat each credit without regard to the type of course completed. But our
community college transcript data also report the subject matter taught in each class.
Therefore, we can measure how the impacts of schooling depend on the subject matter of
the classes. Accordingly, we also extend specification (5) and define ci as a vector
denoting different types of completed credits. In the most parsimonious extension of our
model, we divided the completed courses into the Group 1 and Group 2 courses described
above.
17

The intervals are 0 – 5 credits, 6 – 10 credits, 11 – 20 credits, 21 – 40 credits, 41 – 75 credits, or more
than 75 credits. As shown by Panel B of Table 1, the number of trainees in the first category is larger than

VI. The Impact of Community College Schooling on Earnings
In this section, we present estimates from our four specifications of the effect of
community college courses. We present our results for the first two specifications in
Figures 3 and 4. The estimates that we present in Figure 3 are from the "dummy variable"
specification given in 4(a). These estimates are analogous to those commonly reported in
the literature on government job training programs. The estimates presented in Figure 4
are the earnings impacts of completing an additional credit, 1t, as defined in 4(b), for
each quarter after leaving school.
In order to summarize the results from Figures 3 and 4 with a few parameters, we
next turn to the results from our parsimonious specification given in (5). We report
estimates of the schooling parameters from this model in Table 3a for males and in Table
3b for females. The top panel of each table presents estimates for displaced workers
under 35; the bottom panel of each table presents estimates for displaced workers 35 and
older. In the odd numbered columns, we show the results for the case in which we
assume the training effects are the same during each post-schooling quarter. In the even
numbered columns, the coefficients associated with the labels “Post-College” and the
“Credits*Post-College” variables give the long-run impacts of schooling on earnings.
To investigate the sensitivity of our estimates to alternative specifications of the
schooling effect, we present, for each of the four demographic groups, results based on 6
different versions of the schooling specification given in (5). In five of these
specifications, we alternatively set one or more of the four parameters in (5) to equal
zero. In columns (1) and (2), we present results for the case in which we set the
parameters τ1 and τ3 to be zero. This specification corresponds to the dummy variable
and the number of trainees last category is smaller than the number trainees in the other credit categories.
27

specification commonly used in evaluations of government programs. We do not use the
information on the number of credits completed, but only whether a displaced worker
enrolled and completed at least one community college course. One purpose of this
exercise is to highlight whether it is important to use detailed program data on how much
retraining participants completed. To do this we first show what we would have
concluded about the effectiveness retraining had we lacked this information.
In columns (3) and (4), we present estimates when we assume that the impact of
community college schooling is proportional to the number of completed credits. In this
case, we set the parameters τ0 and τ2 equal to zero. This specification corresponds to
other studies that estimate the impact of community college schooling on the earnings of
younger students (Kane and Rouse, 1995). We present these estimates in order to
compare the results from our study to estimates from previous studies of community
college schooling.
Finally, in columns (5) and (6), we present the results for the hybrid specification
shown in equation (5). This hybrid specification is a parsimonious version of 4(b), which
produced the estimates presented in Figure 4. We find that the impact of completing
another community college course is smaller when we allow for a “just showing-up”
effect than it is when we set these parameters, τ0 and τ2, equal to zero as we do in
columns (3) and (4). Our estimates of the “just-showing up” effects, τ0 + τ2(1/t – li), in
columns (5) and (6) are never statistically significant, but as we explain below, the point
estimates in column (6) are relatively large. If we ignore the “just showing-up” effects
and simply attribute them to unaccounted for heterogeneity, then we assume that we
capture the entire long-run effect of retraining with parameter, τ1, the estimates of which

we present in column 6. This estimate is usually the most conservative estimate of the
long-run impact of community college schooling shown in Tables 3a and 3b.
After presenting results from our parsimonious specification of schooling, we turn
to reconsider whether the relation between credits and earnings are linear as required by
(5). In Table 4, we present estimates of the semi-parametric specification of credits as
described in the text above. The semi-parametric estimates presented in the four columns
of the table correspond to the model (5), whose coefficient estimates we present in
columns 6 of Table 3a and 3b.
A. The Impact of Community College Schooling on Quarterly Earnings
As shown by Figure 3, point estimates based on (4a) indicate that both older and
younger displaced workers who participated in community college schooling
subsequently experienced increased earnings in all but the first few quarters after leaving
school. The lower earnings during the immediate post-schooling period suggest a
transition period after trainees leave school. This evidence indicates that evaluations of
the short-term impacts of human capital interventions for this population may not very
valuable for policy purposes and indeed may be misleading without also considering
these interventions’ long-term effects. In the long-term, impacts for the older trainees are
around $900 per quarter, and are about twice as large as the impacts for the younger
participants.
Another important finding in Figure 3 is that the impact of community college
schooling appears to appreciate over time for both older and younger displaced workers.
They do not depreciate as is sometimes assumed in studies with limited sampling

frames.18 This result is relevant for cost-benefit analyses of training programs that
usually must extrapolate impacts beyond the evaluation’s sampling frame (Glazerman
and McConnell, 2001). The possibility of further appreciation in per-period impacts also
affects our interpretation of our own cost-benefit analyses. The long-term impacts
reported here are larger than those we report below based on the more parsimonious
specification of the relation between credits and earnings. Since our cost-benefit analysis
uses the estimates from the parsimonious specification, the results in Figure 3 suggest we
could be net-benefits understating the net-benefits of retraining in those calculations,
especially for older displaced workers.
Turning to Figure 4, we observe the same pattern for the earnings impact of
completing an additional community college credit. We observe that the effects of
community college credits start off negative, become positive within a year after leaving
school, and then rise, but at a decreasing rate over the next eight years. Again, this
evidence indicates that training effects do not necessarily depreciate over time, but may
appreciate significantly. The idea that the impacts from classroom training might
appreciate over time is consistent with the idea that those who acquire more training
subsequently acquire jobs that offer more on-the-job training (Ashenfelter and LaLonde,
1997).
B. Impacts from the Parsimonious Specification
Results from the Training Program Specification

Ashenfelter's study of the 1964 MDTA cohort was for many years one of the few evaluations of
government training programs that followed participants for more than a few years after completing
training (Ashenfelter 1975; 1978). In his study, the impacts for the male trainees (but not the females)
depreciated over time. Evaluators in their cost-benefit analysis of training programs have used the rate of
depreciation observed in Ashenfelter's study when projecting the pattern of future earnings gains in their
studies.

We next consider impact estimates from alternative versions of our parsimonious
schooling specification (5). As explained above, each of these versions correspond to a
different approach to identifying the impact of retraining. We begin with estimates based
on the conventional “dummy variable” specification used in evaluations of government
job training programs. These results are analogous to those that we report in Figure 3,
except now we report estimates separately for males and females.
The results in column (1) and column (2) of Tables 3a and 3b indicate that (A) the
long-term effects of retraining are larger than the short-term effects of retraining; (B) the
long-term effects of retraining older displaced workers are at least as large as those for
younger displaced workers; and (C) the long-term impact of retraining amounts to about
$1,500 annually for older males and to about $1,100 annually for older females. These
annual impacts compare favorably to impacts reported for government training programs
targeted at disadvantaged populations (Heckman, LaLonde, and Smith, 1999).
As shown by the bottom panel of Table 3a, the impact of schooling on older
males’ subsequent quarterly earnings amounts to only -$141.85. But, when we allow the
short-term and long-term impacts of schooling to differ, the average long-term impact of
community college schooling rises to $392. This impact amounts to about 7 percent of
post-displacement earnings. However, on average these men completed about two-thirds
of an academic year of schooling. Therefore, if we scale-up this estimate, it suggests that
in the long run, one academic year of community college schooling raises older men's
annual earnings by about 11 percent. This percentage compares favorably to

conventional estimates reported in the returns to schooling literature (Card, 1999;
Heckman, Lochner, and Todd, 2003).19
Results from Return to Community College Credits Specification
The results in columns (3) and (4) of Tables 3a and 3b focus on the impact of
completing community college credits on earnings (Kane and Rouse, 1995). Here, we
identify the impact of community college schooling by relying on variation in completed
credits among displaced workers. These impacts indicate that (A) again the long-term
effects of schooling are larger than the short-term effect; (B) the impacts generated from
this specification of participation in retraining are smaller than the impacts based on the
conventional “dummy variable” specification; and (C) the earnings gains from
completing an additional credit are about the same for older displaced workers’ as the
gains experienced by younger workers.
The results in the bottom panel of Table 3a, indicate that completing community
college courses raises older men’s quarterly earnings by $10.83 per credit. Therefore, we
expect the quarterly earnings of someone who completed two-thirds of an academic year
of schooling to be $292 greater than someone who completed no retraining. If we scale
up this figure, it implies that one academic year of such schooling raises older males'
earnings by about 8 percent.
The earnings impact associated with the community college credits specification,
although still in line with standard estimates reported in the schooling literature, is less
than the percentage we reported above for the conventional “dummy variable”
19

In the schooling literature, measures of the impact of one year of schooling usually hold constant
potential experience. In our analysis we hold constant age. Because of the relation between age and
potential experience, all other things equal, we expect our estimates of the impact of one year of
community college schooling to be somewhat smaller than the ones we would obtain if we instead held

specification. Information on the intensity of training appears to lower affect estimates of
its average impact. We reached the same conclusion for both older females and younger
males. Only for younger females is the impact estimate from the community college
credits specification larger than the impact estimate from the conventional dummy
variable specification.
Results from Hybrid Specification
Finally, in columns (5) and (6) of Tables 3a and 3b, we present estimates of
specification (5) that allows participation in community college to have its own effect on
earnings that is separate from the effect of additional completed credits. For older workers,
the estimated earnings gains from retraining are significant and comparable to those that we
obtain for their younger counterparts. For older males, completing community colleges
courses is associated with a $148.10 “just showing up" effect plus an additional $8.93
impact per completed credit. We expect, then, that on average retraining raised the earnings
of older males in our sample by $393 or by about 7 percent of post-displacement
earnings.20 This amount is nearly identical to the estimate we reported above for the
training program specification. If we scale this estimate up to a full year of schooling this
gain amounts to about 11 percent of post-displacement earnings.
Whether we believe the foregoing estimate of the impact of community college
credits for older workers is credible depends partly on how we interpret the “just-showing
up” effect. Although statistically insignificant, its magnitude of $148.10 suggests that
enrolling in a community college course is about as valuable to the trainee as staying

constant potential experience.
20
We arrive at this figure by multiplying the per credit impact of $8.93 times 27.4 credits the number of
credits completed on average by older male trainees (see Panel B, Table 1) and then adding the "just
showing up" effect of $148.01.

enrolled in community college and completing nearly 3 additional courses. Although we
believe it is plausible that enrolling in community college courses facilitates job search, and
that this benefit could raise short-term earnings, it seems less implausible that this effect
could raise individuals' long-term earnings by 3 percent.
Accordingly, we believe that our point estimates of the impact of schooling are
potentially biased upward if they include the “just showing up” effect. Despite our rich
controls for unobserved heterogeneity and the cost of worker displacement, our estimates
may be picking up unobserved differences between the trainees and the comparisons. This
conjecture is consistent with our observation above that age was a very important
determinant of the enrollment process. In other work, we found more generally that many
baseline characteristics appeared to be significant determinants of participation (Jacobson,
LaLonde, and Sullivan, 1999). By contrast, age as well as other observed characteristics are
less important predictors of how many credits displaced workers completed once they
enrolled. This evidence suggests to us that if a large set of observed characteristics are not
that important when predicting how many credits a trainee completes, then it is plausible
that other unobserved characteristics may not be that important either. Therefore, we
consider our estimates of τ1ci, the gains from completing additional credit, in column (6),
rather than τ0 + τ1ci, to be a more plausible estimate of the long-term impact of community
college schooling.
Given the foregoing interpretation of our results, we conclude that community
college retraining raised older male trainees' long-term quarterly earnings by an average of
$241. Therefore, we expect that one academic year of such retraining would raise their
quarterly earnings by about $400 or 7 percent of post-displacement earnings. This impact

for older displaced workers is in line with conventional estimates of the impact of formal
schooling on earnings, but is about 30 percent less than the impact we would have reported
had we relied on estimates from the conventional dummy variable specification (shown in
column 2). It also is less than the impacts from the community college credit specification
(shown in column 4).
The impacts of community college schooling for two of the other three
demographic groups are similar. For the third group, young females, completing an
additional credit is associated with about a $12 increase in quarterly earnings. This impact
amounts to about 14 percent of post-displacement earnings. But as indicated by the robust
standard errors, this impact is not statistically significantly different from the $9.56 impact
per credit that we report for older females. Therefore, at least among displaced workers
who participate in retraining, old dogs acquire new skills as effectively as their younger
counterparts.
C. The Impact of Community College Schooling While In School
Besides reporting the impact of community college credits on earnings after leaving
school we also report in Tables 3a and 3b their impacts on earnings while trainees are
enrolled in school. The negative coefficients associated with the “In-College*Credits per
Quarter” variables in all 6 specifications and for all four demographic groups indicate that
being enrolled in school is associated with significantly lower earnings. One interpretation
of this finding is that trainees forego earnings by delaying their return to regular full-time
work.
In column (6) of Table 3a, we see that for older males enrolling and completing
community college courses during a given quarter is associated with a reduction in earnings

of $275.10 for every credit completed. Therefore, we expect that an older male who
enrolled in community college and completed 1 course for 5 credits to have on average
$1,376 lower earnings during the quarter. Therefore, these estimates suggest that the
average opportunity cost of completing one academic year of credits completed over three
calendar quarters equals about $12,350.21 Looking across the remaining columns of the
table we see that these estimates of the opportunity cost of schooling do not depend on
which specification we use.
If the foregoing figures measure the opportunity costs of schooling, then because
their labor market earnings are the highest, we expect older males to experience the largest
earnings losses associated with participating in retraining, and the younger females to
experience the smallest losses. Consistent with this reasoning, our estimates in Tables 3a
and 3b imply that acquiring one academic year of community college schooling reduces the
earnings of older males while in school by about 22 percent more than it does for younger
males, and by about 33 percent more than it does for similarly aged women. We find that
the youngest women incur the smallest opportunity costs associated with enrolling in
schooling. Their costs are about 44 percent less than those of older men.
According to this interpretation of the "In-College" estimates, the opportunity costs
of retraining are somewhat larger than the direct costs of retraining for older males and
comparable to the direct costs of retraining for the other demographic groups. Kane and
Rouse report that the cost of providing a student with an academic year of community
college schooling is about $8,000, of which individuals pay about one-fifth of this amount
through their tuition and fees (Kane and Rouse, 1999). If we interpret our "In-College"
21

We arrive at this figure as follows: We assume each quarter the trainees completes 15 credits. This
amounts to a loss of –$4,114 per quarter (or 12.60 plus –275.10 times 15 credits), or $12,342 over three

estimates as the opportunity cost of schooling, the private cost of one academic year of
community college schooling for an older male is $14,150. The social costs of this
schooling is $20,350 and higher still if we also account for the welfare cost of the taxes
raised to subsidized community college schooling (Browning, 1987; Heckman, LaLonde,
and Smith, 1999).
Our “In-College” estimates may not measure the opportunity cost of schooling.
Instead, they could reflect individuals’ unsuccessful job search. As a result, they simply tell
us that those who did not find jobs right away enrolled in community college courses and
the least successful job searchers among the trainees complete the most classes. Under this
interpretation, our “In-College” estimates overstate the opportunity cost of completing
retraining and any net-benefit or internal rate of return calculations based on them are
biased downward. Consequently, later in this paper, when we estimate the internal rates of
return to retraining, we consider three cases: (A) we treat our "In-College" estimates as
estimates of the opportunity costs of retraining; (B) we estimate the opportunity costs of
schooling to be equal to one-half of the cost estimates in (A); and (C) we assume the
opportunity cost of retraining is zero.
D. Evidence of Non-linear Effects of Community College Credits
Next, we examine how sensitive our results are to our linear specification of the
relation between credits and earnings. In Table 4 we present estimates of the long-run
impact of the indicated amount of schooling, τ1K, from the semi-parametric specification of
credits described above in (6). Our results indicate that (A) displaced workers can benefit
from completing just a few community college courses; (B) completing many classes
usually produces greater per period impacts than completing just a few classes; and (C)
quarters. We use this figures with computing the net benefit and internal rates of return later in the papper.
37

overall, there is not much evidence here of diminishing returns to completing courses. Our
earlier assumption that earnings gains rise linearly with credits appears to be a reasonable,
though imprecise, approximation of the relation between credits and earnings (c.f. Kane
and Rouse, 1995). As shown by column (2) of Table 4, beyond two classes, older males’
earnings rise monotonically with the number of credits completed. If we compare the
difference between the earnings of trainees who completed 21 to 40 credits and their
counterparts who completed 6 to 10 credits, we observe more than a $500 difference in
quarterly earnings. Given that on average the difference between these two groups’
completed schooling is about one-half a year, this impact is quite substantial in terms of
percentages. But, the relation between credits and earnings is not as uniformly monotonic
during the first year and one-half of schooling for other the four demographic groups. For
example, as shown by column 4, older women who completed 21 to 40 credits earn about
than $450 less per quarter than their counterparts who completed only 11 to 20 credits.
The results in Table 4 reveal that trainees who completed the most schooling
experienced the largest earnings gains. As shown by the last row of the table, older males
who completed about 2 years of schooling experienced a nearly $1,000 gain in their
quarterly earnings. This gain indicates that two years of training raises older male earnings
by about 17 percent. Among the other three demographic groups, the estimated earnings
gains are larger. This evidence again suggests that completing two years of community
college retraining can reduce much of the long-term earnings losses associated with
displacement (Ruhm, 1991; Jacobson, LaLonde, and Sullivan 1993; Farber 1993, 2003).
The impacts for trainees who completed approximately two years of community
college schooling also provide evidence that our assumption in (5) requiring an affine

relation between credits and earnings is reasonable. These impacts for are roughly
consistent with what we would have predicted from the results reported in columns (4) and
(6) of Tables 3a and 3b. The impact of $952 reported in Table 4 for older males works out
to a gain of $10.58 per credit. This amount is smaller than the $10.83 figure we reported in
column (4) of Table 3a, but it is larger than the estimated impact of $8.94 reported in
column (6). A similar calculation for the other three demographic groups leads to a similar
conclusion.22 Therefore, we do not find any strong evidence of diminishing returns to
completed credits.
E. Impacts By Content of Courses
In this subsection, we use information on course content to show that the foregoing
estimated impacts mask differences in impacts by type of course. But, within our broadly
defined categories of community college credits, the per-period earnings impacts of older
and younger trainees continue to be similar, especially for male trainees.
To examine these differences, we extended specification (5) in earlier work to
account for the nine categories of credits listed in Appendix Table A (Jacobson, LaLonde,
and Sullivan, 1997). After reviewing the results, we found it helpful for expositional
purposes to aggregate these categories into two groupings. As described above in Section
III, the first grouping (Group 1 courses) consists of academic courses in the sciences and
mathematics as well as courses teaching more technically oriented vocational subject
matter, including courses in the health occupations. The second grouping (Group 2
courses) consists of all other community college courses.

For younger males, this calculation implies a $11.35 per credit effect compared with the $9.09 effect
reported in Table 3a; for older females a $11.87 per credit effect compared with the $9.56 effect reported in
Table 3b; and for younger females. For younger females this estimate implies an effect of $11.88 per credit,
which in contrast to the other three groups is about the same as the $12.13 estimate reported in Table 3b.

As shown by Table 5, we find consistent differences between the long-term impacts
of Group 1 courses and Group 2 courses. But within each of these broadly defined
categories of courses, the impact of schooling is similar for older and younger displaced
workers. Completing a Group 1 credit increased the long-term quarterly earnings of both
older and younger males by approximately $12. Among females these courses were
associated with larger earnings increases, especially for younger women.
The estimates reported in Table 5 imply that completing one academic year of
Group 1 courses raises both older and younger males’ long-term quarterly earnings by
about $550. This increase amounts to about 10 percent of older males’ post-displacement
earnings and to about 12 percent of younger male's post-displacement earnings. For older
females, our estimates imply that completing one academic year of Group 1 courses raises
quarterly earnings by $830 or by about 21 percent of post-displacement earnings. For
younger females, our point estimates imply even larger earnings impacts. For women, these
earnings gains suggest that completing just one year of Group 1 schooling can eliminate
much of the permanent earnings losses reported associated with displacement (Ruhm,
1991; Jacobson, LaLonde, and Sullivan, 1993a; Farber 1993, 2003).
By contrast to the foregoing results for Group 1 courses, we find that completing all
other community college courses has a much smaller long-term impact on earnings. As
shown by the second of the last row of Table 5, the long-term impacts of Group 2 courses
are about $4 to $5 per credit for all four groups of displaced workers. Group 2 courses have
about 1/5 to 1/2 the impact of the Group 1 courses.23 The figures amount to only 3 to 5

We also have found that this relation between Group 1 and Group 2 courses holds among displaced
workers who tend to take a majority of their courses in Group 2 subjects. Therefore, we do not interpret our
Group1 and Group 2 findings as having arisen because different types of workers concentrate in these
different subject areas. We note here that our econometric specification includes both fixed effects and

percent of post-displacement earnings. Moreover, these impacts are smaller than those
usually reported in studies of schooling.
F. Backcasting and Interpreting Evidence of Specification Error
Our specification of the impact of community college schooling is over identified.
In particular, we have assumed that completed credits do not predict earnings prior to
enrolling in school. We can explicitly test this assumption, because we observe earnings for
many quarters prior to the start of training. If our econometric specification is correct
retraining should not appear to affect earnings prior to completing community college
credits.
We examine whether training affects earnings prior to completing community
college schooling during two specific pre-training periods. The first period is after
displacement, but before entering school. The second period is the year prior to
displacement. We included indicator variables in (5) that were equal to one when the
quarter was in one of these periods and the individual subsequently completed community
college credits (the "Pre-College" variable in Table 6). In addition, we included the
interaction between these indicator variables and how many credits the individual
completed in Group 1 and Group 2 subject areas (the interactions with the "Pre-College"
variable in Table 6). Together these variables show the relation between community
college attendance and completion of credits on earnings during each of these preschooling periods. Each column in the Table 6 corresponds to a different specification of
the cost of displacement in (5).

worker-specific time trends. So we have accounted for unobserved heterogeneity that corresponds to these
controls.

We begin our analysis of our backcasting test with column 5 of Table 6. These
figures are based on the specification that we used to produce the results that we reported
in Table 5. They indicate that participating in community college schooling predicts
earnings prior to enrolling in school. The first set of results reveals that during the year
prior to displacement individuals who completed more Group 1 and Group 2 credits
tended to have earnings that were above their own expected (trend) levels. The estimated
coefficients associated with the Group 1 and Group 2 credit interactions are similar for all
four demographic groups. This result indicates that individuals whose pre-displacement
earnings were above expected levels subsequently completed more of both types of
community college courses. This finding on earnings before job loss suggests that our
results in Tables 3a and 3b as well as Table 5 might overstate the impacts of retraining.
By contrast, the next set of figures in Table 6 suggests that the bias in our results
runs in the opposite direction. For all four demographic groups, these estimates indicate
that during the post-displacement pre-training period individuals who subsequently
completed a lot of retraining had earnings that were below their expected levels. This
second set of results suggests that individuals who did worse than expected after losing a
job participated in more retraining. If our specification of the cost of displacement fails to
capture this variation among displaced workers, then our estimates of the impact of
community college schooling are likely too low.
The findings we report in Table 6 may signal that the impact estimates we report
in Tables 3a, 3b, and 5 are biased. Despite our rich econometric specification there may
be important sources of selection that we have failed to take into account. The two sets of
results in Table 6 suggest one explanation that we explore here. Participation is especially

high among displaced workers whose earnings were above expected levels prior to their
job losses and were below expected levels just after their job losses. Therefore, workers
who experienced particularly large unexplained drops in earnings between the pre- and
post-displacement periods tend to enroll and complete more community college
schooling. If these drops in earnings reflect the permanent cost of job loss, then the
estimated impacts of community college credits reported in Tables 3a, 3b, and 5 are
likely too small. If we could do a better a job controlling for these earnings drops, we
might be able to reduce the magnitude of the estimates reported in Table 6 and be more
confident in our non-experimental estimates.
To explore the merits of foregoing contention, we examine how sensitive our
backcasting test is to the way we specify the cost of displacement. The columns in Table
6 correspond to different controls for the temporal pattern of the impact of displacement
on workers' earnings. In the first column of the table, we exclude all controls for the
effect of displacement on earnings. In the second column, we introduce the vector of
dummy variables that account for the average temporal pattern of displacement. (See
Appendix B.) In column 3, we allow this pattern to vary according to workers' gender,
minority status, age, prior schooling, prior tenure, and region of the state. In column 4 we
allow this pattern also to vary by a worker's prior industry. In Table 1 we observed that
prior industry was related to the likelihood that a displaced worker participated in
retraining, and previous studies have shown that the costs of displacement vary across
industries (Jacobson, LaLonde, and Sullivan, 1993a,b). Finally, in the last column, we
allow this pattern to vary according to labor market conditions at the time of an
individual's job loss.

Contrary to our expectation, we find that our backcasting results change little as
we refine our specification of the cost of displacement. The results in the second column
indicate that controlling for the average temporal pattern of the effect of displacement
may matter a good deal. The coefficients, especially the variable indicating whether a
displaced worker participated in community college retraining, declines in magnitude.
But as we move across the columns of the table, we observe that additional controls for
how the temporal pattern of the displacement effect varies among individuals has less of
an impact on the results of our backcasting test. During the year prior to displacement,
the results for credits change hardly at all. During the period after displacement but
before school, the magnitude of the schooling credit coefficients become smaller, but
these estimates still are statistically significant at conventional levels.
The results of the foregoing exercise suggest that our estimates of the impact of
community college schooling probably depend on our controls for the average temporal
pattern of the cost of displacement. However our impact estimates do not depend on
further refinements of this specification. To check this conjecture, we reestimated (5)
using the different specifications of the displacement effect described in Table 6. We
present these alternative impact estimates in Table 7.
As shown by column 1 of Table 7, if we include no controls for the displacement
effect, the estimated impacts of both Group 1 and Group 2 courses are smaller than we
report in Table 5 (and repeat here in column 5). When we control for the average
temporal pattern of the effect of displacement, our estimates are close to those that we
reported above in Table 5.24 Allowing this pattern to vary according to other
24

The point estimates of the "just showing up" effects vary substantially across the columns Table 7, but
as indicated by the standard errors, these effects are imprecisely measured.

characteristics and to labor market conditions does not substantively alter our estimates.25
Since our specification of the displacement effect in column 5 includes more than 130
parameters, it does not seem promising to explore whether further controls would
improve the results of our backcasting test or alter our results on the impact of
community college schooling.
VII. Discussion and Conclusion
A. Can We Teach Old Dogs New Tricks?
In this paper, we use administrative data to examine how community college
schooling affects the short and long-term earnings prospects of older displaced workers.
The question that motivated our research is whether older workers gain as much from this
important source of retraining as younger workers. Our analysis indicates that older
workers, those 35 or over when they lost their jobs, experienced similar per-period
impacts from community college retraining as younger displaced workers. For males, we
find that the per-period impacts for older and younger workers are nearly identical. Even
when we consider more quantitatively oriented (Group 1) subject matter separately from
other community college courses (Group 2), we still find that both older and younger
male trainees experienced nearly identical earnings gains from similar types of retraining.
Our findings for females are similar to those for males. We find weak evidence of
larger per-period impacts for women, especially for younger women. Our point estimates
of the per-period impact of the Group 1 courses are up to two times larger than the
estimates we report for males. However, the standard errors associated with these
estimates indicate that we can not reject the hypotheses that (A) older and younger
25

We also explored including both leads and lags of our three labor market condition variables in our
specification of the displacement effect. Our results we unaffected by this addition to our model.

displaced females experience similar per-period impacts from broader similar types of
retraining; and (B) that the per-period impacts for the female trainees are the same as
those of the male trainees.
Overall, we find that community college schooling raised older displaced workers'
earnings by about $9 per credit. In the short-term, we find that the impact of such
retraining was negative during the first year after leaving school. But this impact grew
over time and showed no sign of deteriorating after 10 years. Indeed, this impact appears
to be getting larger with time since leaving school, and we have reason to believe that our
estimate of $9 per credit may understate the long-term effects of retraining.
Our per-period impact estimates imply that one academic year of community
college retraining raises older males' earnings by about 8 percent and older females'
earnings by about 10 percent. Our point estimates also suggest that if displaced workers
concentrate entirely on Group 1 subjects, these percentages are about 33 percent higher
for older males and about 100 percent higher for older females. These impact estimates
are consistent with the earnings gains we expect from formal schooling acquired by
younger persons. As a result, we conclude "you can teach (at least some) old dogs new
tricks."
In section II, we argued that is reasonable to expect the per-period impacts for older
trainees to be larger than for younger trainees, even if on average the population of older
displaced workers consists of less efficient learners than does the population of younger
displaced workers. Instead, we find that the gains for older and younger trainees are
comparable. Given that older workers have shorter remaining work lives and almost
certainly higher opportunity costs, this finding suggests that the distribution of impacts

among the displaced worker population not only may have a different mean, but also a
different variance for older workers. Estimating the shape of these impact distributions
requires stronger assumptions that we have imposed here (Aakvik, Heckman and Vytlacil
2003). But knowing the shapes of these distributions is important for policy analysis. We
leave this question for future research.
We do not have the same confidence in our findings that we would have if our estimates
were generated from a social experiment. As we discussed above, our specification failed
the backcasting test, and this failure may signal that our impact estimates are biased.
However we have found no reason to believe that our estimates are more likely to
overstate than they are to understate the impact of community college retraining. Our
non-experimental estimates are based on richer a specification of unobserved
heterogeneity than is used in nearly all other studies of the returns to schooling, and on a
rich specification of the way displacement affects individuals' short and long-term
earnings prospects. We find that further controls for these displacement effects are
unlikely to alter our results. If biases do remain in our impact estimates, more work is
needed to model what must be a complicated selection process into retraining that likely
also requires even richer data.26
Finally, even if we were completely confident that our non-experimental impact estimates
replicated the results that we would have reported using a social experiment, our
conclusions apply only to the group of displaced workers in our sample who chose to
participate in retraining. The question of the external validity of our results remains.
More research is required on whether our results represent the experiences of displaced
26

Recent findings from research on this problem in other training settings underscore the difficulty of
this task. (Heckman, Ichimura, Smith, and Todd, 1998; Smith and Todd, 2003).

workers in other states and at other points in time. In addition, our results do not
necessarily predict the impact of retraining for Washington State displaced workers who
chose not to participate or for ones who might have been induced to participate in
retraining because of expanded public subsidies.

B. Should We Teach Old Dogs New Tricks?
Even without the foregoing caveats and qualifications, our results do not necessarily
imply that society should subsidize or even encourage the retraining of older displaced
workers. Although older and younger workers experience similar per-period impacts
from retraining, the net-benefits and rates of return (IRR) from these investments are
likely different.
To examine this issue more closely, we use the information from column 6 of Table 3a
and 3b to compute the private and social net-benefits and the IRR from investments in
community college retraining. We assume that displaced workers complete one academic
year of the same mix of Group 1 and Group 2 courses as the individuals in our
Washington State sample. We also assume that individuals expect to pay one-fourth of
their increased earnings in taxes and that the welfare cost of the taxes raised to subsidize
community college schooling amounts to $3,250 per academic year of schooling. This
amount assumes that the deadweight loss associated with raising $1 in taxes is $0.50
(Browning 1987; Heckman, LaLonde, and Smith, 1999).
In Panel A of Table 8, we present the net-benefit of retraining from the perspective of the
participant and of society. Here, we assume that the opportunity cost of retraining is equal
to one-half the cost implied by the "In-College" effects. In Panel B, we present alternative

IRR calculations from the perspective of society. (The private IRR calculations are in a
footnote, below.) We examine how sensitive our calculations are to alternative
interpretations of the “just showing up” and the “In-College” effects.
As shown by Panel A of Table 8, our calculations indicate that our sample of displaced
workers likely experienced substantial net-benefits from their investments in community
college schooling.27 But the (private) net-benefits of retraining are markedly larger for
younger displaced workers than for older displaced workers. The benefit to cost ratios
indicate that for every dollar that younger displaced workers invested in their retraining,
they got back (in present value terms) between $3.07 and $5.40.28 By contrast, the
corresponding ratios for older displaced workers are smaller ranging from $1.69 to $3.05.
For both groups of displaced workers, retraining seems likely to have been a sound
investment. But, the differences between older and younger trainees’ benefit to cost ratios
provide a reason for why we find substantially lower participation rates in retraining by
older displaced workers.
The results of our cost-benefit analysis of community college retraining are less
impressive from the perspective of society. The difference between results for the two
perspectives occurs, because community college schooling is heavily subsidized by
taxpayers, and because of the welfare cost of taxation that we incorporate in our

As noted above in the text, we have standardized these calculations to one academic year of schooling.
As shown in Panel B of Table 1, the trainees in our sample acquired a little less than two-thirds of a year of
schooling. Recall in section VI.D we found no evidence of diminishing impacts of community college
credits for any of the four demographic groups. Thus, the average net benefit of retaining for our sample of
displaced workers is approximately one-third less than the figures in Table 8.
28
The private IRR are larger for younger than for older displaced workers. Assuming the opportunity cost
of retraining equals one-half the amount implied by the estimated “in-College” effects, we estimate that the
private IRR for younger trainees ranges from 13.1 percent for younger men to 21.2 percent for younger
women. For older trainees our private IRR estimates range from 11.4 percent for older men to 15.7 percent
for older women. If we alternatively assume that our "In-College" estimates reflect the opportunity cost of
retraining, then our estimates range from 5.4 percent for older males to 9.4 percent for older females.

calculations. Our more conservative calculations suggest that society approximately
broke even when an older displaced male worker was retrained. For older females, our
results indicate that society only received a modest net benefit: for every dollar invested,
it got back about $1.27 (in present value terms).29 By contrast, the benefit to cost ratios
are larger for younger displaced workers, especially for younger females.
We also examine the benefits of retraining by considering alternative IRR calculations.
The figures in Panel B help to underscore the policy importance of alternative
interpretations of the "just showing up" and the "In-College" effects. As shown in column
4, if we assume that the opportunity cost of schooling is zero and we include the "just
schooling up" effect as part of the impact of retraining, we find that the implied (social)
IRR from retraining is impressive for all four demographic groups. By contrast, if we
assume, as we do in column 3, that the "In-College" effects measure the opportunity cost
of retraining and we assume the "just showing up" effects are not part of the impacts of
retraining, then our IRR estimates are relatively low for all groups, expect younger
females. Under these assumptions, our estimates imply that the social IRR from
completing one academic year of schooling is only 1.4 percent for older males. However,
as shown by Panel B, our IRR estimates are very sensitive to our estimate of the
opportunity cost of retraining. More research is needed on whether displaced workers
really forgo job opportunities when they participate in community college schooling.
Despite our cautious interpretation of the IRR figures in Panel B, the 1.4 percent IRR
estimate for older males underscores the importance of accounting for labor market
conditions before encouraging older workers to seek retraining. Even for our sample of
29

The benefit to cost ratios that include the “just showing up effect” are somewhat larger: 1.34 for older males

and 1.49 for older females.

older displaced workers, all who voluntarily chose to participate in retraining, it does not
appear to have made sense for them to delay their return to work to acquire one academic
year of community college schooling. If this decision caused them to forego what our
"In-College" effects suggest are about one-half of a year of earnings, then we estimate
that the social IRR from their investment was likely very low. By contrast, if they
attended school while trying to generate a job offer, we estimate that the social IRR from
their investment may have been quite substantial, possibly as high as 11 percent.
Finally, we observe that our conclusions about the returns to retraining also are
sensitive to the type of courses completed by displaced workers. So far, we have based
our net-benefit and IRR calculations on the assumption that displaced workers complete
the same mix of Group 1 and Group 2 courses observed in our sample. Above, we report
that the types of courses that we classify as Group 1 courses had per-period impacts that
were two to five times larger than the per-period impacts of the Group 2 courses. For
older male workers, this difference in per-period impacts implies that the social IRR from
one academic year of Group 1 courses equals about 8 percent. This figure compares
favorably with conventional estimates of the internal rates of return to schooling.30
By contrast, the IRR from a similar investment in Group 2 courses has a negative
IRR. This finding is important for retraining policy. We observed in Table 1 that about
one-half of the credits completed by male displaced workers and nearly two-thirds of the
credits completed by female displaced workers were in courses teaching Group 2 subject
30

The 8.1 percent figure assumes the "just showing up" effect is not part of the per-period impact of
community college schooling. When we include it in our calculation for older males, the IRR of Group 1
courses rises to 10.3 percent. We computed these percentages under the assumption that the opportunity
cost of retraining equaled one-half the cost implied by the "In-College" effects. Our social IRR figures for
Group 1 courses are comparable to those reported for individuals in the population who complete between
12 and 14 years of schooling. See Heckman, Lochner, and Todd (2003), Table 4. Their calculations also
include consideration of tuition and tax payments.

matter. This raises the question of whether community colleges should steer older
displaced workers toward Group 1 subject areas. Similarly, would programs that operate
under the Workforce Investment Act or Trade Adjustment Assistance Act, which rely on
a lot of community college retaining, be more productive if participants were steered
away from Group 2 courses and toward Group 1 courses?
Although our findings about the benefits of Group 2 courses suggest that program
operators should steer displaced workers away from these subject areas, we are cautious
about making this inference from our study. In our study older displaced workers were
very unlikely to complete one academic year of especially Group 1 schooling. This fact
suggests to us that incentives to concentrate on Group 1 retraining were not apparent to
most trainees, the opportunity costs were indeed substantial, or possibly many displaced
workers were not prepared to successfully complete such retraining. In any case, it is
unclear whether our IRR estimates from Group 1 courses would have been as large as
indicated here, if more displaced workers had trained intensively in these subject areas.
Further research should explore whether programs designed to steer displaced workers,
especially older displaced workers, toward Group 1 type courses improve the
performance of government workforce development initiatives.

References
Aakvik, A. J. Heckman, and E. Vytlacil (2003). “Estimating Treatment Effects For Discrete Outcomes
When Responses To Treatment Vary: An Application to Norwegian Vocational Rehabilitation Programs,”
forthcoming, Journal of Econometrics.
Ashenfelter, Orley (1975). "The Effect of Manpower Training on Earnings: Preliminary Results," in
Proceedings of the Twenty-seventh Annual Winter Meeting of the Industrial Relations Research
Association 1975: 252-260.
Ashenfelter, Orley (1978). "Estimating the Effect of Training Programs on Earnings," Review of
Economics and Statistics, 60, 47-57.
Ashenfelter, Orley, and Card, David (1985). ``Using the Longitudinal Structure of Earnings to Estimate the
Effect of Training Programs.'' Review of Economics and Statistics 67 (November): 648-60.
Ashenfelter, Orley and Robert LaLonde (1997). “The Economics of Training,” Handbook of Human
Resources, D. Lewin, D. Mitchell, and M. Zaidi, eds., JAI Press.
Angrist, Joshua, and Newey, Whitney (1991). ``Over-Identification Tests in Earnings Functions with Fixed
Effects,'' Journal of Business and Economic Statistics, 9(3): 317-323 (July).
Barnett, W. Steven (1993). "The Long-term Effect of Early Childhood Programs on Cognitive and School
Outcomes," The Future of Children. Volume 5(3).
Barnett, W. Steven and Leonard N. Masse (199 ). "A Cost-Benefit Analysis of the Abecedarian Early
Childhood Intervention," National Institute for Education Research.
Browning, E., (1987). “On the Marginal Welfare Cost of Taxation,” American Economic Review, 77(1):
11-23.
Card, David, (1999). "The Causal Effect of Education on Earnings," In Handbook of Labor Economics,
Vol. 3A, edited by Orley Ashenfelter and David Card. (Amsterdam: North Holland), pp 1801-1863.
Carneiro, Pedro Manuel and James Heckman (2003). "Human Capital Policy," National Bureau of
Economic Research Working Paper, No. 9495.
Corson, Walter; Decker, Paul; Gleason; and Nicholson, Walter (1993). International Trade and Worker
Dislocation: Evaluation of the Trade Adjustment Assistance Program, Final Report DOL contract No: 999-0805-75-071-01, Princeton: Mathematica Policy Research Inc.
Currie, Janet (2001). "Early Childhood Education Programs," Journal of Economic Perspectives, 15(2):
213-238.
Currie, Janet and Duncan Thomas (1995). "Does Head Start Make a Difference?" The American Economic
Review, 85(3): 341-364.
Currie, Janet, Eliana Garces, and Duncan Thomas (2002). "Longer Term Effects of Head Start," The
American Economic Review, 92(4): 999-1012.
Farber, Henry (1993). “The Incidence and Costs of Job Loss: 1982 - 1991,” Brookings Papers on Economic
Activity: Microeconomics, 1993(1): 73-119.
Farber, Henry (2003). “Job Loss in the United States, 1981-2001” NBER Working Paper 9707 (May).

Freeman, Richard B. (1974). ``Occupational Training In Proprietary Schools and Technical Institutes,''
Review of Economics and Statistics, 63:310-318.
Grubb, Norton (1993a). ``The Long-Run Effects of Proprietary Schools on Wages and Earnings:
Implications For Federal Policy,'' Education Evaluation and Policy Analysis, 15(1): 17 - 33 (Spring).
Grubb, Norton (1993b). ``The Varied Economic Returns to Post-secondary Education: New Evidence From
the Class of 1972,'' Journal of Human Resources, 28(2): 365-82.
Heckman, James (2001) Policies to Foster Human Capital," Research in Economics, 54(1): 3-56.
Heckman, James, and Robb, Richard (1985). ``Alternative Methods for Evaluating the Impact of
Interventions.'' In Longitudinal Analysis of the Labor Market Data, ed. J.J. Heckman and B. Singer.
Cambridge: Cambridge University Press.
Heckman James J., and Hotz, V. Joseph (1989). ``Choosing Among Alternative Nonexperimental Methods
for Estimating the Impact of Social Programs: The Case of Manpower Training.'' Journal of the American
Statistical Association 84 (December): 862-874.
Heckman James J., H Ichimura, Jeffrey Smith, and Petra Todd (1998). "Characterizing Selection Bias
Using Experimental Data," Econometrica, 66(5): 1017-1098.
Heckman, James J., Robert LaLonde, and Jeffrey Smith (1999). "The Economics and Econometrics of
Active Labor Market Programs," In Handbook of Labor Economics, Vol 3A, edited by Orley Ashenfelter
and David Card. (Amsterdam: North Holland), pp 1865-2097.
Heckman, James J., Lance Lochner, and Petra Todd, (2003). "Fifty Years of Mincer Earnings Regressions,"
Unpublished working paper, University of Chicago, March 28, 2003.
Hollenbeck, Kevin, (1992). “Post-secondary Education as Triage: Returns to Academic and Technical
Programs, W.E. Upjohn Institute For Employment Research, Staff Working Paper 92-10.
Hollenbeck, Kevin, (2002). “Comments on Chapter 7,” in Targeting Employment Services, R. Eberts, C.
O’Leary, and S. Wandner, eds., Kalamazoo, MI: W.E. Upjohn Institute for Employment Research.
Jacob, Brian (2001). "Where the Boys Aren't: Non-Cognitive Skills, Returns to School and the Gender Gap
In Higher Education," National Bureau of Economic Research Working Paper, No. 8964.
Jacobson, Louis, and Daniel Sullivan (1999). “ Evaluation of Timber Retraining Benefits,” report
conducted for Conducted for the Washington State Joint Legislative Audit and Review Committee, by
Westat, Inc.
Jacobson, Louis S.; LaLonde, Robert J.; and Sullivan, Daniel G. (1993a). ``Earnings Losses of Displaced
Workers,'' American Economic Review, 83(4):685-709.
---- (1993b). The Costs of Worker Dislocation, Kalamazoo, Michigan: W.E. Upjohn Institute for
Employment Research.
---- (1997). ``The Returns from Community College Schooling for Displaced Workers,” Federal Reserve
Bank of Chicago Working Paper, WP-97-16 (December).
Jacobson, Louis, Robert LaLonde, and Daniel Sullivan (2000). “Participation in Community College
Schooling and it Effects on Displaced Workers Earnings and Employment Prospects: A Study of Displaced
Workers From Washington State,” Final report to U.S. Department of Labor, Employment and Training
Administration, Westat, Inc., contract number K-6307-7-00-80-30.

Jacobson, Louis S.; LaLonde, Robert J.; and Sullivan, Daniel G. (2003). ``The Returns from Community
College Schooling for Displaced Workers,” Journal of Econometrics, forthcoming
Kane, Thomas; and Rouse, Cecila (1993). ``Labor Market Returns to Two- and Four- year College,''
American Economic Review, 85(3): 600-614.
Kane, Thomas; and Rouse, Cecila (1999). "The Community College: Educating Students at the Margin
Between College and Work," Journal of Economic Perspectives, 13(1,Winter): 63-84.
Karoly, Lynn, P. Greenwood, S. Everingham, J. Hube, R. Kilborn, C. Rydell, M. Sanders, and J. Cheesa.
(1998). "Investing in Our Children: What We Know and Do Not Know About the Costs and Benefits of
Early Childhood Intervention Programs." The Rand Institution, ED 419 621.
Karoly, Lynn, et. al. (2001). "Assessing Costs and Benefits of Early Childhood Intervention Programs:
Overview and Application to the Starting Early Starting Smart Program. Santa Monica: The Rand
Institution.
Leigh, Duane E. (1990). Does Training Work for Displaced Workers: A Survey of Existing Evidence,
Kalamazoo, Michigan: W.E. Upjohn Institute for Employment Research.
Leigh, Duane E. and Andrew M. Gill (1997). “Labor Market Returns to Community Colleges,” The Journal
of Human Resources, 32(2): 334–353 (Spring).
LaLonde, Robert (2003) “Employment and Training Programs,” forthcoming in M. Feldstien and R.
Moffitt eds. Means Tested Transfer Programs in the U.S., Chicago: Unversity of Chicago Press for the
National Bureau of Economic Research, 2002.
Neal, Derek (1995). “Industry-Specific Capital: Evidence from Displaced Workers,: Journal of Labor
Economics 13(October): 653-677
Olds, David, et. al. (1997). "Long-term Effects of Home Visitation on Maternal Life Course and Child
Abuse and Neglect: Fifteen Year Follow-up of a Randomized Trail," Journal of the American Medical
Association, 27, 278(8): 637-643.
Reynolds, Arthur J. et. al. (2001). "Long-term Effects of an Early Childhoold Intervention on Educational
Achievement and Juvenile Arrest: A 15 Year Follow-up of Low Income Children in Public Schools,"
Journal of the American Medical Association, Volume 285, pp.
Ruhm, Christopher (1991), ``Are Workers Permanently Scarred by Job Displacements?'' American
Economic Review, 81:1, 319-323.
Smith, Jeffrey and Petra Todd (2003). "Does Matching Overcome LaLonde's Critique of Non-Experimental
Estimators?," Journal of Econometrics, forthcoming.
U.S. Census (2001). “Table 1: Enrollment Status of the Population 3 Years and Over, By Age, Sex, Race,
Hispanic Origin, Nativity, and Selected Educational Characteristics: October 2000,” June 1, 2001, U.S.
Bureau of the Census, web page.

Appendix A
Notes on Administrative Earnings and Community College Records
We constructed our sample of Washington State displaced workers from three administrative data
bases. We received from the state the UI claims records for every worker who filed a valid unemployment
insurance claim between 1990.II and 1994.IV and who had accumulated at least six quarters job tenure.
We matched these data to these workers' quarterly earnings records in UI covered jobs for the period from
1987 until 2000, and to machine-readable transcripts from 25 of the state's community colleges. The
community college records begin in the fall term of 1989 and extend through 1995.
In order to focus our study on displaced workers, we restricted our sample to:
(A) Adults between 22 and 60 years old at the time of their job loss
(B) Workers who had accumulated at least 3 years job tenure with their pre-displacement
employer
(C) Workers who remained continuously attached to the state's UI covered work force
during the 14 year period covered by our quarterly earnings records.
We defined a displaced worker as being “continuously attached” to the state’s work force if she never had
more than 8 consecutive quarters without UI covered earnings, except during the period following their job
loss and when she was enrolled in community college courses. This restriction of our sample meant that
we excluded approximately two-thirds of the available observations from our analysis.
Our review of these excluded individuals indicated that many never had positive wage and salary
earnings in Washington State following their job loss or enrollment in the state’s community colleges.
Although many of these individuals may have moved out of state, we also found the women and older
workers were more likely in this category. These individuals are generally found to have lower mobility
rates or a considered “tied movers.” Interestingly the participation rates in community college schooling
for the displaced workers who were not continuously attached was similar to the rates we report for our
sample in Table 1 in the text (Jacobson, LaLonde, and Sullivan, 2000).
The Washington State sample used in our analysis in this paper contains 65,321 displaced
workers. During the period around their displacements 10,405 completed at least one community college
course. Of these participants in community college schooling, 5,180, or about 50 percent were 35 or older
when they lost their jobs. This sample is smaller than the one used in an earlier paper that followed
displaced workers for fewer years because here we apply the "continuously attached" criteria for a longer
time period (Jacobson, LaLonde, and Sullivan, 2003).
The community college transcript database included information on the type of courses completed
by students. Table A lists ten major categories of community college courses. In our empirical work in the
text we found it helpful to summarize our findings by aggregating these categories into two groupings.
Table A
Classifications of Washington State Community College Classes
Group 1: Quantitative or Technically Oriented Vocational Courses:
Health related courses
Technical/professional courses
Technical trades
College level science and math academic courses
Group 2: Less Quantitative Courses:
Sales/service courses
Vocational courses (not in Group 1)
Social Science/humanities academic courses
Health/PE/consumer oriented courses
Basic skills education
Other courses.

Appendix B
Specification of the Displacement Parameters
Previous research has documented the temporal pattern of the impact of displacement on workers'
earnings (Jacobson, LaLonde, and Sullivan, 1993a,b). Displaced workers’ earnings tend to decline during the
period prior to displacement; drop sharply following the quarter of their job loss; and then rise relatively
rapidly during the next few quarters before increasing at a slower rate in subsequent periods. It may be
important to account for the effects of displacement in our analysis, because this pattern may be associated
with individuals’ decisions to participate in community college schooling.
In Jacobson, LaLonde, and Sullivan, (2003) we found that the following specification was
sufficiently rich to allow for differences in the temporal pattern of displacement among displaced workers.
To control for the average pattern of displacement, we defined the impact of being displaced in period s on
earnings during quarter t as follows:
δit (si , zi ) = δt - s = δk ,
where k = t - si. Letting Dkit = 1 if worker i was displaced at time t - k, we write the displacement effect as
δit (si , zi ) = Σ δk Dkit
In our empirical work, we allow k to range from -12, the twelfth quarter prior to job loss, to the end of the
sample period, which is more than 40 quarters after displacement for some individuals.
We allow displacement effects to vary by workers' characteristics to account for the possibility that
the impact of displacement is correlated with whether a worker receives community college schooling.
Instead of interacting these characteristics with the full vector of displacement indicators, Dkit , we found that a
more parsimonious specification adequately accounts for differences between the average pattern of
displacement effects, δk, and the pattern for workers with characteristics, zi.
We summarize this departure from the average patterns using four variables defined as follows:
F1it = t - (s - 12), if s - 12 ≤ t ≤ s and is equal to 0 otherwise;
F2it = (F1it)2 ;
F3it = 1 if s < t, and is equal to 0 otherwise;
F4it = 1/(t - s), if s < t, and is equal to 0 otherwise.
This specification allows the displacement effects for workers with characteristics zi to differ from the average
effect. This occurs according to a quadratic function during the twelve quarters prior to displacement, and
according to the inverse of the time since displacement during the post-displacement period. The coefficient
associated with the F3it term indicates the departure from the average long-term impact of displacement for
workers with characteristics zi. Therefore, the displacement effect in our econometric model becomes:
(2) δit (si , zi) = Σ δk Dkit + F1itziφ1 + F2itziφ2 + F3itziφ3 + F4itziφ4 .
The vector zi also includes labor market conditions at the time of displacement. To control for local
labor market conditions, we include in the vector, z i , the rates of unemployment and employment growth in
the county during the year prior to workers' job losses. To control for prevailing conditions in worker's former
industries we include the rate of employment growth in workers' two digit (SIC) industry statewide during the
year prior to job loss.

Table 1: Characteristics and Community College Participation of Displaced Workers in Washington State
Panel A: Characteristics of Younger and Older Workers of Washington State Displaced Workers
Males
Worker Characteristic

Age at job loss
Minority
Greater than 6 years prior tenure
Educational Attainment:
Less than a H.S. Degree
More than a H.S. Degree
Prior Industry:
Aerospace
Wood Products
Other Manufacturing
Region of State:
Seattle-Tacoma MSA
Other Counties with MSA's
Rural Counties
Labor Market Conditions at Time of
Job Loss:
County unemployment rate (%)
County employment growth (%)
2 digit Industry Employment
growth (%)

Under 35
T1
C2
(1)
(2)

35 and Over
T1
C2
(3)
(4)

Females
Under 35
35 and Over
T1
C2
T1
C2
(5)
(6)
(7)
(8)

28.7
(3.6)
.12
.12

29.6
(3.5)
.17
.13

43.0
(5.9)
.10
.25

44.0
(6.3)
.13
.23

28.9
(3.7)
.11
.16

28.8
(3.4)
.17
.15

43.6
(5.8)
.09
.28

44.5
(6.2)
.14
.27

.09
.43

.18
.28

.06
.55

.12
.43

.06
.49

.12
.38

.04
.53

.12
.41

.19
.09
.24

.11
.08
.24

.18
.16
.34

.10
.07
.23

.13
.02
.14

.09
.02
.14

.11
.04
.15

.07
.02
.15

.55
.13
.32

.55
.12
.33

.51
.13
.37

.57
.11
.31

.59
.12
.29

.60
.11
.29

.53
.13
.33

.58
.12
.30

7.04
1.50
0.41

7.20
1.54
1.08

7.31
1.13
-0.12

7.06
1.47
1.17

6.94
1.44
1.31

7.00
1.45
1.72

7.09
1.45
1.51

7.04
1.48
2.02

Table 1: Characteristics and Community College Participation of Displaced Workers in Washington State (continued)
Males
Under 35
T1
C2
(1)
(2)
Mean Earnings Prior to Job Loss:
1 - 4 quarters before (in 000's)

5 - 8 quarters before (in 000's)

Number of Observations

Females
35 and Over
T1
C2
(3)
(4)

Under 35
T1
C2
(5)
(6)

35 and Over
T1
C2
(7)
(8)

$26.5
(11.6)

$25.7
(12.1)

$34.5
(15.3)

$33.3
(17.6)

$21.1
(9.7)

$20.5
(10.0)

$24.5
(11.8)

$23.4
(13.2)

$26.7
(11.7)

$26.2
(12.4)

$35.8
(14.8)

$34.5
(17.5)

$21.1
(9.2)

$20.6
(10.2)

$24.7
(11.4)

$23.5
(12.6)

2,936

14,560

2,371

19,342

2,291

7,462

2,809

13,552

Panel B: Participation Rates and Community College Credits Completed by Displaced Workers

All Credits:
Males Under 35
Males 35 and Over
Females Under 35
Females 35 and Over

Rate3
.168
.109
.235
.172

Mean4

Std5

29.5
27.4
27.3
23.5

33.3
34.0
32.3
30.8

Proportion with Number of Completed Community
College Credits
1-5
6-10
11-20
21-40
41-75
75+
.27
.33
.32
.39

.16
.16
.17
.16

.16
.16
.15
.14

.15
.12
.14
.11

.13
.11
.12
.10

.13
.12
.12
.10

Table 1: Characteristics and Community College Participation of Displaced Workers in Washington State (continued)

Group1 Credits:

Mean4

Std5

Males Under 35
Males 35 and Over
Females Under 35
Females 35 and Over

15.2
15.3
8.8
8.4

24.3
24.9
16.8
16.0

Group1 Credits:

Mean4

Std5

Males Under 35
Males 35 and Over
Females Under 35
Females 35 and Over

14.3
12.2
18.5
15.1

20.9
19.8
23.5
22.4

Proportion with Number of Completed Group 1
Community College Credits
0
1-5
6-20
21+
.34
.29
.46
.41

.21
.28
.24
.29

.23
.21
.18
.19

.22
.21
.12
.11

Proportion with Number of Completed Group 2
Community College Credits
0
1-5
6-20
21+
.29
.33
.15
.21

.23
.24
.27
.31

.26
.26
.30
.27

.22
.18
.28
.21

Notes: Panel A: <1>T denotes the training groups. We define displaced workers as trainees or community college participants if they complete at least one
credit. <2> C denotes the comparison group. The comparison group consists of displaced workers who either never enrolls in community college or who enroll
but dropped out before completing one course. We excluded from our sample workers who completed more than three academic years (135 credits) of
community college schooling. Fractions are the proportions of indicated group with the given characteristic. The numbers in parentheses are the standard
deviations. Panel B: Credits accumulated in Washington State community colleges by workers displaced between 1990 and 1995. Group 1 credits are from
courses that teach more quantitatively oriented vocational material, including courses training for the health occupations and the construction trades, and that
teach academic math and science courses. Group 2 credits are from all other community college courses. <3> Rate is the participation rate in community college
schooling around the time of workers job loss. This fraction is the ratio of displaced workers who complete at least one credit to all displaced workers in the
indicated demographic group. <4> Mean is the mean number of credits completed among those who completed at least one community college credit. <5> Std
denotes the sample standard deviation.

Table 2a: Participation in Community College by Age of Displaced Workers
Credits Completed

Probability of
Completing One or
More Credits

Probability of
Probability of Earning Credits Earned Given
Enrolling in a Credit Credits Given
At Least One Credit
Course
Enrollment

Males:

20-24

6.72
(0.53)

0.182
(0.012)

0.222
(0.013)

0.013
(0.038)

12.61
(3.63)

25-29

3.87
(0.45)

0.108
(0.010)

0.132
(0.011)

0.010
(0.037)

9.99
(3.48)

30-34

2.68
(0.45)

0.071
(0.010)

0.091
(0.011)

-0.009
(0.037)

9.19
(3.84)

35-39

2.30
(0.45)

0.053
(0.010)

0.065
(0.011)

0.008
(0.037)

10.31
(3.51)

40-44

1.51
(0.46)

0.040
(0.010)

0.051
(0.011)

-0.006
(0.038)

6.67
(3.59)

45-49

1.60
(0.48)

0.038
(0.010)

0.039
(0.011)

0.047
(0.039)

8.17
(3.69)

50-54

0.97
(0.51)

0.026
(0.011)

0.031
(0.012)

0.008
(0.041)

4.87
(3.93)

55-60

0.0

39,208

Observations

0.0

6,567

5,306

Notes: Figures in columns 1 and 5 of the table are from a regression with the indicated column heading as the dependent variable and with an intercept and
indicators for the age ranges shown. The figures in columns 2 through 4 are coefficients from a linear probability model with an intercept and indicators for the
age ranges shown. No other controls are included in the regressions. Information on the sample is given in the text and in Appendix A. Numbers in parentheses
are standard errors.

Table 2a: Participation in Community College by Age of Displaced Workers (continued)
Credits Completed

Females:
20-24

Probability of
Completing One or
More Credits

Probability of
Probability of Earning Credits Earned Given
Enrolling in a Credit Credits Given
At Least One Credit
Course
Enrollment

10.51
(0.72)

0.227
(0.017)

0.258
(0.018)

0.004
(0.035)

21.40
(3.19)

25-29

4.95
(0.60)

0.128
(0.014)

0.151
(0.015)

0.013
(0.033)

12.92
(2.98)

30-34

3.30
(0.58)

0.081
(0.013)

0.099
(0.014)

-0.003
(0.032)

10.97
(2.97)

35-39

3.21
(0.58)

0.071
(0.013)

0.087
(0.014)

-0.004
(0.032)

11.79
(2.97)

40-44

2.83
(0.58)

0.061
(0.013)

0.074
(0.014)

0.004
(0.033)

10.95
(2.98)

45-49

2.30
(0.60)

0.047
(0.014)

0.057
(0.015)

-0.001
(0.034)

9.99
(3.07)

50-54

1.05
(0.64)

0.028
(0.015)

0.031
(0.016)

0.009
(0.036)

4.55
(3.28)

55-60

0.0

26,113

Observations

0.0

6,156

5,099

Table 2b: Adjusted Participation in Community College by Age of Displaced Workers
Credits Completed

Probability of
Completing One or
More Credits

Probability of
Probability of Earning Credits Earned Given
Enrolling in a Credit Credits Given
At Least One Credit
Course
Enrollment

Males:

20-24

6.77
(0.53)

0.191
(0.011)

0.229
(0.013)

0.023
(0.035)

10.65
(3.63)

25-29

3.61
(0.45)

0.107
(0.010)

0.130
(0.011)

0.027
(0.033)

7.90
(3.44)

30-34

2.47
(0.44)

0.070
(0.010)

0.090
(0.010)

-0.005
(0.033)

7.36
(3.42)

35-39

1.95
(0.44)

0.046
(0.010)

0.061
(0.010)

0.002
(0.033)

8.48
(3.44)

40-44

0.98
(0.42)

0.027
(0.010)

0.042
(0.010)

-0.017
(0.034)

5.07
(3.50)

45-49

1.15
(0.47)

0.024
(0.010)

0.032
(0.010)

0.006
(0.035)

5.71
(3.60)

50-54

0.79
(0.50)

0.021
(0.011)

0.030
(0.011)

-0.035
(0.037)

3.78
(3.83)

55-60

0.0

39,208

Observations

0.0

6,568

5,306

Notes: See Table 2A. Figures are coefficients for the indicators of the age ranges shown in the table. All models include controls for prior schooling, prior
industry, earnings in year prior to displacement, tenure on pre-displacement job, minority status, region of state, county unemployment employment growth rates,
the statewide employment growth rate in the individual’s prior two digit industry and quarter and year of job loss. Numbers in parentheses are standard errors.

Table 2b: Adjusted Participation in Community College by Age of Displaced Workers (continued)
Credits Completed

Probability of
Completing One or
More Credits

Probability of
Probability of Earning Credits Earned Given
Enrolling in a Credit Credits Given
At Least One Credit
Course
Enrollment

Females:

20-24

10.30
(0.72)

0.225
(0.017)

0.258
(0.018)

0.050
(0.032)

21.15
(3.17)

25-29

4.55
(0.59)

0.121
(0.014)

0.147
(0.015)

0.028
(0.030)

12.00
(2.95)

30-34

2.92
(0.58)

0.073
(0.013)

0.094
(0.014)

0.013
(0.030)

10.18
(2.92)

35-39

2.72
(0.57)

0.059
(0.013)

0.079
(0.014)

0.009
(0.030)

11.20
(2.92)

40-44

2.32
(0.57)

0.048
(0.013)

0.067
(0.014)

-0.004
(0.030)

9.94
(2.93)

45-49

1.72
(0.59)

0.032
(0.014)

0.483
(0.015)

-0.010
(0.031)

8.57
(3.01)

50-54

0.80
(0.62)

0.023
(0.015)

0.029
(0.016)

-0.005
(0.032)

4.50
(3.22)

55-60

0.0

26,113

Observations

0.0

6,156

5,099

Table 3a: Impact of Community College Schooling on Male Displaced Workers' Earnings
(Estimates from alternative specifications on quarterly earnings)
Modela
Under 35:
In College

(1)

(2)

(3)

(4)

(5)

(6)

347.52
(81.85)

429.12
(81.82)

353.67
(75.90)

400.68
(75.61)

347.94
(81.97)

417.19
(81.60)

In College*Credits/Qtr

-254.71
(10.72)

-253.85
(10.71)

-256.24
(11.17)

-241.30
(10.89)

-255.86
(11.32)

-242.64
(11.05)

Post-Collegeb

-26.72
(66.36)

352.09
(82.34)

-15.46
(78.98)

109.04
(100.85)

Post-College*1/kc

-976.87
(91.24)

-310.05
(122.26)
-0.71
(1.47)

Post-College*Credits

-0.52
(1.74)

-27.39
(2.06)

Post-College*Credits*1/k
35 and Older:
In College

10.54
(1.81)

9.09
(2.23)
-22.92
(2.80)

-83.31
(109.26)

38.78
(108.09)

-43.93
(101.98)

-7.40
(101.77)

-67.61
(110.02)

12.60
(108.84)

-285.64
(12.08)

-284.30
(12.07)

-297.45
(13.08)

-273.71
(12.74)

-295.55
(13.28)

-275.10
(12.93)

In College*Credits/Qtr
Post-Collegeb
Post-College*1/kc
Post-College*Credits

-1379.82
(120.91)

-524.29
(83.77)
-4.69
(1.82)

10.83
(2.13)
-38.34
(2.46)

Post-College*Credits*1/k

-3.89
(2.17)

8.94
(2.63)
-30.98
(3.28)

Notes: a. Dependent variable is quarterly earnings. All models include demographic, heterogeneous displacement, and in-college controls as well as individual
and period-specific fixed effects and worker-specific time trends. Robust standard errors are in parentheses.
b. Post-College is an indicator variable for whether the current quarter is after the training participant left community college.
c. 1/k denotes the reciprocal of the number of quarters after the trainee left school.

Table 3b: Impact of Community College Schooling on Female Displaced Workers' Earnings
(Estimates from alternative specifications on quarterly earnings)
Modela
Under 35:
In College

(1)

(2)

(3)

(4)

(5)

(6)

232.12
(80.27)

289.12
(79.70)

265.59
(75.33)

295.18
(75.02)

227.50
(80.65)

263.59
(79.83)

In College*Credits/Qtr

-191.86
(10.45)

-191.22
(10.44)

-190.14
(11.26)

-174.17
(11.09)

-186.68
(11.45)

-171.64
(11.27)

Post-Collegeb

-59.62
(59.25)

212.38
(74.26)

-101.41
(68.68)

-76.46
(88.94)

Post-College*1/kc

-696.94
(67.06)

-33.86
(86.89)
0.60
(1.59)

Post-College*Credits

2.00
(1.84)

-25.54
(1.57)

Post-College*Credits*1/k
35 and Older:
In College

11.08
(1.98)

12.13
(2.36)
-24.97
(2.03)

190.72
(70.39)

257.23
(69.93)

196.29
(64.45)

217.69
(64.27)

194.41
(71.17)

232.30
(70.33)

In College*Credits/Qtr

-215.10
(9.58)

-214.37
(9.58)

-218.27
(10.81)

-198.64
(10.54)

-218.23
(11.07)

-200.02
(10.77)

Post-Collegeb

-23.01
(57.99)

279.26
(70.73)

-3.74
(66.14)

72.13
(82.65)

Post-College*1/kc
Post-College*Credits

-779.48
(81.68)

-182.55
(103.09)
-1.14
(1.60)

10.56
(1.91)
-28.35
(2.03)

Post-College*Credits*1/k

-1.11
(1.83)

9.56
(2.23)
-25.60
(2.61)

Table 4: Impact of Community College Credits on Earnings Using a Step Function for Completed Credits
Displaced Males
Credits

Under 35

Displaced Females

35 and Older

Under 35

35 and Older

226.71
(143.82)
180.63
(169.98)

166.41
(171.07)
65.72
(249.23)

-23.03
(118.97)
78.01
(148.57)

42.15
(103.47)
206.89
(145.24)

11 – 20

126.29
(187.71)

403.43
(221.48)

418.33
(167.14)

598.18
(170.36)

21 – 40

459.13
(179.73)

594.14
(244.32)

-256.33
(183.54)

154.38
(179.85)

41 – 75

398.16
(193.79)

715.07
(272.17)

507.48
(217.73)

348.07
(193.69)

76+

1021.92
(209.51)

924.07
(242.89)

1281.70
(228.31)

1069.35
(224.87)

1–5
6 – 10

Notes: The figures in the table are estimates of the long-run impact of the indicated number of community college credits on
quarterly earnings. All models allow for different long-term and short-term effects of schooling. See specification (6) in text and
notes to Table 3a and 3b.

Table 5: Impact of Community College Courses by Type of Credits
(Short and Long-Run Impacts of Group 1 and Group 2 Courses)
Males
Under 35
In College
In College*Group 1
Credits/Qtr
In College*Group 2
Credits/Qtr
Post-Collegeb
Post-College*1/kc
Post-College*Group 1
Credits
Post-College*Group 1
Credits*1/k
Post-College*Group 2
Credits
Post-College*Group 2
Credits*1/k

Females
35 and Older

Under 35

35 and Older

488.41
(81.88)

20.45
(109.39)

262.36
(79.76)

236.57
(70.14)

-209.68
(13.34)

-256.41
(15.43)

-165.51
(20.94)

-198.15
(19.79)

-288.70
(15.16)

-302.33
(19.59)

-174.05
(12.80)

-202.06
(13.24)

107.89
(101.08)

151.00
(126.99)

-56.10
(88.46)

77.07
(82.64)

-275.65
(122.64)

-508.65
(157.97)

-33.59
(109.45)

-183.03
(103.08)

12.05
(3.14)

12.39
(3.77)

23.72
(6.29)

18.48
(5.33)

-15.93
(3.81)

-27.28
(4.46)

-25.84
(6.58)

-26.85
(5.46)

5.82
(3.74)

4.16
(4.43)

5.47
(2.99)

4.29
(3.10)

-32.58
(4.45)

-36.85
(5.47)

-24.61
(3.45)

-24.89
(3.56)

Notes: a. The dependent variable is quarterly earnings. All models include demographic, heterogeneous displacement, and in-college controls as well as
individual and period-specific fixed effects and worker-specific time trends. Robust standard errors are in parentheses. See Appendix A for definition of Group 1
and Group 2 courses.
b. Post-College is an indicator variable for whether the current quarter is after the training participant left community college.
c. 1/k denotes the reciprocal of the number of quarters after the trainee left school.

Table 6: Does Community College Participation Predict Earnings Prior to Retraining?
(Predicted "Effect" of Community College Participation and Completed Credits on Pre-Enrollment Earnings)
Controls for Temporal Pattern of Cost of Displacement
None

Just
Overall
Dummies a

Add
Demographicsb

Add
Prior
Industryc

Add
Labor
Marketd

Males Under 35
One Year Prior to
Displacement:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits
Post-Displacement/Prior to
Enrollment:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits

-143.79
(45.65)

28.41
(42.97)

66.50
(43.23)

87.81
(43.28)

89.59
(43.30)

5.21
(1.45)

4.84
(1.35)

4.77
(1.36)

5.18
(1.38)

5.15
(1.38)

5.08
(1.62)

3.12
(1.54)

3.04
(1.58)

3.29
(1.60)

3.27
(1.60)

-1,495
(147.06)

205.99
(141.65)

-115.01
(144.23)

1.72
(143.10)

3.54
(142.14)

-26.66
(5.32)

-17.11
(4.83)

-15.92
(5.11)

-12.40
(4.79)

-11.41
(4.81)

-35.51
(6.22)

-21.69
(5.51)

-23.38
(5.98)

-18.29
(5.59)

-18.28
(5.54)

Table 6: Does Community College Participation Predict Earnings Prior to Retraining? (continued)
(Predicted "Effect" of Community College Participation and Completed Credits on Pre-Enrollment Earnings)
Controls for Temporal Pattern of Cost of Displacement
None

Just
Overall
Dummies a

Add
Demographicsb

Add
Prior
Industryc

Add
Labor
Marketd

Males 35 and Older
One Year Prior to
Displacement:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits
Post-Displacement/Prior to
Enrollment:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits

-253.97
(61.66)

-121.75
(59.37)

-66.49
(59.52)

-49.14
(59.38)

-53.98
(59.36)

7.40
(1.80)

7.99
(1.69)

8.11
(1.70)

9.23
(1.70)

9.29
(1.70)

8.85
(2.14)

10.07
(2.04)

10.84
(2.02)

12.70
(2.03)

12.68
(2.01)

-2,989
(184.71)

-1,138
(178.09)

-600.11
(178.83)

-495.78
(177.87)

-470.19
(176.28)

-24.93
(6.06)

-15.67
(5.85)

-17.57
(5.88)

-12.64
(5.72)

-11.99
(5.71)

-25.15
(7.49)

-13.88
(7.11)

-17.45
(7.00)

-11.70
(6.15)

-12.89
(6.02)

Just
Overall
Dummies a

Add
Demographicsb

Add
Prior
Industryc

Add
Labor
Marketd

Females Under 35
One Year Prior to
Displacement:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits
Post-Displacement/Prior to
Enrollment:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits

-94.97
(42.11)

56.79
(38.91)

-73.42
(39.32)

-57.11
(39.20)

-56.04
(39.22)

7.19
(2.36)

5.82
(2.27)

5.98
(2.28)

6.65
(2.28)

6.72
(2.28)

6.19
(1.38)

6.22
(1.23)

6.14
(1.22)

6.35
(1.21)

6.34
(1.21)

-1,110.9
(116.75)

634.36
(111.47)

0.97
(113.17)

38.97
(109.70)

44.22
(108.96)

-35.68
(6.68)

-19.33
(6.33)

-18.24
(6.41)

-14.19
(6.02)

-13.05
(6.07)

-18.80
(4.31)

-11.16
(3.95)

-14.89
(4.01)

-13.45
(3.70)

-13.01
(3.71)

Just
Overall
Dummies a

Add
Demographicsb

Add
Prior
Industryc

Add
Labor
Marketd

Females 35 and Older
One Year Prior to
Displacement:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits
Post-Displacement/Prior to
Enrollment:
Pre-College
Pre-College*Group 1
Credits
Pre-College*Group 2
Credits

17.36
(42.76)

141.45
(39.79)

48.14
(40.20)

39.14
(39.83)

39.50
(39.83)

8.20
(2.59)

6.97
(2.50)

6.11
(2.51)

6.68
(2.50)

6.85
(2.50)

0.41
(1.54)

1.94
(1.41)

2.71
(1.41)

4.06
(1.39)

4.01
(1.39)

-2,082
(115.58)

-194.99
(111.31)

-195.68
(112.33)

-224.94
(108.74)

-223.93
(107.42)

-32.97
(7.21)

-20.36
(7.01)

-14.17
(6.83)

-9.70
(6.21)

-10.16
(6.15)

-22.42
(4.58)

-10.77
(4.37)

-18.13
(4.18)

-14.10
(3.97)

-13.71
(3.91)

Notes: a. Dependent variable is quarterly earnings. All models include demographic, in-college controls as well as individual and
worker-specific fixed effects and worker-specific time trends. Robust standard errors are in parentheses. The dummy variables that
control for the temporal pattern of the cost of displacement are described in Appendix B. b. Allows for heterogeneity in the temporal
pattern of displacement according to an individuals' minority status, region of the state, prior schooling, and prior tenure on the job. c.
Allows in addition to b. heterogeneity according to an individual's prior industry. d. Allows in addition to c. heterogeneity according
to an individual's labor market conditions at displacement. e. Pre-College is an indicator variable during the period indicated in the
table for whether the individual subsequently enrolled and completed at least one community college credit.

Table 7: Sensitivity of Long-run Impacts of Community College
Retraining to Specification of the Displacement Effects

(1)

None

Controls for Displacement Effects
(2)
(3)
(4)
(3) Plus Prior
Dummy
(2) Plus
Industry
Variables
Demographics

(5)
(4) Plus Labor
Market

Males Under 35:
177.75
(110.52)

174.70
(104.17)

46.11
(102.74)

111.71
(101.34)

107.89
(101.08)

Group 1*Post College

8.13
(3.56)

11.80
(3.27)

12.95
(3.24)

12.14
(3.13)

12.05
(3.14)

Group 2*Post College

-1.72
(4.04)

4.82
(3.86)

6.23
(3.87)

5.69
(3.74)

5.82
(3.74)

-140.39
(135.55)

-52.19
(129.04)

90.26
(127.31)

159.24
(127.17)

151.00
(126.99)

Group 1*Post College

10.70
(4.04)

13.56
(3.84)

12.84
(3.83)

12.29
(3.77)

12.39
(3.77)

Group 2*Post College

-0.97
(4.92)

7.57
(4.62)

5.30
(4.54)

4.11
(4.42)

4.16
(4.42)

Post College

Males 35 or Older:
Post College

Table 7: Sensitivity of Long-run Impacts of Community College
Retraining to Specification of the Displacement Effects (continued)
Controls for Displacement Effects
(1)

(2)

(3)

(4)

(5)

None

Dummy
Variables

(2) Plus
Demographics

(3) Plus Prior
Industry

(4) Plus Labor
Market

Females Under 35:
157.41
(96.51)

168.46
(91.79)

-98.50
(89.90)

-56.62
(88.74)

-56.10
(88.46)

Group 1*Post College

17.24
(6.54)

20.47
(6.51)

24.57
(6.38)

23.89
(6.29)

23.72
(6.29)

Group 2*Post College

1.09
(3.39)

4.28
(3.19)

6.15
(3.08)

5.47
(3.00)

5.47
(2.99)

-35.36
(90.27)

90.15
(84.50)

59.41
(83.88)

71.91
(82.79)

77.07
(82.64)

Group 1*Post College

14.30
(5.89)

19.24
(5.47)

18.71
(5.44)

18.40
(5.33)

18.48
(5.33)

Group 2*Post College

0.87
(3.42)

4.99
(3.20)

4.64
(3.22)

4.47
(3.11)

4.29
(3.11)

Post College

Females 35 or Older:
Post College

Note: See Tables 3a, 3b, and 6. For discussion of specification of the effects of displacement on earnings see Appendix B. In column 3 we allow the temporal
pattern of displacement effects to vary by minority status, age at displacement, prior schooling, region of the state, prior tenure on the pre-displacement job. In
column 4, we add to this specification controls for prior industry. In column 5, we add controls for labor market conditions at the time of job loss. The conditions
are the country unemployment rate, employment growth during the prior year in the country, and statewide employment growth during the prior year in the
individuals’ prior 2-digit SIC industry.

Table 8: The Net Benefit and Internal Rates of Return from an Academic Year of
Community College Schooling for Displaced Workers
Panel A: Cost-Benefit Analysis of Investments in Displaced Workers' Retraining
Exclude "Just Showing Up" Effect
Males
Females
Young
Old
Young
Old
(1)
(2)
(3)
(4)
From Perspective of Participants:
$13.1
$5.2
$21.9
9.1
…Net Benefit (in 000s)
3.07
1.69
5.40
2.61
…Benefit to Cost Ratio
From Perspective of Society:
$9.9
-$0.3
$21.0
$4.2
…Net Benefit (in 000s)
1.61
0.98
2.43
1.27
…Benefit to Cost Ratio

Include "Just Showing Up" Effect
Males
Females
Young
Old
Young
Old
(5)
(6)
(7)
(8)
$18.2
3.88

$9.9
2.30

$17.7
4.56

$11.6
3.05

$16.1
2.04

$5.9
1.34

$15.5
2.05

$7.6
1.49

Panel B: Alternative Social Internal Rates of Return for One Academic Year of Retraining
Exclude "Just Showing Up" Effect
Include "Just Showing Up" Effect
Treatment of "In-School Effects" as Opportunity Costs
No
1/2
Yes
No
1/2
(1)
(2)
(3)
(4)
(5)
10.4%
7.4%
5.4%
12.6%
9.2%
Younger Men
7.8%
3.9%
1.4%
10.8%
6.5%
Older Men
13.9%
11.1%
9.1%
12.0%
9.4%
Younger Women
9.5%
6.2%
4.0%
11.0%
7.8%
Older Women

Yes
(6)
7.1%
3.9%
7.6%
5.5%

Notes: Calculations based on estimates in column 6 of Table 3a for males and column 6 of Table 3b for females. We assume that the remaining work life is 22
years for older displaced workers and 36 years for younger workers. In panel A, we discount future per-period earnings impacts at a real rate of 4 percent. We
also assume that individuals pay taxes of 25 percent on their increased incomes. We assume the total costs of a year of school equal $8,000 per year and that
students pay about 20 percent or $1,500 of this direct cost through their tuition. The remaining amount is paid by taxpayers. For the calculations in Panel A, we
assume the opportunity cost of schooling equal 1/2 the costs implied by the "In-College" estimates reported in Tables 3a and 3b. In Panel, B we make the
indicated alternative assumptions about the opportunity cost of retraining. All figures in Panel B are the social internal rates of return. Finally, we assume that the
welfare costs associated with the taxes raised to subsidized community college schooling equals 50 percent of the subsidy or $3,250.

Figure 1: Earnings of Trainees and Comparisons
Workers Under 35 Displaced In 1991

8ooo~==============================================~

...

!...

7000

ca
:I

...

8000

f!
ca

5000

•a.

,:g

•

........••••

4000

...................----·..-----..........
...
...
•

I...

3000
2000

-8

-4

quarter relative to displacement

PLOT

group 1 concentrators
comparisons

• • • • • • • · group 2 concentrators

Figure 2: Earnings or Trainees and Comparisons
Workers 35 and OVer Displaced In 1991
8000

...

...
!...
ca
:I

7000
8000
5000

f!
ca

4000

I...

-~,

............··..~'
\.

•...a.

,:g

~'
••
•

, .. .,.

,,,~

If
.v ~·
~:~'

3000

...

2000

---

~·

11111!: ....

--.::;,;1"

··--,~
......

.. ........

__
-------~
._...

.............

-.•""""

I I

....- ........... _
••..........,..

..........

······

...................

~·

1000
I

-8

I I

-4

I I

quarter relative to displacement

PLOT

group 1 concentrators
comparisons

• • • • • • • · group 2 concentrators

dollars per quarter

Program Evaluation Model
1000
900
800
700
600
500
400
300
200
100
0
-100
-200
-300
-400
-500
0

quarters after leaving school
PLOT

young workers

old workers

Figure 4: Time Pattern of Training Impacts
Hybrid Model

dollars per quarter per credit

-10

-20
0

quarters after leaving school
PLOT

young workers

old workers

Working Paper Series
A series of research studies on regional economic issues relating to the Seventh Federal
Reserve District, and on financial and economic topics.
Dynamic Monetary Equilibrium in a Random-Matching Economy
Edward J. Green and Ruilin Zhou

WP-00-1

The Effects of Health, Wealth, and Wages on Labor Supply and Retirement Behavior
Eric French

WP-00-2

Market Discipline in the Governance of U.S. Bank Holding Companies:
Monitoring vs. Influencing
Robert R. Bliss and Mark J. Flannery

WP-00-3

Using Market Valuation to Assess the Importance and Efficiency
of Public School Spending
Lisa Barrow and Cecilia Elena Rouse
Employment Flows, Capital Mobility, and Policy Analysis
Marcelo Veracierto
Does the Community Reinvestment Act Influence Lending? An Analysis
of Changes in Bank Low-Income Mortgage Activity
Drew Dahl, Douglas D. Evanoff and Michael F. Spivey

WP-00-4

WP-00-5

WP-00-6

Subordinated Debt and Bank Capital Reform
Douglas D. Evanoff and Larry D. Wall

WP-00-7

The Labor Supply Response To (Mismeasured But) Predictable Wage Changes
Eric French

WP-00-8

For How Long Are Newly Chartered Banks Financially Fragile?
Robert DeYoung

WP-00-9

Bank Capital Regulation With and Without State-Contingent Penalties
David A. Marshall and Edward S. Prescott

WP-00-10

Why Is Productivity Procyclical? Why Do We Care?
Susanto Basu and John Fernald

WP-00-11

Oligopoly Banking and Capital Accumulation
Nicola Cetorelli and Pietro F. Peretto

WP-00-12

Puzzles in the Chinese Stock Market
John Fernald and John H. Rogers

WP-00-13

The Effects of Geographic Expansion on Bank Efficiency
Allen N. Berger and Robert DeYoung

WP-00-14

Idiosyncratic Risk and Aggregate Employment Dynamics
Jeffrey R. Campbell and Jonas D.M. Fisher

WP-00-15

Working Paper Series (continued)
Post-Resolution Treatment of Depositors at Failed Banks: Implications for the Severity
of Banking Crises, Systemic Risk, and Too-Big-To-Fail
George G. Kaufman and Steven A. Seelig

WP-00-16

The Double Play: Simultaneous Speculative Attacks on Currency and Equity Markets
Sujit Chakravorti and Subir Lall

WP-00-17

Capital Requirements and Competition in the Banking Industry
Peter J.G. Vlaar

WP-00-18

Financial-Intermediation Regime and Efficiency in a Boyd-Prescott Economy
Yeong-Yuh Chiang and Edward J. Green

WP-00-19

How Do Retail Prices React to Minimum Wage Increases?
James M. MacDonald and Daniel Aaronson

WP-00-20

Financial Signal Processing: A Self Calibrating Model
Robert J. Elliott, William C. Hunter and Barbara M. Jamieson

WP-00-21

An Empirical Examination of the Price-Dividend Relation with Dividend Management
Lucy F. Ackert and William C. Hunter

WP-00-22

Savings of Young Parents
Annamaria Lusardi, Ricardo Cossa, and Erin L. Krupka

WP-00-23

The Pitfalls in Inferring Risk from Financial Market Data
Robert R. Bliss

WP-00-24

What Can Account for Fluctuations in the Terms of Trade?
Marianne Baxter and Michael A. Kouparitsas

WP-00-25

Data Revisions and the Identification of Monetary Policy Shocks
Dean Croushore and Charles L. Evans

WP-00-26

Recent Evidence on the Relationship Between Unemployment and Wage Growth
Daniel Aaronson and Daniel Sullivan

WP-00-27

Supplier Relationships and Small Business Use of Trade Credit
Daniel Aaronson, Raphael Bostic, Paul Huck and Robert Townsend

WP-00-28

What are the Short-Run Effects of Increasing Labor Market Flexibility?
Marcelo Veracierto

WP-00-29

Equilibrium Lending Mechanism and Aggregate Activity
Cheng Wang and Ruilin Zhou

WP-00-30

Impact of Independent Directors and the Regulatory Environment on Bank Merger Prices:
Evidence from Takeover Activity in the 1990s
Elijah Brewer III, William E. Jackson III, and Julapa A. Jagtiani
Does Bank Concentration Lead to Concentration in Industrial Sectors?
Nicola Cetorelli

WP-00-31

WP-01-01

Working Paper Series (continued)
On the Fiscal Implications of Twin Crises
Craig Burnside, Martin Eichenbaum and Sergio Rebelo

WP-01-02

Sub-Debt Yield Spreads as Bank Risk Measures
Douglas D. Evanoff and Larry D. Wall

WP-01-03

Productivity Growth in the 1990s: Technology, Utilization, or Adjustment?
Susanto Basu, John G. Fernald and Matthew D. Shapiro

WP-01-04

Do Regulators Search for the Quiet Life? The Relationship Between Regulators and
The Regulated in Banking
Richard J. Rosen
Learning-by-Doing, Scale Efficiencies, and Financial Performance at Internet-Only Banks
Robert DeYoung
The Role of Real Wages, Productivity, and Fiscal Policy in Germany’s
Great Depression 1928-37
Jonas D. M. Fisher and Andreas Hornstein

WP-01-05

WP-01-06

WP-01-07

Nominal Rigidities and the Dynamic Effects of a Shock to Monetary Policy
Lawrence J. Christiano, Martin Eichenbaum and Charles L. Evans

WP-01-08

Outsourcing Business Service and the Scope of Local Markets
Yukako Ono

WP-01-09

The Effect of Market Size Structure on Competition: The Case of Small Business Lending
Allen N. Berger, Richard J. Rosen and Gregory F. Udell

WP-01-10

Deregulation, the Internet, and the Competitive Viability of Large Banks
and Community Banks
Robert DeYoung and William C. Hunter

WP-01-11

Price Ceilings as Focal Points for Tacit Collusion: Evidence from Credit Cards
Christopher R. Knittel and Victor Stango

WP-01-12

Gaps and Triangles
Bernardino Adão, Isabel Correia and Pedro Teles

WP-01-13

A Real Explanation for Heterogeneous Investment Dynamics
Jonas D.M. Fisher

WP-01-14

Recovering Risk Aversion from Options
Robert R. Bliss and Nikolaos Panigirtzoglou

WP-01-15

Economic Determinants of the Nominal Treasury Yield Curve
Charles L. Evans and David Marshall

WP-01-16

Price Level Uniformity in a Random Matching Model with Perfectly Patient Traders
Edward J. Green and Ruilin Zhou

WP-01-17

Earnings Mobility in the US: A New Look at Intergenerational Inequality
Bhashkar Mazumder

WP-01-18

Working Paper Series (continued)
The Effects of Health Insurance and Self-Insurance on Retirement Behavior
Eric French and John Bailey Jones

WP-01-19

The Effect of Part-Time Work on Wages: Evidence from the Social Security Rules
Daniel Aaronson and Eric French

WP-01-20

Antidumping Policy Under Imperfect Competition
Meredith A. Crowley

WP-01-21

Is the United States an Optimum Currency Area?
An Empirical Analysis of Regional Business Cycles
Michael A. Kouparitsas

WP-01-22

A Note on the Estimation of Linear Regression Models with Heteroskedastic
Measurement Errors
Daniel G. Sullivan

WP-01-23

The Mis-Measurement of Permanent Earnings: New Evidence from Social
Security Earnings Data
Bhashkar Mazumder

WP-01-24

Pricing IPOs of Mutual Thrift Conversions: The Joint Effect of Regulation
and Market Discipline
Elijah Brewer III, Douglas D. Evanoff and Jacky So

WP-01-25

Opportunity Cost and Prudentiality: An Analysis of Collateral Decisions in
Bilateral and Multilateral Settings
Herbert L. Baer, Virginia G. France and James T. Moser

WP-01-26

Outsourcing Business Services and the Role of Central Administrative Offices
Yukako Ono

WP-02-01

Strategic Responses to Regulatory Threat in the Credit Card Market*
Victor Stango

WP-02-02

The Optimal Mix of Taxes on Money, Consumption and Income
Fiorella De Fiore and Pedro Teles

WP-02-03

Expectation Traps and Monetary Policy
Stefania Albanesi, V. V. Chari and Lawrence J. Christiano

WP-02-04

Monetary Policy in a Financial Crisis
Lawrence J. Christiano, Christopher Gust and Jorge Roldos

WP-02-05

Regulatory Incentives and Consolidation: The Case of Commercial Bank Mergers
and the Community Reinvestment Act
Raphael Bostic, Hamid Mehran, Anna Paulson and Marc Saidenberg
Technological Progress and the Geographic Expansion of the Banking Industry
Allen N. Berger and Robert DeYoung

WP-02-06

WP-02-07

Working Paper Series (continued)
Choosing the Right Parents: Changes in the Intergenerational Transmission
of Inequality  Between 1980 and the Early 1990s
David I. Levine and Bhashkar Mazumder

WP-02-08

The Immediacy Implications of Exchange Organization
James T. Moser

WP-02-09

Maternal Employment and Overweight Children
Patricia M. Anderson, Kristin F. Butcher and Phillip B. Levine

WP-02-10

The Costs and Benefits of Moral Suasion: Evidence from the Rescue of
Long-Term Capital Management
Craig Furfine

WP-02-11

On the Cyclical Behavior of Employment, Unemployment and Labor Force Participation
Marcelo Veracierto

WP-02-12

Do Safeguard Tariffs and Antidumping Duties Open or Close Technology Gaps?
Meredith A. Crowley

WP-02-13

Technology Shocks Matter
Jonas D. M. Fisher

WP-02-14

Money as a Mechanism in a Bewley Economy
Edward J. Green and Ruilin Zhou

WP-02-15

Optimal Fiscal and Monetary Policy: Equivalence Results
Isabel Correia, Juan Pablo Nicolini and Pedro Teles

WP-02-16

Real Exchange Rate Fluctuations and the Dynamics of Retail Trade Industries
on the U.S.-Canada Border
Jeffrey R. Campbell and Beverly Lapham

WP-02-17

Bank Procyclicality, Credit Crunches, and Asymmetric Monetary Policy Effects:
A Unifying Model
Robert R. Bliss and George G. Kaufman

WP-02-18

Location of Headquarter Growth During the 90s
Thomas H. Klier

WP-02-19

The Value of Banking Relationships During a Financial Crisis:
Evidence from Failures of Japanese Banks
Elijah Brewer III, Hesna Genay, William Curt Hunter and George G. Kaufman

WP-02-20

On the Distribution and Dynamics of Health Costs
Eric French and John Bailey Jones

WP-02-21

The Effects of Progressive Taxation on Labor Supply when Hours and Wages are
Jointly Determined
Daniel Aaronson and Eric French

WP-02-22

Working Paper Series (continued)
Inter-industry Contagion and the Competitive Effects of Financial Distress Announcements:
Evidence from Commercial Banks and Life Insurance Companies
Elijah Brewer III and William E. Jackson III

WP-02-23

State-Contingent Bank Regulation With Unobserved Action and
Unobserved Characteristics
David A. Marshall and Edward Simpson Prescott

WP-02-24

Local Market Consolidation and Bank Productive Efficiency
Douglas D. Evanoff and Evren Örs

WP-02-25

Life-Cycle Dynamics in Industrial Sectors. The Role of Banking Market Structure
Nicola Cetorelli

WP-02-26

Private School Location and Neighborhood Characteristics
Lisa Barrow

WP-02-27

Teachers and Student Achievement in the Chicago Public High Schools
Daniel Aaronson, Lisa Barrow and William Sander

WP-02-28

The Crime of 1873: Back to the Scene
François R. Velde

WP-02-29

Trade Structure, Industrial Structure, and International Business Cycles
Marianne Baxter and Michael A. Kouparitsas

WP-02-30

Estimating the Returns to Community College Schooling for Displaced Workers
Louis Jacobson, Robert LaLonde and Daniel G. Sullivan

WP-02-31

A Proposal for Efficiently Resolving Out-of-the-Money Swap Positions
at Large Insolvent Banks
George G. Kaufman

WP-03-01

Depositor Liquidity and Loss-Sharing in Bank Failure Resolutions
George G. Kaufman

WP-03-02

Subordinated Debt and Prompt Corrective Regulatory Action
Douglas D. Evanoff and Larry D. Wall

WP-03-03

When is Inter-Transaction Time Informative?
Craig Furfine

WP-03-04

Tenure Choice with Location Selection: The Case of Hispanic Neighborhoods
in Chicago
Maude Toussaint-Comeau and Sherrie L.W. Rhine

WP-03-05

Distinguishing Limited Commitment from Moral Hazard in Models of
Growth with Inequality*
Anna L. Paulson and Robert Townsend

WP-03-06

Resolving Large Complex Financial Organizations
Robert R. Bliss

WP-03-07

Working Paper Series (continued)
The Case of the Missing Productivity Growth:
Or, Does information technology explain why productivity accelerated in the United States
but not the United Kingdom?
Susanto Basu, John G. Fernald, Nicholas Oulton and Sylaja Srinivasan

WP-03-08

Inside-Outside Money Competition
Ramon Marimon, Juan Pablo Nicolini and Pedro Teles

WP-03-09

The Importance of Check-Cashing Businesses to the Unbanked: Racial/Ethnic Differences
William H. Greene, Sherrie L.W. Rhine and Maude Toussaint-Comeau

WP-03-10

A Structural Empirical Model of Firm Growth, Learning, and Survival
Jaap H. Abbring and Jeffrey R. Campbell

WP-03-11

Market Size Matters
Jeffrey R. Campbell and Hugo A. Hopenhayn

WP-03-12

The Cost of Business Cycles under Endogenous Growth
Gadi Barlevy

WP-03-13

The Past, Present, and Probable Future for Community Banks
Robert DeYoung, William C. Hunter and Gregory F. Udell

WP-03-14

Measuring Productivity Growth in Asia: Do Market Imperfections Matter?
John Fernald and Brent Neiman

WP-03-15

Revised Estimates of Intergenerational Income Mobility in the United States
Bhashkar Mazumder

WP-03-16

Product Market Evidence on the Employment Effects of the Minimum Wage
Daniel Aaronson and Eric French

WP-03-17

Estimating Models of On-the-Job Search using Record Statistics
Gadi Barlevy

WP-03-18

Banking Market Conditions and Deposit Interest Rates
Richard J. Rosen

WP-03-19

Creating a National State Rainy Day Fund: A Modest Proposal to Improve Future
State Fiscal Performance
Richard Mattoon

WP-03-20

Managerial Incentive and Financial Contagion
Sujit Chakravorti, Anna Llyina and Subir Lall

WP-03-21

Women and the Phillips Curve: Do Women’s and Men’s Labor Market Outcomes
Differentially Affect Real Wage Growth and Inflation?
Katharine Anderson, Lisa Barrow and Kristin F. Butcher

WP-03-22

Evaluating the Calvo Model of Sticky Prices
Martin Eichenbaum and Jonas D.M. Fisher

WP-03-23

Full text of Working Papers (Federal Reserve Bank of Chicago) : Should We Teach Old Dogs New Tricks? : The Impact of Community College Retraining on Older Displaced Workers, Working Paper 2003-25

FRASER