You are on page 1of 5

Fringe Benefits and Labor Mobility Author(s): Olivia S.

Mitchell Source: The Journal of Human Resources,


Vol. 17, No. 2 (Spring, 1982), pp. 286-298 Published by: University of Wisconsin Press Stable URL:
https://www.jstor.org/stable/145474 Accessed: 11-02-2019 07:54 UTC JSTOR is a not-for-profit service
that helps scholars, researchers, and students discover, use, and build upon a wide range of content in a
trusted digital archive. We use information technology and tools to increase productivity and facilitate
new forms of scholarship. For more information about JSTOR, please contact support@jstor.org. Your use
of the JSTOR archive indicates your acceptance of the Terms & Conditions of Use, available at
https://about.jstor.org/terms University of Wisconsin Press is collaborating with JSTOR to digitize,
preserve and extend access to The Journal of Human Resources(Mitchell, 1982) This content
downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use subject to
https://about.jstor.org/terms FRINGE BENEFITS AND LABOR MOBILITY* Voluntary job changes are the
major mechanism in our economy aligning desired and actual labor supply.1 In this paper I examine one
aspect of this mechanism: the effect of fringe benefits on job change by workers. Fringe benefits are not
usually portable across jobs and thus are often described as deterrents to mobility.(Mitchell, 1982)
However, until recently this prediction has not been examined in depth, depite the fact that it is relevant
to a number of policy issues. For instance, fringe-benefit improvements are often favored during periods
of wage-price guidelines, and it is of interest to determine the probable impact on labor mobility flows.
Recent and proposed legislation in the private pension area will influence the speed with which workers
gain pension rights, and this, too, may affect mobility. Finally, a national health-care plan may eventually
replace medical-care packages currently available through em- ployers, and this may affect mobility
patterns as well. Here I report the results of testing a microeconomic model of the effects of fringe
benefits on labor mobility. I. TOWARD AN EMPIRICAL SPECIFICATION A fringe benefit is a form of
nonwage compensation provided by a worker's firm. Examples include medical and disability insurance,
pension programs, life insurance, and similar benefits. Frequently fringe benefits decline in value when a
worker leaves his job.2 For instance, coverage by employer-sponsored * Financial support for this
research was provided by the Employment and Training Admin- istration, U.S. Department of Labor. The
capable research assistance of Gloria Bazzoli is gratefully acknowledged. Helpful comments were
provided by colleagues at Cornell University, Janet Johnson, and the editor. None of the above persons
or institutions is responsible for the views contained herein. [Manuscript received April 1980; accepted
September 1980.] 1 The Michigan Quality of Employment Panel Survey indicates that among wage and
salary workers, 22 percent of the males and 29 percent of the females surveyed over a four-year period
quit their jobs. Total job-change figures are 35 and 41 percent, respectively (see Table 1). Mattila [11]
and Freeman [7] also find that quits exceed layoffs as a reason for job change. 2 This paper investigates
worker mobility across employers and does not consider movements within a firm. See Olson [16] on
intrafirm job change. The Journal of Human Resources * XVII * 2 0022-166X/82/0002-0286 $01.50/0 o
1982 by the Regents of the University of Wisconsin System This content downloaded from
119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use subject to https://about.jstor.org/terms
Communications 1 287 health plans typically terminates shortly after the worker leaves the firm, and
although private insurance is available, its cost is usually higher than for group plans. Even if the worker
obtains medical coverage through a new employer, most group plans require him or her to undergo an
initial waiting period before becoming a participant. In addition, preexisting medical con- ditions are
often exempted from coverage. Therefore, the loss of medical insurance coverage upon changing jobs
may be quite costly to the worker. Life and disability insurance plans operate in much the same way.3 An
employer-sponsored pension program will also fall in value for a worker who severs his attachment to
that firm. If the worker is not vested, pension contributions made on his behalf are lost. Even if he is
vested, the value of the pension promise may fall when he leaves his job. For example, many pension
plans base retirement benefits on the worker's final salary. A pension promise to a job leaver would be
calculated on the pre-quit rate rather than the pre-retirement salary, so that benefits are worth less to
the worker the lower his wage. High inflation further erodes the real value of eventual pension benefits
for job leavers prior to retirement. Finally, the worker who changes employers will typically not be able
to transfer accu- mulated pension funds to a different plan, even if one is available on his new job.4
Therefore, he will have little opportunity to influence pension investment strategies either directly or
through his union. All of these factors suggest that the value of a pension promise declines when
workers change jobs. Other fringe benefits also typically increase with tenure. Senior workers are less
likely to be laid off and often are eligible for more vacation days and more medical coverage.5 Previous
writers have suggested that this loss of fringe benefits is likely to deter worker mobility. Ross [21]
provided one of the first empirical analyses of this thesis, and more recent studies include those by
Pencavel [18] and Burton and Parker [5]. Unfortunately, empirical confirmation of this prediction is as yet
tenuous, since aggregate data often used in these studies are measured inaccurately and aggregate
equations are plagued by simultaneity bias.6 Analysts of individual-level data have recently begun to
examine how 3 Group medical, life, and disability insurance characteristics are discussed in Gregg and
Lucas [9] 4 Portability is common in only a few industries, notably trucking and construction. Private
pension plan characteristics are reviewed in McGill [12]. 5 Profit-sharing and stock-option programs are
another type of fringe benefit that may decline in value for job changers. These plans offer employees
either cash or deferred benefits, with contributions to the plans depending on the firm's profitability. In
the event that benefits are not received until retirement, the job-changer will incur a loss of effective
income as in the case of employer-sponsored pensions. 6 See Pencavel's discussion [18]. This content
downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use subject to
https://about.jstor.org/terms 288 | THE JOURNAL OF HUMAN RESOURCES nonwage factors influence
labor market adjustment.7 However, these studies have focused primarily on a particular type of
compensation-unemployment insurance-and a particular type of mobility-the decision to move out of
unemployment into a job. Microeconomic empirical research on worker move- ments across employers
is still in its infancy,8 and as yet no individual-level study has examined the influence of a variety of fringe
benefits on job change.9 The following model is proposed to examine the role of fringe benefits in
determining worker quit behavior in individual-level data. The decision to quit is postulated to depend
on the difference between the current compensation package (Cc) and that available from an alternative
job (Ca): (1) D = Cc- Ca The greater the difference between CC and Ca, the less likely is job change. In
general, the probability of a quit can be summarized by an index Q, which is a function of the
determinants of mobility: (2) Q = f(Wc, Fc, Wa, Fa,, r, t) where W = wage rate, F = fringe benefits, r =
discount rate, and t = tenure. The hypothesis that current fringes deter mobility implies that aQ/ dFc < 0,
ceteris paribus. This discussion suggests that an empirical model of the quit decision should include both
wage and fringe-benefit streams available on current and alternative jobs. More general models could
also incorporate uncertainty by including layoff and new hire probabilities with different employers.
Retirement ages could also be examined endogenously. And finally, the ideal data base would contain
longitudinal information on workers so as to reveal mobility patterns over time. Such a formidable data
base is not available; indeed, most researchers in this area have had great difficulty obtaining any
information linking workers to their places of employment for confidentiality reasons. In particular, no
nationally representative panel data set contains any information on the portability of fringe benefits.
The survey examined in the present paper does provide far more infor- mation on workers and their
compensation packages than previously available and contains variables which can be used to explore
some of the portability issues. It is a panel data set from the nationally representative Quality of 7 See
the studies reviewed in Gustman [10] and Parsons [17]. 8 See Borjas and Rosen [4] and Mincer and
Jovanovich [14]. Freeman [8] has examined the effects of unionism on workers' mobility in micro data. 9
Schiller and Weiss [22] evaluated the impact of private pension characteristics on worker mobility, but
did not consider other fringes and examined only behavior of male employees in the nation's largest
firms. This content downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use
subject to https://about.jstor.org/terms Communications I 289 Employment Survey (QES) conducted by
the Michigan Survey Research Center.10 The QES contains complete information on 530 males and 252
females interviewed in 1973 and 1977," including data on individual workers' earnings, job
characteristics, human capital, availability of several fringe benefits, and mobility patterns over the panel
period.'2 A simple model that may be estimated with QES data uses as a dependent variable a
dichotomous term taking on the value of one if the individual changed jobs over the panel period, and
zero otherwise. Earnings in 1973 are used as a measure of WC13 and a vector of fringe benefits available
on the baseline job is used to measure Pc. Dollar values of Pc are not provided directly so that availability
dummies are interacted with several other terms to reflect the fact that senior employees are likely to be
eligible to receive higher nonwage benefits and that they may lose more of the pension fund if they
quit.14 The analysis also interacts fringe benefit variables with firm size and union terms. Mitchell and
Andrews [15] suggest that pension benefits may be provided more cheaply in larger firms because of
risk-pooling and because larger firms can hire pension trust managers. Freeman [8] argues that unions
are more able to force employers to pay these promised benefits than are nonunion workers. Thus,
fringe-benefit terms are interacted with union and firm-size dummies to evaluate whether such factors
lower the probability of quits. The desired measure of Wa is, of course, the alternative wage stream
available should an individual change jobs. In this study several determinants of the "offered wage" are
entered directly, including the worker's education, 10 The data are described in Quinn and Staines [19]
and Quinn, Mangione, and Seashore [20]. The panel is described in the 1974-77 Codebook [23]. 11 In
1973, the QES included only those individuals who worked 20 hours or more per week. However, this
rule does not appear to make the sample unrepresentative. Bureau of Labor Statistics data indicate that
in 1973 more than 95 percent of the male employees and 80 percent of the females worked more than
35 hours per week. The cutoff used in the QES is substantially lower and will therefore exclude fewer
individuals. 12 The QES contains information on 596 males and 324 females between the ages of 18 and
65, who were civilian nonfarm wage and salary workers in 1973. Item nonresponses and data editing
problems require a reduction in the sample size to 530 males and 252 females (89 and 78 percent of the
samples, respectively). In some cases wage rates were coded incorrectly; for instance, one plumber listed
his wage rate as $27,000 per hour, a figure too high even in these inflationary times. Other individuals
who had missing data on experience, industry, occupation, and wage income in 1973 were also excluded
since no baseline wage could be derived for these persons. Exploratory analysis of job-change patterns
which used the full sample indicated that exclusion of those without wage data does not alter results
reported below. 13 Additional variables reflecting job security and promotion possibilities in the baseline
job may also be used to parameterize the stream of wages with the present employer, We. 14 An
interesting test of the effect of pension vesting allows tenure to enter as a spline function since most
workers vest after 10 years of service. Age interactions are also examined below. This content
downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use subject to
https://about.jstor.org/terms 290 | THE JOURNAL OF HUMAN RESOURCES TABLE 1 MEANS (STANDARD
DEVIATIONS) OF THE QES SAMPLE Males (N = 512) Females (N = 252) Quit (1 = yes) .22 (.48) .29 (.46) Job
change (1 = yes) .35 (.42) .41 (.49) Race (1 = nonwhite) .08 (.27) .15 (.36) Education (years) 12.75 (3.06)
12.54 (2.56) Union (1 = yes) .39 (.49) .26 (.44) Experience (years) 18.78 (11.11) 13.93 (10.30) Tenure
(years) 7.27 (9.02) 5.51 (6.58) Wage (in logs) 2.33 (.47) 1.66 (.56) Age (years) 36.52 (11.54) 37.00 (12.40)
Pension (1 = yes) .75 (.42) .63 (.48) Medical (1 = yes) .86 (.35) .76 (.42) Life (1 = yes) .76 (.42) .64 (.48)
Stock (1 = yes) .20 (.40) .13 (.34) Profit (1 = yes) .22 (.42) .15 (.34) race, and the number of years of labor
force experience (available directly from the data). A worker must also take into account the likelihood of
finding attractive fringe benefits in an alternative job. One approach would be to estimate the value of
benefits available elsewhere, given information on workers who quit to take new jobs. This is not
possible in the QES data set (or any other representative data set) since no dollar values of workers'
fringes are provided. The survey does, however, ask workers to evaluate their chances of finding another
job with compensation attributes at least as great as those which they currently receive. One would
expect this term to be positively related to quits. The empirical form of equation (2) is estimated using
probit analysis, where the dependent variable is equal to one if the worker changed employers between
1973 and 1977. All explanatory variables are measured as of the baseline interview in 1973. Summary
statistics appear in. Table 1. This model assumes that all workers' quit probabilities are described by the
same functional relationship so that observed differences in behavior are due to different job-changing
costs across workers. With more waves of lon- gitudinal data, the model could be extended to a
framework that also allows taste differences in the population to affect the functional relationship
between quit rates and job-changing costs. Unfortunately, the QES panel, which has the several useful
features mentioned above, does not permit this extension since only one quit decision is reported.
Estimates are presented separately by sex, however, which may control for some of these effects.
Additionally, recent work on unions and turnover by Freeman [8] that controls for hetThis content
downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use subject to
https://about.jstor.org/terms Communications 1 291 erogeneity finds that estimates of union
coefficients are not affected. Thus it is likely that the functional relationship utilized here is a reasonable
speci- fication of underlying behavior. III. EMPIRICAL RESULTS The QES panel consists of individuals
surveyed in both 1973 and 1977, regardless of their employment status in 1977.15 Thus some workers
moved from wage and salary work to self-employment, retirement, or out of the labor force. In any
event, all individuals were asked in 1977 whether or not they were with the same employer as in 1973. If
not, the circumstances under which they left that firm were detailed. Twenty-nine percent of the
females and 22 percent of the males quit their jobs over the four-year period, figures roughly
comparable to statistics from other longitudinal data sources.16 How- ever, quits were identified as the
motivation for only about half of all job changes; other separations included layoffs or discharges.
Because of the large difference between reported quits and total job changes in these data, the
empirical analysis below examines each variable in turn to compare the effects of fringes on quit and
total mobility. Quit Patterns and Their Determinants The empirical work below suggests three main
conclusions about the impact of fringe benefits on quits: 1. Mobility is lower for workers with fringe
benefits, and this effect is robust to the inclusion (or exclusion) of other variables. A male worker with a
pension plan is 10 percent less likely to quit than his counterpart without a pension. 2. Higher wages
reduce a worker's quit probability, but this effect is cut in half when fringe benefits are incorporated in
the model. Studies which ignore fringe benefits as a deterrent to quits may overstate the effects of
wages. 3. Quit patterns differ by sex. Females are less responsive than males to the loss of fringe
benefits. Pensions. Results in Table 2 suggest that a pension promise significantly 15 The present data
base is therefore more representative than those used in studies which eliminate persons who were not
wage earners at every interview date (cf. Freeman [7]). 16 For instance, Freeman [7] reports quit rates of
10 percent for older males in the National Longitudinal Survey between 1966 and 1971, and 30 percent
among respondents of all ages in the Michigan Panel Data from 1968 to 1972. Schiller and Weiss [22]
report turnover rates of about 11 percent in one year for their sample of white males in large firms.
Borjas and Rosen [4] indicate that 34.5 percent of the NLS Young Men changed jobs between 1971 and
1973. This content downloaded from 119.160.119.122 on Mon, 11 Feb 2019 07:54:17 UTC All use
subject to https://about.jstor.org/terms 292 | THE JOURNAL OF HUMAN RESOURCES TABLE 2 QUIT
PROBABILITY MODELSa Males Females 1 2 3 4 5 6 Constant 1.306** 1.073** 1.064** -.704 -.707 .674
(.435) (.438) (.444) (.505) (.505) (.510) Race -.083 -.087 -.116 .043 .046 .092 (.276) (.279) (.284) (.245)
(.246) (.251) Education -.005 .005 .008 .073* .075* .072* (.026) (.026) (.027) (.041) (.041) (.041) Union
-.485** -.377** -.392** -.643** -.580** -.615** (.153) (.158) (.168) (.232) (.236) (.242) Experience
-.024** -.026** -.027** -.014 -.014 -.015 (.008) (.008) (.008) (.010) (.010) (.010) Tenure -.097** -.089**
-.088** -.033 -.031 -.028 (.025) (.026) (.026) (.026) (.025) (.026) Wage -.435** -.275** -.224 -.179 -.104
-.027 (.169) (.175) (.179) (.176) (.183) (.192) Pension -.458** -.443** -.274 -.140 (.168) (.183) (.198)
(.221) Medical -.200 .022 (.240) (.252) Life .116 -.319 (.204) (.227) Stock -.200 .171 (.196) (.257) Profit
-.019 -.377 (.182) (.306) -21nL 432.3 424.8 422.6 273.9 271.9 268.1 a Probit coefficients from a model
where the dependent variable equals one if the individual quit his 1973 job; zero otherwise. Asymptotic
standard errors appear in parentheses. **t-statistic > 1.96. *t-statistic > 1.65 (

Mitchell, O. S. (1982). Fringe Benefits and Labor Mobility. The Journal of Human Resources, 17(2), 286.

https://doi.org/10.2307/145474

You might also like