Job Search Behavior of Older Americans†
Hugo Benı́tez-Silva‡
SUNY at Stony Brook
Huan Ni
Kennesaw State University
June 13, 2010
Abstract
This paper presents an empirical analysis of job search behavior among older Americans using the Health and Retirement Study. Increasing longevity, improving health conditional on
age, increasing labor supply flexibility stemming from an increase in part-time work, selfemployment and the use of technological advances to promote second careers, and increasing
labor force participation, make the study of search behavior at the end of the life cycle an important research topic. The analysis shows that older Americans actively search for new jobs,
both on-the-job and when out of work, and that this behavior is strongly correlated with the
labor market outcome of those individuals, making searchers more likely to switch jobs or find
new jobs. We also discuss the key variables that help us understand the different job search
behavior of employed and non-employed individuals, as well as across genders.
Keywords: Job Search Behavior, Binary Choice Models, Health and Retirement Study.
JEL classification: J14, I28, D0
† We are grateful to John Rust, Moshe Buchinsky, Ann Huff-Stevens, George Hall, Sergi Jiménez-Martı́n, and Sofia
Cheidvasser for their help, encouragement and comments. And to Michael Boozer, Jenny Hunt, Olivia S. Mitchell,
Caroline Austin, Giorgio Pauletto, T. Paul Schultz, Deb Dwyer, Kevin Austin, and the participants in a number of conferences and seminars for insightful comments and discussions. We also want to thank the Department of Economics
at Universitat Autònoma de Barcelona, Universitat Pompeu Fabra, and the Economics Department at the University of
Maryland for their hospitality. Wayne-Roy Gayle and Frank Heiland provided excellent research assistance. Benı́tezSilva is especially grateful to the Center for Retirement Research at Boston College that funded the first steps of this
research through a Steven H. Sandell Grant, and to project number SEJ2005-08783-C04-01 from the Spanish Ministry
of Science and Technology.
‡ Corresponding author: Economics Department, State University of New York at Stony Brook, Stony Brook, N.Y.
11794-4384, phone: (631) 632-7551, fax: (631) 632-7516, e-mail: [email protected]
1 Introduction
This paper analyzes the job search behavior of older Americans using data from the Health and
Retirement Study (HRS), presenting one of the first empirical studies of this behavior among this
population. Job search behavior among older Americans, although mentioned in the literature for
almost two decades, has been less formally modeled than, for example, retirement incentives and
their policy implications.
Several factors currently compound to make this topic an important one: increasing longevity,
improving health status conditional on age, increasing labor supply flexibility stemming from an
increase in the use of part-time work and the use of technological advances which promote second
careers, and increasing labor force participation (it has reached levels not seen since the early
1980s for males 55 to 64 and has continued to increase for females in that age group. Additionally,
for males 65 to 69 the participation rates are at the highest in over 35 years). All of these make
the study of search behavior at the end of the life cycle an important contribution, in order to
eventually achieve what a key researcher in the study of older Americans already emphasized more
than two decades ago: “If in the next century the nation is to effectively use older workers’ skills
and experience, the development of new retirement and employment policies must begin today.”
(Sandell 1987, p.p. 245).
In Sandell (1987), and also in Borus et al. (1988), the focus was on the difficulties that older
workers face in the labor market. The contributors to those volumes identified four aspects that
would help improve the situation of older individuals in the labor force: improving economic conditions; increase in labor market flexibility; investment in training and retraining; and improving
job search. More than two decades after those remarks were made, the economic conditions have
in general improved for both younger and older workers, and labor market flexibility has increased
1
substantially among older workers if we consider the increasing trend towards part-time work and
self-employment among older individuals. However, there has been relatively little improvement
in the understanding of the processes that foster job search behavior and human capital accumulation at the end of the life cycle. In this study we present the empirical evidence that backs the
assertion that job search among older Americans is an important.
The empirical analysis of search behavior at the end of the life cycle presented here represents
one of the first efforts to characterize job search as an important issue to consider for older individuals. Up until now most of the research in the Economics of Aging has not paid much attention to
search behavior since relatively few people return to work after retirement or work beyond the traditional retirement ages.1 However, with the aging of the US population, the potential shortage of
labor supply makes it extremely important to promote the return to the labor force of older Americans. Additionally, increased life expectancy and improved health, along with new technological
opportunities, allow individuals to consider second careers and to search for what researchers have
called bridge jobs or part-time jobs, as a way of phasing out of the labor force (Quinn 1998, and
Friedberg 1999a). This makes the study of the search decision of this population an interesting and
novel project. Only a few research efforts, including Hutchens (1988, 1993) and the volumes by
Sandell (1987) and Borus et al. (1988), deal directly with the issue of aging and search behavior.
However, even those researchers only concentrate on the disadvantages that older workers face to
find jobs comparable to those of younger individuals.
Turning to the search literature, the emphasis has mostly been on the empirical analysis of
1
Benı́tez-Silva (2000) in an unpublished manuscript, and using the first three waves of the HRS analyzes the role
that search plays in labor force transitions among older Americans. Similarly, Maestas and Li (2006), in a more recent
working paper, and using six waves of the HRS, focus on the issue of discouraged workers, and the effect of search
on employment transitions but without accounting for the possible endogeneity of job search in that framework, and
without analyzing the differential effect by gender.
2
young and middle-aged individuals.2 Furthermore, theoretical search models have not investigated
the implications of the searching parties’ distinct characteristic of approaching or having reached
retirement age.3
In this paper, we empirically show the importance of job search in terms of affecting labor
market outcomes, and we present the estimation results focusing on job search, using cross-section
and panel data models, which shows the importance of age, marital status, education, and especially previous work attachment and health limitations in the decision to search for a new job. We
find clear differences on the effect of these characteristics between employed or non-employed
individuals, as well as across genders.
Section 2 presents the data set we are using and some evidence of the importance of job search
among older Americans, in terms of how prevalent this behavior is among this population, and
how it affects labor market outcomes. Section 3 shows cross-section and panel data estimates
of the decision to search for a job by employed and non-employed older Americans. Section 4
concludes.
2
Lippman and McCall (1976) provide an excellent literature review of the early contributions to the economics
of job search. More recent discussion of the theory of job search with a macroeconomic emphasis can be found in
Sargent (1987), Ljungqvist and Sargent (2000), and Wright (2000). Bhattacharya, Mulligan, and Reed (2001) present
a standard model of labor market search which analyzes the effects of retirement policies that foster early retirement
on the employment of young individuals.
3 Most of the initial efforts were directed at understanding the behavior of unemployed searchers, essentially introducing a theory of voluntary unemployment (McCall 1970, Mortensen 1970, Gronau 1971, Peterson 1972, Whipple
1973, Barron 1975, and Feinberg 1977). That theory tackled the problem from different angles (firms, workers, institutions), and has eventually evolved into the equilibrium models of the labor market that the growing search literature
builds upon (Jovanovic 1979, Albrecht and Axell 1984, Burdett and Mortensen 1998, Van den Berg and Ridder 1998).
The empirical research has been instrumental in making job search theory a topic of growing interest, as shown by
the importance of the work by Kiefer and Neumann (1979), Sandell (1980a, 1980b), Wolpin (1987), Stern (1989),
Eckstein and Wolpin (1990), and the empirical literature reviewed in Devine and Kiefer (1991), and Van den Berg
(1999). Researchers have also recognized the importance of on-the-job search, and have studied it in order to better
understand the evidence on turnover, quit behavior, and job mobility (Black 1981, Miller 1984, Parsons 1991, Topel
and Ward 1992, Neal 1999). Also, the debate on whether individuals actually quit in order to search for jobs, and
whether this is an optimizing behavior has also fostered an interesting body of work with mixed empirical results
(Barron and McCafferty 1977, Kahn and Low 1982, Holzer 1987, and Blau and Robins 1990) Finally, Seater (1977)
was the first to set up a unified model of consumption, labor supply, and job search, a model that interestingly closely
relates to the human capital literature, for example in Ben-Porath (1967) and Heckman (1976a, 1976b).
3
2 Data, Importance of Job Search, and Summary Statistics
The HRS is a nationally representative longitudinal survey of 7,700 households headed by an
individual aged 51 to 61 as of the first round of interviews in 1992-93. The primary purpose of
the HRS is to study the labor force transitions with particular emphasis on sources of retirement
income and health care needs. It is a survey conducted by the Survey Research Center (SRC) at
the University of Michigan and funded by the National Institute on Aging. 4 In this study we use
the first six waves of the survey. The last five of those waves were conducted by phone using the
computer assisted technology (CATI) which allows for much better control of the skip patterns and
reduces recall errors.
Figure 1 presents the percentage of individuals who said they were searching for a job by age.
We consider employees and self-employed individuals to be searching for a job if they answered
yes to the question: Are you currently looking for another job? For non-employed respondents
they answered yes to the question: Have you been doing anything to find work during the last
four weeks? The same graph also shows the responses of the employed and the non-employed.
Given the nature of the HRS we essentially have data for individuals between the ages of 50 and
approximately 78.5 Notice the relatively large percentage of individuals in their fifties who are still
actively searching for jobs, a fact that motivates further investigation on the nature of this activity
by older Americans.
Figures 2 and 3 present the information about search behavior for non-employed and employed
individuals using the HRS and show that job search is undertaken by a non-trivial proportion of
4
For more detailed information on the HRS see Juster and Suzman (1995), and Gustman et al. (1995).
There are a few individuals in the sample older than 75, but they were not searching for a job. Since the HRS was
supposed to be representative of the U.S. population of individuals 51 to 61 as of the first round of interviews (in the
field during 1992 and the beginning of 1993) we do not really lose much information by eliminating those respondents
from our sample.
5
4
individuals in the sample, and that their responses to a variety of questions regarding job search
show that they are active job seekers.
Figure 2 shows that 4.5% of non-employed individuals were searching for a job in the month
before the interview, and that almost half of those were searching exclusively for a full-time job,
and around 32% only wanted a part-time job. Both types of searchers rely quite heavily on direct
contacts to try to find a job, and full-time job seekers use employment agencies and informal
channels more frequently. The chart also shows that among non-searchers a non-trivial proportion
actually wants a job, and that almost 60% of them want a full-time job.
Figure 3 presents a similar breakdown for employed workers. Employed individuals search in
a higher proportion, 8.8%, and almost 2 out of 3 of them were searching for a full-time job. Both
part-time and full-time job searchers rely heavily on informal channels to search for their new job.
Among non-searchers, almost 70% said they would not consider other jobs, mainly because of
their fear of losing their pension and health insurance benefits.
However, while we have argued that job search is an important behavior among older Americans, whether it is worth further studying is in part a function of whether it actually matters for the
individuals in the population analyzed. If job search was not necessarily linked with labor market
outcomes it is unclear whether a study of job search would be of real economic importance. 6 Table
1, shows the strong correlation between labor market outcomes and job search behavior. We can
see that both on-the-job search and search from out of work are very effective in the sense that job
searchers are much more likely to change jobs or find new jobs. In fact among those who search
on-the-job in a given wave of the sample, one out of three individuals (33%) changes employers
in the subsequent wave, while the change only happens in one out of 9 (11%) for non-searchers.
6
We build upon the discussions in Benı́tez-Silva (2000), who in an unpublished manuscript, and using the first
three waves of the HRS analyzes the role that search plays in labor force transitions.
5
The difference is more striking among those who are out of work. 48% of those who did not have
a job but were searching had a job come the next wave of the sample, while this proportion was
below 7% among those who did not search for a job. This unconditional evidence points to a very
important role of job search in the labor force transitions of older Americans. Notice as well, that
a similar proportion of individuals search on-the-job and when out of work, suggesting that this
behavior is worth analyzing for both sub-samples of the population.
Table 1: Job Search and Labor Supply Transition
Transition in Labor Force
Non-searcher P(trans Cond. Job searcher
non-searcher)
Employed (last wave)
Same employer (next wave)
18,773
88.98%
1,349
Changed employer
2,325
11.02%
671
Sub total
21,098
2,020
% of all employees
91.26%
8.74%
Non-employed (last wave)
Non-employed (next wave)
25,952
93.32%
652
Employed (next wave)
1,857
6.68%
604
Sub total
27,809
1,256
% of all non-employed
95.68%
4.32%
P(trans Cond.
searcher)
Total
66.78%
33.22%
20,122
2,996
23,118
51.91%
48.09%
26,604
2,461
29,065
Tables A1 and A2 in the Appendix present the multivariate analysis that supports the discussions resulting from the correlations presented in Table 1. The Appendix presents the analysis
in the framework of a bivariate probit model with a structural shift, in line with the celebrated
work by Heckman (1978) to take into account the possibility that the job search indicator might
be endogenous in the job transition equations for those employed and those out of work. We also
show our test on the exogeneity of the job search indicator, and find that it can indeed be taken
as exogenous. In Tables A1 and A2 of job transition equations, job search is a highly significant
and strong predictor of the transitions to new jobs and to employment even after controlling for
a wide array of characteristics and conditions. The second part of this latter result is interpreted
as indicating that the distinction between unemployment and out of the labor force is behaviorally
6
meaningful among older individuals, extending the result of Flinn and Heckman (1983) who used
data on young men. Also, the result is consistent with the findings of Barron and Mellow (1981),
who, using the CPS, find that more search intensity leads to a higher probability of finding a job
for non-employed individuals. The first part of the result comes to emphasize the importance of
on-the-job search among older Americans.
Summary Statistics
The HRS provides researchers with a large array of socio-economic and demographic variables,
health indicators, ADLs, IADLs, and even some variables that measure expectations. Table 2,
presents an exploratory analysis of the data. If we compare searchers with our full sample of
respondents, we can observe that searchers are more likely to be younger, male, and with a slightly
higher level of education, and less likely to be white and married. They are likely to have a much
lower level of net worth and housing wealth, but a bit higher level of income in the previous year.
They are also less likely to have health insurance, and less likely to be receiving Medicare or
Medicaid. Searchers are in overall better general health by objective and subjective measures, and
are less likely to have health limitations, but are more likely to be smokers and moderate drinkers.
It is important, however, to make a clear distinction between those searching on-the-job and
non-employed searchers. Compared with those searching on-the-job, non-employed searchers are
older, less likely to be male, married, and white, and have less years of education. They are on
average wealthier than their employed counterparts, but with lower income levels. Non-employed
searchers are also less likely to have any type of health insurance, but they rely more on Medicare
and Medicaid. Finally, non-employed searchers are in overall worse health, as measured by both
self-reported indicators and more objective ones. Additionally, a higher proportion of them has
health limitations compared with those searching on-the-job.
7
3 Cross-Section and Panel Data Estimates
The objective of this section is to find the determinants of the job search decision among employed
and non-employed older Americans in a multivariate analysis framework. Given the discussion in
the previous section, one of the main motivations for this exercise is the need to better understand
the labor force transitions of older Americans, given the role of job search in those outcomes.
We will use both cross-section and panel data specifications of a binary choice model, where
the dependent variable is the decision to search for a job. The usual latent variable model applies
convincingly in this case since we will assume that, although there can be many levels of job search
intensity, an individual reports him or herself searching for a job only if he (she) has done enough
to be considered a job searcher. That a particular person considers him or herself a job searcher
will usually be linked to a given action, such as inquiring about a job, responding to an advertisement in the newspaper, or even spreading the word among a certain network of individuals. The
threshold of what it means to be an active job searcher can potentially be very different depending
on the individual. We will be able to take into account some of this (observed and unobserved)
heterogeneity through the multivariate specifications presented below.
Tables 3 to 8 show the results of the empirical analysis. All of them have a similar structure, the
first three columns of each table show the Maximum Likelihood estimates of a Probit model with
their standard errors and marginal effects. The last three columns show the Maximum Likelihood
estimates of a Random Effects Probit model with standard errors and marginal effects, where we
exploit the longitudinal nature of the data to control for unobserved heterogeneity.
It is well known that using the standard probit model with pooled data produces consistent but
inefficient estimates (Maddala 1987). However, as shown by Guilkey and Murphy (1993), in finite
samples, and if the number of time periods is larger than two, the standard probit model performs
8
quite poorly compared with the Random Effects probit model. For all the models presented below,
a Likelihood Ratio test, comparing the pooled probit estimator and the random effects probit estimator, always rejected the pooled estimator in favor of the panel characterization. The LR tests
whether the panel variance component is of relevance in the model.
The Probit estimates are the result of fitting the following standard Probit model:
α
Si
β Xi
vi (1)
where S is a dichotomous variable indicating whether the individual had searched for a job in the
month preceding the interview, if out of work, or whether he was currently searching for a job, if
employed. X is a vector of exogenous explanatory variables including a constant, β is a vector of
coefficients, and v is a normally distributed disturbance with mean zero and variance σ vv .
The Random Effects Probit model adds an individual random effect to the equation above, and
takes into account the multi-period nature of the data. For a given individual i (i
time period t (t
1 N) and
1 T ) the model can be written as follows:
Sit
α
β Xit
ui
vit (2)
where ui is the random individual specific effect, which is normally distributed with mean zero
and variance σuu . The remaining error term, vit , representing unobserved individual characteristics
that vary with time, is also assumed normal with mean zero and variance σ vv . We will take ui ,
and vit to be independent of each other and of the Xit explanatory variables. We integrate out the
unobservable component using Gauss-Hermite quadrature with 12 nodes. Increases in the number
of nodes had basically no effect on the estimates shown below. The maximum number of time
periods we observe in the data is six.
Table A.3. in the Appendix shows the results using the full sample of individuals, and con9
trolling for an employee indicator, which is positive and highly significant in the cross-sectional
estimation but not significant in the panel specification, suggesting that eventhough there is a much
higher proportion of searchers among the employed, once we control for an array of characteristics and unobserved heterogeneity, the difference is not significant. The rest of the tables in this
section show results for the search decision of employed and unemployed individuals, as well as
by gender, separately, to better characterize the difference between these two populations.
Table 3 presents estimates, standard errors, and marginal effects, of the decision to search
for a job by the sample of employed individuals.7 Being married has a fairly large negative and
significant effect, we will later see that this is driven mainly by the effect of marriage on women’s
decisions. Males are significantly more likely to search for jobs, and individuals with more years
of education are also more likely to search for jobs. Whites, other things equal, are less likely to
search for jobs, and age has the expected negative effect, which is consistent with the notion that
as individuals have less years of active work ahead of them they are less likely to want to incur job
search costs.8
Net worth has a negative and significant effect on the probability of searching for a job, which
is consistent across specifications, and also consistent with a theoretical framework where a higher
level of individual wealth allows a person to maintain a higher consumption level and reduces
his incentives for job search. The latter argument is similar to the result in a search model like
that of Sargent (1987), where higher financial wealth translates into a higher reservation wage, or
7
Following the recommendations of an anonymous referee we have removed from the estimation a number of
variables linked to the individuals’ work histories, given that they could be considered as endogenous to the search
decision, and hard to instrument in this setting.
8 The introduction of demographic variables is justified by the fact that we want to control for this type of characteristic to better assess the effect of the other variables of interest. Notice that the age effect is almost monotonic, with
the exception of the indicator for age 65 to 69, which has a smaller effect than for those age 63 to 64, probably due to
the fact that for the former age group the earnings test in this period was either lower or did not apply, making future
work comparatively more appealing at the margin.
10
a more detailed model like that of Lentz and Tranaes (2000). The same negative effect can be
seen for the income variable, with higher income decreasing the probability of job search among
the employed. This negative effect is consistent with the theoretical characterizations of Gronau
(1971) which justifies it with the intuitive argument that current earnings are a proxy for the cost
of search in the same way that they are a proxy for the cost of leisure. Stigler (1962) argues that
employment agencies might charge proportionally higher fees the higher the earnings of the job
searcher.
Lack of access to health insurance has a large positive and significant marginal effect on the
probability of searching for a job among this sample of individuals, a result consistent with the
large literature that emphasizes the importance of health insurance among older people (Currie
and Madrian 1999). Having private health insurance also has a positive and significant effect (especially in the panel specification) but the effect is much smaller. Another relevant variable that we
introduce in the tradition of Stigler (1962) and McCall (1970) is a proxy for the expected period of
employment, the self-reported probability of living to age 85. We would expect it to be positively
correlated with the likelihood of searching for a job and, the results confirm that conjecture. However, the results are not very significant, probably suggesting that the age indicators capture most
of this effect.
We also control for a variety of objective health measures, with only one of them appearing
as significant for employed searchers. A binary indicator for self-reported psychological problems
has a very large positive effect on the probability of searching for a job, which might reflect that
some of these problems are likely to be work related (e.g. stress, bad working environment).
A dummy variable indicating whether the individual has a health limitation for work has a
positive effect on the decision to search for a job, but it is not significant. Interestingly it will be
11
shown below that the effect is highly significant and negative for non-employed searchers. Selfreported indicators of general health show that those in worse health are more likely to search for
new jobs (in the panel specification), but those with objective health conditions are less likely to
search for new jobs. However, these results are not significant. 9
Finally, indicators of the last three interview periods, 1998 to 2002, have a negative (but significant only for 2000) effect on the probability of searching for a job, something expected given
the excellent labor market conditions of the late 1990s, and the substantial increase in labor force
participation among individuals 55 to 64 (and also for individuals 65 and over) in the same period,
as discussed in Benı́tez-Silva (2000) and Burtless and Quinn (2000, 2002).
We can see in this table, and those that follow, that the cross-sectional and panel models provide
similar results. In general the preferred panel specification provides stronger effects, once we
account for the predicted probability of the job search outcome, and in some cases, effects that
lack significance in the cross-sectional specification become so once unobserved heterogeneity is
accounted for.
Table 4 shows the estimates of the search decision for non-employed individuals. The first
clear difference with Table 3 is the positive, significant, and comparatively large, effect of income,
probably proxying for the work attachment of those currently out of work. The negative effect
of net wealth is in line with that of employed individuals. Among non-employed searchers the
importance of health insurance is also very strong, with those without insurance becoming very
likely to search for jobs. This is not very surprising given the importance of health insurance for
9
See Benı́tez-Silva et al. (2004) for a discussion of the relationship between self-reported health limitation and
disability status. In the case of job changes, there is the complex issue of the possible role that pre-existing conditions
can play in the possibility of finding a job with good health insurance in the presence of health conditions or limitations. This could probably help explain the negative effect of these measures among the non-employed, but also the
different effect that health limitations and health conditions can have, as in Table 3, given the strict laws preventing
discrimination against individuals with disabilities.
12
older Americans before they reach age 65 when they can receive Medicare.
Another important difference with the sample of employed individuals is in the variable that
proxies for the expected period of employment, the self-reported probability of living to age 85.
For non-employed individuals, it is significant and it has a large effect. Yet another difference is
the negative effect of having a health limitation for work on the probability of searching for a job,
meaning that those with some kind of disability are less likely to try to come back to the labor force.
This negative effect is in line with the effects of self-reported measures of health, as well as the
mobility limitation index. Interestingly, having psychological problems still has a positive effect
on the likelihood of searching for a job, maybe linked to the fact that those problems (possibly
some type of anxiety or depression) could be the result of the stresses of being out of work when
work is needed to maintain a given standard of living. Finally, notice the much better explanatory
power of this model compared with the model for on-the-job search.
Tables 5 and 6 separately estimate the on-the-job search decision for males and females. One
very interesting result is the asymmetric effect of marriage for these two populations. For males,
marriage has a small and positive, effect on the decision to search for a new job, but for females, it
has a large, negative, and highly significant effect on the decision to search for new jobs. This result
might be a hint of a behavioral difference between males and females regarding their approach to
on-the-job search, with females being more ‘loyal’ to their employers, maybe due to a better job
match or due to different attitudes towards work. Parsons (1991) finds the same asymmetry in
his study of young adults from the NLSY in the early 1980s. Also, notice the clear effect of
income (mostly earnings from work in the previous year), which is negative and significant for
both genders. On top of the interpretations we discussed for this result when analyzing Table 3, it
is also possible that higher earnings are, for women, a proxy for attachment to a given job or the
13
labor market, or the quality of the job match. The effect of lack of insurance is also similar for
both genders, with a large positive effect on on-the-job search. Interestingly, the effect of having
private health insurance differs across genders; while for males has a negative and insignificant
effect on the job search decision, for females has a significantly positive effect on job search. This
could be capturing the fact that employed women might be more likely to hold types of jobs that
require them to buy their own coverage, making them more likely to keep searching for a better
job match. It is clear from these results that the separate analysis provides an interesting picture of
the different on-the-job search behavior of the two genders.
Finally, Tables 7 and 8, show the Probit and Random Effects Probit Maximum Likelihood
estimates for non-employed males and females. Again, we can observe the asymmetric effect of
marriage among these two populations, from being an insignificant positive regressor for males,
to having a fairly large negative and very significant effect for females. Net wealth also shows a
different effect across genders; while for males wealth plays a very small and insignificant role, the
effect is negative and significant for females, indicating the differential effect of family resources
and the role of females in the household. The psychological problems indicator also has a very
different effect on non-employed job search for males and females. While for males the effect
is negative, small, and insignificant, the effect is positive and significant for females, which may
indicate the different way that women react to those types of problems. It is worth noticing the
interesting differences in the size of the effects of the mobility and muscular difficulties indices,
as well as the health limitation indicator, with large and significant negative effects for males, but
much smaller and in some cases insignificant effects for females, probably due to the different
types of jobs men and women have done.
14
4 Conclusions
This paper has shown the importance of analyzing the job search behavior of older Americans.
Using the HRS we have first motivated the analysis by showing that job search is highly correlated
with labor force transitions, both unconditionally and in a bivariate model of employment transitions and job search. Then using cross-section and panel data models we have characterized the
decision to search for a job by employed and non-employed individuals, emphasizing the importance of age, wealth, income , health insurance, and health indicators. Special attention is paid
to the sources of the differences between employed and non-employed older Americans and also
between males and females in both groups.
Interestingly, this analysis of job search behavior at the end of the life cycle, represents one of
the few and more recent efforts we are aware of to characterize job search as an important issue to
consider for older individuals. Previously most of the research in the Economics of Aging has not
paid much attention to search behavior. However, increased life expectancy, improved health conditional on age, and new technological opportunities allow individuals to consider second careers
and to search for more flexible jobs.
There are a number of extensions of this study worth considering as avenues of future research.
Firstly, job search can be extended (eventhough data limitations have prevented us from doing so)
to have more than two states, allowing us to discuss not only the fact that an individual searches
for a job but also the intensity of that search, with probably different effects in terms of wage
rewards and lower disutility of labor. Secondly, a separate analysis of job search strategies by
older Americans can provide useful policy insights regarding labor force participation in a labor
market environment that will consistently contain a higher proportion of older workers.
Thirdly, it is probably worth studying job search behavior of older Americans within a theoret15
ical framework to analyze human capital formation at older ages. The links of such a model with
the work of Ben-Porath (1967) and Heckman (1976a, 1976b) were already emphasized by Seater
(1977). In the same way that older workers are responsive both to the cost of job search and the
rewards resulting from that investment of time and tangible resources, they are also likely to be responsive to the cost and rewards of human capital investment. Therefore, a framework of analysis
can be developed to assess public policies that can affect the costs of human capital investment later
in life, or the rewards from it. For example, policies that facilitate adult education to allow older
individuals to keep up with technological change will make it easier for older workers to be more
active in the labor market and eventually work longer. This is especially true in a future economic
environment that might need those workers to ease a labor shortage resulting from the current demographic trends, which if not covered by immigration, could ignite inflationary pressures. This
increased labor force participation (which has continued to increase in the last couple of years
in spite of the recession) could also ease the pressures on a struggling social insurance system. 10
McCall (1970) discusses the trade-offs between policies that facilitate job search through lowering
information costs, and policies that promote training in the context of reducing the proportion of
discouraged workers. Whipple (1973) introduces the concept of Skill Maintenance Clinics as a
way of avoiding the depreciation of skills that can negatively affect the employability of individuals and the efficacy of job search. Mortersen (1970) discusses the possibility of using retraining
programs to reduce the duration of search, which can result in lower unemployment among a given
population. The study of these substitutabilities and complementarities between search and training among older workers is likely to become an important policy research issue.
The analysis we have presented here can be considered an empirical effort in the direction of
10
See Friedberg (1999b) for a discussion of investment incentives in computer skills among older workers. See
Burtless and Quinn (2000) for a discussion of policies to encourage labor force participation of older Americans.
16
eventually solving and estimating an empirical model of consumption, savings, labor supply and
job search of older Americans.
17
Figure 1: Search in the Health and Retirement Survey
Percent of Individuals Searching
Search in the HRS
20.00%
15.00%
10.00%
5.00%
0.00%
50
54
58
62
Age
Employed
Un-employed
18
66
70
74
Figure 2: Search By Non-Employed Respondents
NONEMPLOYED
4.5%
95.5%
NOT
SEARCHING
SEARCHING
32.0%
27.3%
72.7%
WANT JOB
DO NOT
WANT JOB
46.7%
FULL-TIME
PART-TIME
51.2%
18.6%
63.6%
62.6%
40.4%
71.8%
DC
EA
INF
DC
EA
INF
Search Strategies
DC-Direct Contact
EA-Employment Agency
INF-Informal
19
36.6%
56.6%
PART-TIME
FULL-TIME
Figure 3: Search By Employed Respondents
EMPLOYED
8.8%
91.2%
NOT
SEARCHING
SEARCHING
23.8%
67.3%
PART-TIME
31.8%
WOULD
CONSIDER
OTHER JOB
FULL-TIME
10.2%
19.7%
44.2%
17.8%
24.9%
51.7%
DC
EA
INF
DC
EA
INF
Search Strategies:
DC-Direct Contact
EA-Employment Agency
INF-Informal
20
68.2%
WOULD NOT
CONSIDER
OTHER JOB
21.5%
52.0%
LIKE
CURRENT
JOB
FEAR
LOSS OF
PEN/H.I.
Table 2: Sample Statistics
Full
Searchers
Employed
Sample
Sub-sample Searchers
N
56,296
3,896
2,656
age
59.6526
56.2323
55.6476
( 6.3362 )
( 5.5683 )
( 5.4057 )
white
0.8171
0.7762
0.7839
( 0.3866 )
( 0.4169 )
( 0.4117 )
male
0.4204
0.4725
0.4718
( 0.4936 )
( 0.4993 )
( 0.4993 )
married
0.7481
0.7097
0.7188
( 0.4341 )
( 0.4540 )
( 0.4497 )
Yrs of Education
12.4736
12.8013
13.1318
( 3.0102 )
( 3.1149 )
( 2.8964 )
Total net Wealth
5.0061
2.4439
2.0421
(in $100,000 of 1992) ( 23.0213 ) ( 13.5123 ) ( 11.5313 )
Housing Wealth
20.5926
16.5690
13.5988
(in $100,000 of 1992) ( 101.8631 ) ( 91.9365 ) ( 75.0987 )
Resp. Income
2.0625
2.2490
2.5151
(in $10,000 of 1992)
( 3.1013 )
( 2.5367 )
( 2.6233 )
Employee
0.4703
0.6016
0.8812
( 0.4991 )
( 0.4896 )
( 0.3236 )
Self employed
0.0928
0.0811
0.1188
( 0.2902 )
( 0.2731 )
( 0.3236 )
Non-employed
0.4368
0.3173
0
( 0.4960 )
( 0.4655 )
(0)
No Health Ins.
0.0852
0.2082
0.1645
( 0.2791 )
( 0.4060 )
( 0.3708 )
Gov. Health Ins.
0.3083
0.1363
0.1003
( 0.4618 )
( 0.3431 )
( 0.3005 )
Employer Health Ins.
0.6634
0.6363
0.7127
( 0.4725 )
( 0.4811 )
( 0.4526 )
Private Health Ins.
0.1269
0.1140
0.1130
( 0.3328 )
( 0.3178 )
( 0.3166 )
Health limit
0.3060
0.2400
0.2044
( 0.4608 )
( 0.4271 )
( 0.4034 )
0.2322
0.1585
0.1127
Health limit
to paid work
( 0.4222 )
( 0.3652 )
( 0.3162 )
Variable
21
Non-Employed
Searchers
1,240
57.4847
( 5.7054 )
0.7597
( 0.4275 )
0.4742
( 0.4995 )
0.6903
( 0.4625 )
12.0936
( 3.4336 )
3.3046
( 16.9693 )
23.6575
( 122.9144 )
1.6790
( 2.2372 )
0
(0)
0
(0)
1
(0)
0.3016
( 0.4591 )
0.2132
( 0.4098 )
0.4726
( 0.4994 )
0.1161
( 0.3205 )
0.3161
( 0.4652 )
0.2567
( 0.4370 )
Table 2: Sample Statistics (continued)
Full
Searchers Employed Non-Employed
Sample Sub-sample Searchers
Searchers
N
56,296
3,896
2,656
1,240
High blood pressure
0.4381
0.3858
0.3648
0.4306
( 0.4962 ) ( 0.4868 ) ( 0.4815 )
( 0.4954 )
Diabetes
0.1322
0.0991
0.0926
0.1129
( 0.3387 ) ( 0.2988 ) ( 0.2900 )
( 0.3166 )
Heart Problems
0.1639
0.1194
0.1148
0.1290
( 0.3702 ) ( 0.3242 ) ( 0.3189 )
( 0.3354 )
Stroke
0.0383
0.0221
0.0162
0.0347
( 0.1919 ) ( 0.1470 ) ( 0.1263 )
( 0.1832 )
Cancer
0.0834
0.0547
0.0516
0.0613
( 0.2765 ) ( 0.2274 ) ( 0.2213 )
( 0.2400 )
Arthritis
0.4934
0.4053
0.3859
0.4468
( 0.5000 ) ( 0.4910 ) ( 0.4869 )
( 0.4974 )
ADL mobility index
0.1599
0.1232
0.1108
0.1498
( 0.2484 ) ( 0.2003 ) ( 0.1837 )
( 0.2296 )
ADL muscular index
0.2374
0.2017
0.1886
0.2297
( 0.2849 ) ( 0.2556 ) ( 0.2415 )
( 0.2815 )
Prob. Living to 85
0.4735
0.4767
0.4734
0.4837
( 0.3185 ) ( 0.3263 ) ( 0.3222 )
( 0.3348 )
Short term memory
0.5265
0.5116
0.5234
0.4864
( 0.1895 ) ( 0.1944 ) ( 0.1919 )
( 0.1974 )
Long term memory
0.4310
0.4174
0.4284
0.3938
( 0.2138 ) ( 0.2156 ) ( 0.2166 )
( 0.2116 )
Psych. Test
0.1365
0.1514
0.1427
0.1702
( 0.3433 ) ( 0.3585 ) ( 0.3498 )
( 0.3759 )
Excellent Health
0.1762
0.2059
0.2203
0.1750
( 0.3810 ) ( 0.4044 ) ( 0.4145 )
( 0.3801 )
Very Good Health
0.3162
0.3211
0.3471
0.2653
( 0.4650 ) ( 0.4670 ) ( 0.4761 )
( 0.4417 )
Good Health
0.2953
0.3116
0.2997
0.3371
( 0.4562 ) ( 0.4632 ) ( 0.4582 )
( 0.4729 )
Fair Health
0.1495
0.1273
0.1118
0.1605
( 0.3566 ) ( 0.3334 ) ( 0.3152 )
( 0.3672 )
Poor Health
0.0628
0.0341
0.0211
0.0621
( 0.2426 ) ( 0.1816 ) ( 0.1437 )
( 0.2414 )
Smoker
0.2125
0.2749
0.2616
0.3035
( 0.4091 ) ( 0.4465 ) ( 0.4396 )
( 0.4599 )
Drinker
0.5450
0.6127
0.6284
0.5790
( 0.4980 ) ( 0.4872 ) ( 0.4833 )
( 0.4939 )
Variable
22
23
Table 3: Estimates of the Search Decision for Employed Individuals
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.6546
0.0843
–
-2.0745
0.1065
–
-0.0922
0.0348
-0.0141
-0.1301
0.0442
-0.0104
White
0.2594
0.0312
0.0391
0.3000
0.0387
0.0232
Male
-0.1115
0.0312
-0.0170
-0.1236
0.0389
-0.0097
Married
0.0462
0.0056
0.0068
0.0543
0.0068
0.0040
Yrs of Edu
-0.1797
0.0304
-0.0245
-0.2208
0.0382
-0.0146
Age 55-57
-0.2649
0.0360
-0.0340
-0.3271
0.0456
-0.0197
Age 58-59
-0.3421
0.0392
-0.0420
-0.4471
0.0500
-0.0248
Age 60-61
-0.5277
0.0633
-0.0547
-0.6756
0.0819
-0.0291
Age 62
-0.6085
0.0593
-0.0612
-0.8014
0.0742
-0.0327
Age 63-64
-0.5941
0.0654
-0.0599
-0.7685
0.0816
-0.0317
Age 65-69
-0.6956
0.1181
-0.0626
-0.8565
0.1419
-0.0307
Age 70-74
-0.0462
0.0084
-0.0068
-0.0524
0.0081
-0.0039
Resp Income
-0.0026
0.0009
-0.0004
-0.0029
0.0011
-0.0002
Net Wealth
0.5091
0.0419
0.0997
0.6055
0.0520
0.0713
No Health Ins
0.0847
0.0405
0.0130
0.1113
0.0485
0.0089
Priv. Health Ins
0.0771
0.0430
0.0113
0.0778
0.0515
0.0058
Prob. Liv. to 85
-0.0364
0.0283
-0.0053
-0.0399
0.0351
-0.0029
Hypertension
-0.0356
0.0484
-0.0051
-0.0888
0.0576
-0.0061
Diabetes
0.2317
0.0421
0.0386
0.3013
0.0513
0.0278
Psych. Prob.
-0.0548
0.0283
-0.0080
-0.0609
0.0355
-0.0045
Arthritis
0.0770
0.0842
0.0113
0.1100
0.0997
0.0081
Mobility Diff. Index
0.0683
0.0665
0.0100
0.0800
0.0815
0.0059
Muscular Diff. Index
0.0361
0.0351
0.0054
0.0464
0.0444
0.0035
Health limit
0.0227
0.0432
0.0034
0.0161
0.0523
0.0012
Fair Health
-0.0013
0.0944
-0.0002
0.0313
0.1184
0.0024
Poor Health
0.0015
0.0357
0.0002
-0.0208
0.0460
-0.0015
Fourth Wave
-0.0492
0.0323
-0.0070
-0.0751
0.0411
-0.0053
Fifth Wave
0.0102
0.0371
0.0015
-0.0122
0.0470
-0.0009
Sixth Wave
0.0832
-7,197.46 26,498
0.0406
Log L/Obs./Avg. Prob. -7,530.91 26,498
0.0485
0.0419
Pseudo-R2
24
Table 4: Estimates of the Search Decision for Non-Employed Individuals
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.1095
0.1012
–
-1.3224
0.1119
–
-0.1209
0.0440
-0.0083
-0.1416
0.0489
-0.0048
White
0.3043
0.0382
0.0210
0.3615
0.0418
0.0126
Male
-0.2085
0.0383
-0.0146
-0.2401
0.0435
-0.0085
Married
0.0101
0.0061
0.0006
0.0142
0.0065
0.0004
Yrs of Edu
-0.0899
0.0465
-0.0054
-0.0878
0.0556
-0.0025
Age 55-57
-0.2442
0.0514
-0.0129
-0.2889
0.0623
-0.0069
Age 58-59
-0.3799
0.0523
-0.0185
-0.4392
0.0609
-0.0095
Age 60-61
-0.6611
0.0725
-0.0250
-0.7953
0.0853
-0.0122
Age 62
-0.8309
0.0613
-0.0311
-1.0013
0.0729
-0.0156
Age 63-64
-0.8557
0.0599
-0.0366
-1.0284
0.0685
-0.0193
Age 65-69
-1.0908
0.0850
-0.0366
-1.3081
0.0979
-0.0183
Age 70-74
0.0549
0.0109
0.0035
0.0648
0.0073
0.0020
Resp Income
-0.0024
0.0010
-0.0002
-0.0028
0.0010
-0.0001
Net Wealth
0.7317
0.0431
0.0823
0.8514
0.0505
0.0591
No Health Ins
0.0513
0.0472
0.0034
0.0677
0.0552
0.0022
Priv. Health Ins
0.1038
0.0506
0.0066
0.1213
0.0591
0.0038
Prob. Liv. to 85
-0.0085
0.0350
-0.0005
-0.0229
0.0394
-0.0007
Hypertension
-0.1215
0.0496
-0.0072
-0.1471
0.0574
-0.0041
Diabetes
0.0899
0.0461
0.0061
0.1084
0.0523
0.0036
Psych. Prob.
-0.0194
0.0353
-0.0012
-0.0197
0.0405
-0.0006
Arthritis
-0.2552
0.0901
-0.0163
-0.3104
0.0996
-0.0096
Mobility Diff. Index
-0.1286
0.0808
-0.0082
-0.1792
0.0891
-0.0055
Muscular Diff. Index
-0.1930
0.0451
-0.0122
-0.2185
0.0489
-0.0067
Health limit
-0.0981
0.0502
-0.0059
-0.1272
0.0550
-0.0036
Fair Health
-0.3005
0.0743
-0.0155
-0.3697
0.0812
-0.0085
Poor Health
-0.2297
0.0526
-0.0123
-0.2487
0.0630
-0.0062
Fourth Wave
-0.2877
0.0440
-0.0161
-0.3379
0.0514
-0.0088
Fifth Wave
-0.1491
0.0431
-0.0088
-0.1840
0.0515
-0.0051
Sixth Wave
0.0833
-4021.02 25,248
0.0613
Log L/Obs./Avg. Prob. -4135.84 25,248
0.1658
0.1632
Pseudo-R2
25
Table 5: Estimates of the Search Decision for Employed Males
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.8075
0.1237
–
-2.3274
0.1632
–
-0.0600
0.0567
-0.0094
-0.1023
0.0742
-0.0075
White
0.1159
0.0580
0.0167
0.1694
0.0721
0.0105
Married
0.0573
0.0079
0.0087
0.0688
0.0100
0.0047
Yrs of Edu
-0.1237
0.0473
-0.0180
-0.1632
0.0612
-0.0104
Age 55-57
-0.1817
0.0551
-0.0254
-0.2418
0.0706
-0.0144
Age 58-59
-0.2513
0.0590
-0.0338
-0.3515
0.0752
-0.0194
Age 60-61
-0.4699
0.0873
-0.0532
-0.6307
0.1158
-0.0263
Age 62
-0.5911
0.0857
-0.0636
-0.8127
0.1084
-0.0311
Age 63-64
-0.5350
0.0898
-0.0594
-0.7481
0.1140
-0.0298
Age 65-69
-0.6703
0.1591
-0.0647
-0.8364
0.1930
-0.0285
Age 70-74
-0.0507
0.0106
-0.0077
-0.0604
0.0107
-0.0042
Resp Income
-0.0028
0.0016
-0.0004
-0.0038
0.0019
-0.0003
Net Wealth
0.5804
0.0690
0.1225
0.7120
0.0882
0.0868
No Health Ins
-0.0047
0.0653
-0.0007
-0.0052
0.0831
-0.0004
Priv. Health Ins
0.0780
0.0686
0.0119
0.0547
0.0809
0.0038
Prob. Liv. to 85
-0.0578
0.0422
-0.0087
-0.0668
0.0541
-0.0046
Hypertension
0.0097
0.0697
0.0015
-0.0461
0.0838
-0.0031
Diabetes
0.2305
0.0795
0.0401
0.2942
0.0937
0.0257
Psych. Prob.
-0.0400
0.0429
-0.0060
-0.0325
0.0564
-0.0022
Arthritis
0.1728
0.1309
0.0263
0.2171
0.1669
0.0150
Mobility Diff. Index
0.0742
0.1079
0.0113
0.1152
0.1405
0.0079
Muscular Diff. Index
0.0025
0.0536
0.0004
0.0113
0.0695
0.0008
Health limit
0.1161
0.0618
0.0188
0.1525
0.0800
0.0118
Fair Health
-0.0035
0.1398
-0.0005
-0.0076
0.1792
-0.0005
Poor Health
0.0582
0.0530
0.0091
0.0553
0.0729
0.0040
Fourth Wave
-0.0561
0.0486
-0.0083
-0.0754
0.0648
-0.0050
Fifth Wave
0.0076
0.0568
0.0012
-0.0133
0.0760
-0.0009
Sixth Wave
0.1067
-3150.20 11,359
0.0521
Log L/Obs./Avg. Prob. -3348.72 11,359
0.0495
0.0430
Pseudo-R2
26
Table 6: Estimates of the Search Decision for Employed Females
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.3611
0.1128
–
-1.6764
0.1396
–
-0.1051
0.0445
-0.0154
-0.1349
0.0540
-0.0115
White
-0.2415
0.0394
-0.0360
-0.2793
0.0465
-0.0243
Married
0.0351
0.0080
0.0049
0.0401
0.0094
0.0032
Yrs of Edu
-0.2299
0.0404
-0.0293
-0.2699
0.0490
-0.0186
Age 55-57
-0.3602
0.0495
-0.0418
-0.4232
0.0607
-0.0257
Age 58-59
-0.4552
0.0551
-0.0496
-0.5636
0.0685
-0.0309
Age 60-61
-0.6150
0.0939
-0.0567
-0.7489
0.1184
-0.0326
Age 62
-0.6576
0.0851
-0.0602
-0.8244
0.1031
-0.0351
Age 63-64
-0.7030
0.0989
-0.0619
-0.8406
0.1209
-0.0349
Age 65-69
-0.7393
0.1755
-0.0607
-0.8904
0.2132
-0.0334
Age 70-74
-0.0482
0.0148
-0.0067
-0.0494
0.0129
-0.0039
Resp Income
-0.0024
0.0011
-0.0003
-0.0023
0.0013
-0.0002
Net Wealth
0.4670
0.0532
0.0856
0.5435
0.0636
0.0642
No Health Ins
0.1437
0.0517
0.0218
0.1767
0.0593
0.0158
Priv. Health Ins
0.0881
0.0550
0.0123
0.1022
0.0665
0.0081
Prob. Liv. to 85
-0.0151
0.0382
-0.0021
-0.0154
0.0457
-0.0012
Hypertension
-0.0933
0.0661
-0.0123
-0.1363
0.0794
-0.0097
Diabetes
0.2341
0.0495
0.0372
0.3026
0.0600
0.0294
Psych. Prob.
-0.0657
0.0376
-0.0092
-0.0786
0.0452
-0.0062
Arthritis
0.0209
0.1101
0.0029
0.0427
0.1229
0.0034
Mobility Diff. Index
0.0661
0.0852
0.0092
0.0668
0.0990
0.0053
Muscular Diff. Index
0.0525
0.0466
0.0075
0.0605
0.0576
0.0050
Health limit
-0.0584
0.0604
-0.0079
-0.0883
0.0691
-0.0066
Fair Health
0.0088
0.1279
0.0012
0.0619
0.1572
0.0052
Poor Health
-0.0330
0.0486
-0.0045
-0.0580
0.0593
-0.0044
Fourth Wave
-0.0359
0.0434
-0.0049
-0.0620
0.0529
-0.0047
Fifth Wave
0.0249
0.0492
0.0035
0.0041
0.0597
0.0003
Sixth Wave
0.0682
-4013.63 15,139
0.0364
Log L/Obs./Avg. Prob. -4147.16 15,139
0.0551
0.0471
Pseudo-R2
27
Table 7: Estimates of the Search Decision for Non-Employed Males
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-0.8413
0.1590
–
-0.9925
0.1753
–
-0.1031
0.0679
-0.0076
-0.1026
0.0776
-0.0039
White
0.0225
0.0648
0.0015
0.0171
0.0729
0.0006
Married
0.0050
0.0089
0.0003
0.0076
0.0097
0.0003
Yrs of Edu
0.0148
0.0823
0.0010
0.0405
0.1001
0.0015
Age 55-57
-0.0854
0.0859
-0.0055
-0.0945
0.1048
-0.0030
Age 58-59
-0.2998
0.0868
-0.0167
-0.3415
0.1028
-0.0090
Age 60-61
-0.6219
0.1086
-0.0265
-0.7441
0.1274
-0.0137
Age 62
-0.8694
0.0956
-0.0359
-1.0436
0.1130
-0.0190
Age 63-64
-0.9637
0.0922
-0.0472
-1.1402
0.1073
-0.0267
Age 65-69
-1.1874
0.1216
-0.0442
-1.4120
0.1457
-0.0236
Age 70-74
0.0308
0.0100
0.0021
0.0362
0.0112
0.0013
Resp Income
-0.0006
0.0012
-0.00004
-0.0006
0.0013
-0.00002
Net Wealth
1.0147
0.0656
0.1493
1.1685
0.0809
0.1192
No Health Ins
0.0986
0.0762
0.0073
0.1262
0.0883
0.0049
Priv. Health Ins
0.1367
0.0792
0.0094
0.1640
0.0925
0.0057
Prob. Liv. to 85
-0.0691
0.0528
-0.0048
-0.1007
0.0604
-0.0036
Hypertension
-0.1873
0.0746
-0.0117
-0.2316
0.0867
-0.0070
Diabetes
-0.0081
0.0810
-0.0006
-0.0025
0.0972
-0.0001
Psych. Prob.
-0.0507
0.0538
-0.0035
-0.0608
0.0624
-0.0021
Arthritis
-0.4290
0.1443
-0.0296
-0.5451
0.1670
-0.0191
Mobility Diff. Index
-0.1765
0.1346
-0.0122
-0.1942
0.1487
-0.0068
Muscular Diff. Index
-0.3410
0.0725
-0.0236
-0.3891
0.0777
-0.0137
Health limit
-0.0798
0.0752
-0.0053
-0.1144
0.0880
-0.0037
Fair Health
-0.1671
0.1130
-0.0103
-0.2222
0.1222
-0.0065
Poor Health
-0.2872
0.0885
-0.0159
-0.3319
0.1058
-0.0087
Fourth Wave
-0.3471
0.0718
-0.0204
-0.4151
0.0826
-0.0117
Fifth Wave
-0.1116
0.0655
-0.0073
-0.1535
0.0792
-0.0049
Sixth Wave
9,558
0.1155
-1673.45
9,558
0.0940
Log L/Obs./Avg. Prob. -1715.10
0.2246
0.2106
Pseudo-R2
28
Table 8: Estimates of the Search Decision for Non-Employed Females
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.1301
0.1355
–
-1.3505
0.1490
–
-0.1008
0.0579
-0.0063
-0.1312
0.0637
-0.0040
White
-0.3017
0.0529
-0.0200
-0.3421
0.0563
-0.0112
Married
0.0158
0.0089
0.0009
0.0212
0.0090
0.0006
Yrs of Edu
-0.1475
0.0570
-0.0078
-0.1582
0.0680
-0.0038
Age 55-57
-0.3686
0.0673
-0.0164
-0.4389
0.0815
-0.0083
Age 58-59
-0.4460
0.0675
-0.0191
-0.5131
0.0780
-0.0094
Age 60-61
-0.7498
0.1048
-0.0240
-0.8890
0.1229
-0.0112
Age 62
-0.8310
0.0841
-0.0278
-0.9941
0.1005
-0.0133
Age 63-64
-0.7845
0.0798
-0.0300
-0.9510
0.0932
-0.0152
Age 65-69
-1.0572
0.1219
-0.0317
-1.2721
0.1377
-0.0152
Age 70-74
0.0724
0.0300
0.0042
0.0846
0.0104
0.0023
Resp Income
-0.0046
0.0018
-0.0003
-0.0053
0.0017
-0.0001
Net Wealth
0.5507
0.0583
0.0500
0.6395
0.0662
0.0328
No Health Ins
0.0262
0.0602
0.0016
0.0357
0.0713
0.0010
Priv. Health Ins
0.0704
0.0660
0.0041
0.0824
0.0778
0.0023
Prob. Liv. to 85
0.0359
0.0471
0.0021
0.0334
0.0527
0.0009
Hypertension
-0.0713
0.0666
-0.0040
-0.0841
0.0776
-0.0022
Diabetes
0.1243
0.0554
0.0078
0.1461
0.0628
0.0044
Psych. Prob.
0.0061
0.0476
0.0004
0.0159
0.0543
0.0004
Arthritis
-0.1876
0.1180
-0.0110
-0.2164
0.1260
-0.0060
Mobility Diff. Index
-0.1540
0.1035
-0.0090
-0.2305
0.1133
-0.0064
Muscular Diff. Index
-0.0880
0.0583
-0.0051
-0.0972
0.0640
-0.0027
Health limit
-0.1223
0.0679
-0.0067
-0.1481
0.0715
-0.0037
Fair Health
-0.4061
0.0995
-0.0177
-0.4777
0.1111
-0.0089
Poor Health
-0.2037
0.0659
-0.0103
-0.2107
0.0795
-0.0048
Fourth Wave
-0.2579
0.0573
-0.0135
-0.2972
0.0668
-0.0070
Fifth Wave
-0.1922
0.0594
-0.0103
-0.2222
0.0694
-0.0054
Sixth Wave
0.0687
-2282.80 15,690
0.0473
Log L/Obs./Avg. Prob. -2348.50 15,690
0.1371
0.1350
Pseudo-R2
Appendix
In this Appendix we borrow from Heckman (1978), Kiefer (1982) and Greene (1993) to show how
we test for the exogeneity of the searching decision in our estimations. 11
Consider a two equation system:
y1i
X1i β1
y2i
X2i β2
d i α1
d i α2
y2i
γ1
y1i
γ2
U1i
(3)
U2i
(4)
where the dummy variable di is defined by: di 1 iff y2i
0, and di 0 otherwise.
It is important to emphasize that the system above represents two continuous latent variables
that generate observable discrete dummy variables (y1i and y2i ) when they reach a threshold. This
model is flexible enough, as Heckman (1978) shows, to include a number of important specifications. The case that we are interested in is one where equation (1) represents the structural equation
of interest, in our case the decision to become employed in the next period (employment for a third
party and self-employment are not distinguish for the purpose of this test) and (2) represents the
searching decision that might be endogenous. In equations (1) and (2), X1i and X2i , are respectively,
1 K1 and 1 K2 row vectors of bounded exogenous variables. The joint density of continuous
random variables U1i U2i is g U1i U2i which is assumed to be a bivariate normal density, with
mean normalized to be 0 and the 2 2 covariance with variances normalized to 1, and correlation
coefficient ρ 1 1 .
Our objective is to test exogeneity of the searching decision with respect to the structural decision. Without lost of generality we can consider a simple characterization of the system under
the null hypothesis of exogeneity of the search variable, by setting α 2 0 γ1 0 and γ2 0.
This model then reduces to a standard bivariate probit model, where the test for independence of
the probit equations is equivalent for us to a test of exogeneity of the searching decision. If we
cannot reject that ρ 0 then we can safely assume exogeneity in our estimations. If we reject the
null hypothesis we will have to estimate the structural parameters through a bivariate probit with a
structural shift.
Therefore, we test the independence of the probit equations resulting from (1) and (2). The
simplest method to construct the test of the hypothesis that ρ 0 follows Kiefer (1982) and Greene
(1993). The construction of the Lagrange Multiplier (LM) test only requires the estimation of the
two independent probits:
f2
h
LM
(5)
where f and h are calculated as follows:
f
φ w1i φ w2i ∑ q1iq2i Φ w1i Φ w2i i
h
∑ Φ w1i i
(6)
φ w1i φ w2i 2
Φ w1i Φ w2i Φ w2i (7)
where,
11
Benı́tez-Silva et al. (2004) also use this approach in their study of the Social Security Disability award process.
29
q1i
w1i
2y1i 1 and q2i
q1i β1 X1i and w2i
2y2i 1
q2i β2 X2i
(8)
(9)
Tables A1 and A2 presents the estimates of the independent probits for the employment and
search decisions among the employed and unemployed respectively. For the employed the LM test
statistic for this specification is 0.0002615 and follows a χ21 , delivering a P-value of 0.9871, so we
cannot reject the null hypothesis of exogeneity of the search decision. For the unemployed the LM
test statistics is 0.174889 and the P-value is 0.6758.
The LM test statistic is simpler to calculate than the Wald statistic and the Likelihood Ratio that
require the estimation of the bivariate probit, and given that we have around 20,000 observations
in the specifications we present, we believe our results are not sensitive to the test statistic used.
30
31
Table A.1: LM test on the endogeneity of job search in labor force Transition among the employed
Variable
Probit on Transition
Probit on Job Search
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Job Search
0.7767
0.0361
0.2128
Constant
-1.4869
0.0860
-1.7534
0.0988
0.1656
0.0400
0.0310
-0.0707
0.0404
-0.0108
White
0.0882
0.0360
0.0177
0.2716
0.0364
0.0415
Male
-0.0421
0.0336
-0.0085
-0.1033
0.0361
-0.0158
Married
0.0034
0.0058
0.0007
0.0495
0.0065
0.0073
Yrs of Edu
-0.0163
0.0336
-0.0032
-0.2259
0.0347
-0.0307
Age 55-57
-0.1005
0.0399
-0.0192
-0.2590
0.0414
-0.0336
Age 58-59
-0.0518
0.0432
-0.0101
-0.3050
0.0467
-0.0382
Age 60-61
-0.1269
0.0658
-0.0236
-0.5438
0.0794
-0.0559
Age 62
-0.0351
0.0573
-0.0069
-0.6066
0.0738
-0.0608
Age 63-64
-0.0206
0.0602
-0.0041
-0.5464
0.0810
-0.0568
Age 65-69
-0.3435
0.1109
-0.0557
-0.6611
0.1557
-0.0611
Age 70-74
-0.0163
0.0106
-0.0032
-0.0469
0.0100
-0.0069
Resp Income
0.0003
0.0007
0.0001
-0.0034
0.0012
-0.0005
Net Wealth
0.4398
0.0473
0.1078
0.5127
0.0507
0.1015
No Health Ins
0.0548
0.0427
0.0112
0.0828
0.0474
0.0128
Priv. Health Ins
0.0458
0.0446
0.0091
0.1143
0.0499
0.0169
Prob. Liv. to 85
-0.0361
0.0299
-0.0071
-0.0376
0.0328
-0.0055
Hypertension
0.0433
0.0490
0.0088
-0.0527
0.0603
-0.0075
Diabetes
0.0557
0.0468
0.0114
0.2204
0.0492
0.0368
Psych. Prob.
-0.0064
0.0292
-0.0013
-0.0402
0.0323
-0.0059
Arthritis
-0.0153
0.0867
-0.0031
-0.0013
0.0994
-0.0002
Mobility Diff. Index
0.0505
0.0719
0.0101
0.1386
0.0766
0.0204
Muscular Diff. Index
-0.0186
0.0381
-0.0037
0.0473
0.0411
0.0071
Health limit
0.0719
0.0461
0.0148
0.0595
0.0523
0.0091
Fair Health
0.0293
0.1190
0.0059
0.0932
0.1161
0.0147
Poor Health
0.1481
0.0363
0.0315
-0.0252
0.0418
-0.0037
Fourth Wave
0.3927
0.0318
0.0909
-0.0546
0.0385
-0.0078
Fifth Wave
0.0239
0.0399
0.0048
-0.0184
0.0439
-0.0027
Sixth Wave
19,519
0.1339
-5573.6481
19,519
0.0854
Log L/Obs/Avg Prob -7102.4572
32
Table A.2: LM test on the endogeneity of job search in labor force Transition among the non-employed
Variable
Probit on Transition
Probit on Job Search
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Job Search
1.0878
0.0454
0.2680
Constant
-1.0486
0.0890
-0.9172
0.1174
-0.0233
0.0395
-0.0032
-0.0881
0.0503
-0.0056
White
0.1416
0.0330
0.0198
0.3625
0.0435
0.0243
Male
-0.0266
0.0336
-0.0036
-0.2300
0.0432
-0.0153
Married
0.0137
0.0050
0.0019
0.0132
0.0070
0.0008
Yrs of Edu
-0.2103
0.0505
-0.0253
-0.2697
0.0539
-0.0132
Age 55-57
-0.3614
0.0543
-0.0395
-0.4973
0.0600
-0.0204
Age 58-59
-0.4911
0.0541
-0.0505
-0.6033
0.0620
-0.0237
Age 60-61
-0.5068
0.0634
-0.0497
-0.8910
0.0816
-0.0268
Age 62
-0.6168
0.0546
-0.0607
-1.0756
0.0700
-0.0341
Age 63-64
-0.6024
0.0516
-0.0652
-1.1399
0.0717
-0.0436
Age 65-69
-0.9955
0.0661
-0.0837
-1.4004
0.1004
-0.0406
Age 70-74
0.0388
0.0090
0.0053
0.0540
0.0118
0.0032
Resp Income
-0.0005
0.0005
-0.0001
-0.0035
0.0012
-0.0002
Net Wealth
0.3654
0.0457
0.0617
0.7082
0.0492
0.0749
No Health Ins
0.1968
0.0390
0.0297
0.0530
0.0535
0.0033
Priv. Health Ins
0.0859
0.0457
0.0117
0.0987
0.0577
0.0059
Prob. Liv. to 85
-0.0889
0.0292
-0.0121
0.0144
0.0394
0.0009
Hypertension
-0.0262
0.0417
-0.0035
-0.1253
0.0560
-0.0069
Diabetes
-0.0604
0.0408
-0.0080
0.0734
0.0515
0.0046
Psych. Prob.
0.0235
0.0300
0.0032
-0.0140
0.0402
-0.0008
Arthritis
-0.2295
0.0782
-0.0312
-0.1080
0.1035
-0.0065
Mobility Diff. Index
-0.1227
0.0670
-0.0167
-0.0715
0.0905
-0.0043
Muscular Diff. Index
-0.1367
0.0373
-0.0184
-0.2397
0.0513
-0.0141
Health limit
-0.1447
0.0429
-0.0185
-0.1262
0.0574
-0.0070
Fair Health
-0.2916
0.0679
-0.0334
-0.3646
0.0871
-0.0166
Poor Health
0.0060
0.0434
0.0008
-0.3145
0.0563
-0.0150
Fourth Wave
-0.0254
0.0360
-0.0034
-0.3331
0.0479
-0.0174
Fifth Wave
-0.0020
0.0370
-0.0003
-0.1537
0.0469
-0.0086
Sixth Wave
20,145
0.0980
-3239.6610
20,145
0.1040
Log L/Obs/Avg Prob -5389.7622
33
Table A.3: Estimates of the Search Decision for the Full Sample
Variable
Probit
Random Effect Probit
Coeff.
St. Error Marg. Eff.
Coeff.
St. Error Marg. Eff.
Constant
-1.5203
0.0667
–
-1.8492
0.0795
–
-0.1107
0.0274
-0.0127
-0.1532
0.0336
-0.0078
White
0.2387
0.0227
0.0267
0.2801
0.0284
0.0137
Male
-0.1487
0.0238
-0.0171
-0.1747
0.0290
-0.0088
Married
0.0282
0.0040
0.0031
0.0327
0.0047
0.0015
Yrs of Edu
-0.1425
0.0243
-0.0143
-0.1648
0.0299
-0.0068
Age 55-57
-0.2499
0.0283
-0.0232
-0.3065
0.0351
-0.0112
Age 58-59
-0.3324
0.0299
-0.0295
-0.4205
0.0370
-0.0142
Age 60-61
-0.5402
0.0450
-0.0396
-0.6876
0.0566
-0.0175
Age 62
-0.6472
0.0400
-0.0463
-0.8355
0.0498
-0.0207
Age 63-64
-0.6781
0.0410
-0.0499
-0.8922
0.0511
-0.0230
Age 65-69
-0.8592
0.0602
-0.0524
-1.1188
0.0760
-0.0222
Age 70-74
-0.0089
0.0042
-0.0010
-0.0043
0.0047
-0.0002
Resp Income
-0.0028
0.0006
-0.0003
-0.0030
0.0007
-0.0001
Net Wealth
0.1163
0.0226
0.0127
0.0441
0.0256
0.0020
Employed
0.5981
0.0284
0.0949
0.6906
0.0348
0.0579
No Health Ins
0.0633
0.0286
0.0071
0.0834
0.0342
0.0041
Priv. Health Ins
0.0693
0.0318
0.0075
0.0854
0.0382
0.0039
Prob. Liv. to 85
-0.0165
0.0218
-0.0018
-0.0188
0.0262
-0.0009
Hypertension
-0.0693
0.0338
-0.0072
-0.1089
0.0406
-0.0046
Diabetes
0.1712
0.0309
0.0205
0.1998
0.0365
0.0107
Psych. Prob.
-0.0229
0.0220
-0.0025
-0.0301
0.0266
-0.0014
Arthritis
-0.1513
0.0599
-0.0163
-0.2152
0.0687
-0.0099
Mobility Diff. Index
0.0056
0.0503
0.0006
-0.0095
0.0587
-0.0004
Muscular Diff. Index
-0.0430
0.0270
-0.0046
-0.0577
0.0319
-0.0026
Health limit
-0.0350
0.0314
-0.0037
-0.0492
0.0369
-0.0022
Fair Health
-0.2383
0.0545
-0.0217
-0.2804
0.0639
-0.0101
Poor Health
Fourth Wave
-0.0775
0.0271
-0.0080
-0.1059
0.0347
-0.0045
-0.1421
0.0242
-0.0144
-0.1906
0.0305
-0.0079
Fifth Wave
-0.0594
0.0264
-0.0062
-0.0833
0.0331
-0.0036
Sixth Wave
56,844
0.0690
-12397.8010
56,844
0.0329
Log L/Obs./Avg. Prob. -13003.1480
0.0831
0.0716
Pseudo-R2
References
Albrecht, J.W., and B. Axell (1984): “An Equilibrium Model of Search Unemployment,” Journal
of Political Economy, 92 824–840.
Barron, J.M. (1975): “Search in the Labor Market and the Duration of Unemployment: Some
Empirical Evidence,” American Economic Review, 65-5 934–942.
Barron, J.M., and S. McCafferty (1977): “Job Search, Labor Supply, and the Quit Decision:
Theory and Evidence,” American Economic Review, 67-4 683–691.
Barron, J.M., and W. Mellow (1981): “Changes in Labor Force Status Among the Unemployed,”
Journal of Human Resources, 16-3 427–441.
Ben-Porath, Y. (1967): “The Production of Human Capital and the Life Cycle of Earnings,”
Journal of Political Economy, 75-4 Part I 352–365.
Bhattacharya, J., C.B. Mulligan, and R.R. Reed III (2001): “Labor Market Search and Optimal
Retirement Policy,” NBER Working Papers 8591.
Benı́tez-Silva, H., M. Buchinsky, H-M. Chan, J. Rust, and S. Cheidvasser (2004): “How Large is
the Bias in Self-Reported Disability Status?” Journal of Applied Econometrics, 19-6 649–
670.
Benı́tez-Silva, H. (2000): “Micro Determinants of Labor Force Status Among Older Americans,”
manuscript, SUNY-Stony Brook.
Black, M. (1981): “An Empirical Test of the Theory of on-the-job Search,” Journal of Human
Resources, 16-1 129–140.
Blau D., P. K. Robins (1990): “Job Search Outcomes for the Employed and Unemployed,” Journal of Political Economy, 98-3 637–655.
Borus, M.E., H.S. Parnes, S.H. Sandell, and B. Seidman (1988): The Older Worker, IRRA Series.
Burdett, K., and D. T. Mortensen (1998): “Wage Differentials, Employer Size, and Unemployment,” International Economic Review, 39-2 257–273.
Burtless, G., and J.F. Quinn (2000): “Retirement Trends and Policies to Encourage Work Among
Older Americans,” Working Paper 436, Boston College.
Burtless, G., and J.F. Quinn (2002): “Is Working Longer the Answer for an Aging Workforce?”
Working Paper 550, Boston College.
Currie, J. and B.C. Madrian (1999): “Health, Health Insurance and the Labor Market,” in the
Handbook of Labor Economics, Vol. 3C. Orley Ashenfelter and David Card, eds.
Devine, T. J., and N.M. Kiefer (1991): Empirical Labor Economics: The Search Approach.
Oxford University Press, New York.
Dwyer, D.S., and O.S. Mitchell (1999): “Health Problems as Determinants of Retirement: Are
Self-rated Measures Endogenous?” Journal of Health Economics, 18-2 173–193.
34
Eckstein, Z., and K.I. Wolpin (1990): “Estimating a Market Equilibrium Search Model from
Panel Data on Individuals,” Econometrica, 58-4 783–808.
Feinberg, R.M. (1977): “Search in the Labor Market and the Duration of Unemployment: Note,”
American Economic Review, 67-5 1011–1013.
Flinn, C. J., and J. J. Heckman (1983): “Are Unemployment and Out of the Labor Force Behaviorally Distinct Labor Force States?” Journal of Labor Economics, 1-1 28–42.
French, E. (2005): “The Effects of Health, Wealth, and Wages on Labor Supply and Retirement
Behavior,” Review of Economic Studies, 72 395–427.
Friedberg, L. (1999a): “The Trend Towards Part-Time Work Among Older Workers,” manuscript,
University of California, San Diego.
Friedberg, L. (1999b): “The impact of Technological Change on Older Workers: Evidence from
Data on Computers,” University of California, San Diego, Discussion Paper No. 99-11.
Gronau, R. (1971): “Information and Frictional Unemployment,” American Economic Review,
61-3 290–301.
Guilkey, D.K., and J.L. Murphy (1993): “Estimation and testing in the random effects probit
model,” Journal of Econometrics,” 59 301–317.
Gustman, A. L., O. Mitchell, and T.L. Steinmeier (1995): “Retirement Measures in the Health
and Retirement Study,” The Journal of Human Resources 30 (Supplement).
Heckman, J.J. (1974): “Life Cycle Consumption and Labor Supply: an Explanation of the Relationship between Income and Consumption over the Life Cycle,” American Economic Review, 64-1 188–194.
Heckman, J.J. (1976a): “Estimates of a Human Capital Production Function Embedded in a LifeCycle Model of Labor Supply,” in Household Production and Consumption, N. E. Terleckyj
(ed.). Columbia University Press.
Heckman, J.J. (1976b): “A Life-Cycle Model of Earnings, Learning, and Consumption,” Journal
of Political Economy, 84-4.2 S11–S44.
Holzer, H.J. (1987): “Job Search by Employed an Unemployed Youth,” Industrial and Labor
Relations Review, 40 601–611.
Hutchens, R. M. (1988): “Do Job Opportunities Decline with Age,” Industrial and Labor Relations Review, 42-1 89–99.
Hutchens, R. M. (1993): “Restricted Job Opportunities and the Older Worker,” in As the Workforce Ages: Costs, Benefits, and Policy Challenges. Olivia S. Mitchell (ed.) ILR Press,
Ithaca, New York.
Jovanovic, B. (1979): “Job Matching and the Theory of Turnover,” Journal of Political Economy,
87-5 972–990.
35
Juster, T.F., and R. Suzman (1995): “An Overview of the Health and Retirement Study,” The
Journal of Human Resources, 30 (Supplement): S7–S56.
Kahn, L.M., and S.A. Low (1982): “The Relative Effects of Employed and Unemployed Job
Search,” Review of Economics and Statistics, 64-2 234–241.
Kiefer, N.M., and G.R. Neumann (1979): “An Empirical Job-Search Model, with a Test of the
Constant Reservation-Wage Hypothesis,” Journal of Political Economy, 87-1 89–107.
Lentz, R., and T. Tranaes (2000): “Job Search, Savings and Wealth Effects,” manuscript, Northwestern University.
Lippman, S.A., and J.J. McCall (1976): “The Economics of Job Search: A Survey,” Parts I and
II. Economic Inquiry, 14 155–189 and 347–368.
Ljungqvist, L., and T.J. Sargent (2000): Recursive Macroeconomic Theory. MIT Press.
Low, H.W. (1998): “Simulation Methods and Economic Analysis,” manuscript, University College London, University of London.
Low, H.W. (1999): “Self Insurance and Unemployment Benefit in a Life-Cycle Model of Labour
Supply and Savings,” manuscript, University of Cambridge.
Maddala, G.S. (1987): “Limited Dependent Variable Models Using Panel Data,” Journal of Human Resources, 22-3 307–338.
Maestas, N., and X. Li (2006): “Discouraged Workers? Job Search Outcomes of Older Workers.
MRRC Working Paper, WP 2006-133.
McCall, J. J. (1970): “Economics of Information and Job Search,” Quarterly Journal of Economics, 84-1 113–26.
Miller, R.A. (1984): “Job Matching and Occupational Choice,” Journal of Political Economy,
92-6 1086–1120.
Mortensen, D.T. (1970): “Job search, the duration of unemployment, and the Phillips curve,”
American Economic Review, 60-5 847–862.
Neal, D. (1999): “The Complexity of Job Mobility among Young Men,” Journal of Labor Economics, 17-2 237–261.
Parsons, D. O. (1991): “The Job Search Behavior of Employed Youth,” Review of Economics and
Statistics, 73-4 597–604.
Peterson, R.L. (1972): “Economics of Information and Job Search: Another View,” Quarterly
Journal of Economics, 86-1 127–131.
Quinn, J. (1998): “Retirement Patterns and Bridge Jobs in the 1990’s,” manuscript, Boston College.
Rust, J., M. Buchinsky, and H. Benı́tez-Silva (2003): “Dynamic Structural Models of Retirement
and Disability,” manuscript, University of Maryland.
36
Sandell, S.H. (1980a): “Job Search by Unemployed Women: Determinants of the Asking Wage,”
Industrial and Labor Relations Review, 33-3 368–378.
Sandell, S.H. (1980b): “Is the Unemployment Rate of Women Too Low? A Direct Test of the
Economic Theory of Job Search,” Review of Economics and Statistics, 62-4 634–638.
Sandell, S.H. (1987): The Problem Isn’t Age: Work and Older Americans, (ed.) Prager Publishers,
New York.
Sargent, T.J. (1987): Dynamic Macroeconomic Theory. Harvard University Press.
Seater, J. J. (1977): “A Unified Model of Consumption, Labor Supply, and Job Search,” Journal
of Economic Theory, 14 349–372.
Siven, C-H. (1974): “Consumption, Supply of Labor and Search Activity in an Intertemporal
Perspective,” Swedish Journal of Economics, 76-1 44–61.
Stern, S. (1989): “Estimating a Simultaneous Search Model,” Journal of Labor Economics, 7-3
348–369.
Stigler, G. J. (1961): “The Economics of Information,” Journal of Political Economy, 69-3 213–
225.
Stigler, G. J. (1962): “Information in the Labor Market,” Journal of Political Economy, 70-5
94–105.
Topel, R. H., and M. P. Ward (1992): “Job Mobility and the Careers of Young Men,” Quarterly
Journal of Economics, 107-2 439–479.
Van den Berg, D. (1999): “Empirical Inference with Equilibrium Search Models of the Labour
Market,” Economic Journal, 109 283–306.
Van den Berg, D. and G. Ridder (1998): “An Empirical Equilibrium Search Model of the Labor
Market,” Econometrica, 66-5 1183–1221.
van der Klaauw, W., and K.I. Wolpin (2005): “Social Security, Pensions and the Savings and
Retirement Behavior of Households,” manuscript, University of North Carolina at Chapel
Hill.
Whipple, D. (1973): “A Generalized Theory of Job Search,” Journal of Political Economy, 81-5
1170–1188.
Wolpin, K.I. (1987): “Estimating a Structural Search Model: The Transition from School to
Work,” Econometrica, 55-4 801–817.
Wright, R. (2000): “Job Search Theory,” Lecture Notes. University of Pennsylvania.
37