Upload
vodung
View
215
Download
1
Embed Size (px)
Citation preview
Job Mobility in 1990s Britain: Does Gender Matter? *
Alison L Booth and Marco FrancesconiInstitute for Social and Economic Research
University of EssexWivenhoe Park
Colchester CO4 3SQEngland
Tel.: 44-1206-873087Fax: 44-1206-873151
e-mail: [email protected]
Institute for Social and Economic ResearchUniversity of Essex
Wivenhoe ParkColchester CO4 3SQ
EnglandTel.: 44-1206-873534Fax: 44-1206-873151
e-mail: [email protected]
Title as Running Head: Job Mobility in 1990s Britain: Does Gender Matter?
AbstractThe paper examines gender differences in intra-firm and inter-firm job changes, includingworker-initiated and firm-initiated separations, for white full-time British workers over theperiod 1991-96. We document four main findings. First, job mobility is high for both men andwomen, with more than one quarter of the sample changing job each year. Second, thedistinction between promotions, quits and layoffs is important, suggesting that studies thateither aggregate worker-initiated and firm-initiated separations or neglect within-firm mobilitymay provide an inappropriate picture of career mobility. Third, we find that the average maleand female quit and promotion probabilities are remarkably similar, but there are significantgender differences in layoff probabilities. Fourth, we find significant gender differences in theimpact of variables such as union coverage, occupation and presence of young children.
First version, January 1999This version, September 1999
JEL Classification: J24, J41, J62Keywords: Career mobility, gender, promotions, quits, layoffs
* The support of the Economic and Social Research Council under Award No. L212252007 is gratefullyacknowledged. Views expressed here are those of the authors and not necessarily those of the ESRC. We aregrateful to an anonymous referee, Joe Altonji, Anne Preston, and seminar participants at the University ofAmsterdam, University of Essex, and the American Economic Association meetings 1999 (New York) for usefulcomments, and to Jeff Frank for stimulating discussion.
1
NON-TECHNICAL SUMMARY
It is well-known that women workers have a lower attachment to the labour force than men,with potentially important consequences for human capital accumulation, job mobility andoccupational segregation by gender. But are men more likely than women to be promoted?How do career patterns differ by gender for workers who are strongly attached to the labourmarket? Does distinguishing between intra-firm and inter-firm job changes improve ourunderstanding of gender differences in mobility? We address these questions using a sampleof workers from the British Household Panel Survey (BHPS), collected annually over theperiod 1991-1996. This sample comprises only full-time workers, and excludes from theanalysis job-to-nonemployment transitions. Our data allow career (or job-to-job) mobility tobe disaggregated into job changes involving a promotion within a firm and job changesinvolving movements across firms. With this disaggregation, the paper extends previous workin two directions. First, it distinguishes between internal and external job mobility, which hasnot been possible before using other British survey data. Second, within the career mobilityliterature, it extends the approach of most US empirical research by distinguishing betweenworker-initiated (quits) and firm-initiated (layoffs) job changes across firms.
We find that, although women’s promotion and quit rates are higher than men’s in theraw data, such differences vanish once we control for standard individual and jobcharacteristics. This contrasts with the popular view that women quit more often and arepromoted less frequently than men. However, women’s layoff rates remain significantlyhigher than men’s. Our results also demonstrate that, although average job mobility rates ofmen and women in the sample are similar, the rates do respond differently to specific changesin their socio-economic environment. These findings emphasise the importance ofdistinguishing between different forms of job mobility. The turnover effects of certainvariables, such as the presence of young children, union coverage and occupation, differsignificantly by gender across all forms of job mobility. Thus, even with a reasonablyhomogenous sample of workers, as long as women’s and men’s career patterns differ in termsof their response to changes in individual or job-specific characteristics, analyses that focussimply on the overall separation rate and neglect intra-firm mobility may provide anincomplete picture of workers’ career development and reach potentially misleadingconclusions.
2
1. Introduction
Job mobility is a striking feature of the British labour market in the 1990s. According
to new longitudinal data, each year more than a quarter of full-time workers can expect to
change job, and another quarter will move again one year later. Yet relatively little is known
about the incidence of various forms of mobility within and across firms and how they differ
by gender. Are men more likely than women to be promoted? What are the major
determinants of job changes for men and women? It is well-known that women workers have
a lower attachment to the labour force than men, with potentially important consequences for
human capital accumulation, job mobility and occupational segregation by gender (Mincer
and Ofek, 1982; Royalty, 1998). But how do career patterns differ by gender for workers who
are strongly attached to the labour market? Does distinguishing between intra-firm and inter-
firm job changes improve our understanding of gender differences in mobility? We address
these questions using a sample of workers from the British Household Panel Survey (BHPS),
collected annually over the period 1991-1996. This sample comprises only full-time workers,
and excludes from the analysis job-to-nonemployment transitions.
Our data allow career (or job-to-job) mobility to be disaggregated into job changes
involving a promotion within a firm and job changes involving movements across firms. With
this disaggregation, the paper extends previous work in two directions. First, it distinguishes
between internal and external job mobility, which has not been possible before using other
British survey data.1 This distinction is important, because promotions are an integral part of
workers’ careers (Gibbons, 1998; Gibbons and Waldman, 1999; and references therein).
Second, within the career mobility literature, it extends the approach of most US empirical
research by distinguishing between worker-initiated (quits) and firm-initiated (layoffs) job
changes across firms (Sicherman and Galor, 1990; McCue, 1996). This distinction is
3
important when there are informational asymmetries and costly renegotiation between workers
and firms (Hall and Lazear, 1984).
2. Background
Why do workers change job? From the supply side, workers leave their job (to a
different employer or another job at the same employer) if the expected utility from so doing
exceeds current utility less the costs of the change. To the extent that men and women differ in
alternative opportunities and costs, there may be gender differences in job mobility. Voluntary
job separation behaviour has been addressed most frequently in the context of human capital
and job-matching theory (see among others Oi, 1962; McLaughlin, 1991; Harper, 1995).
Gender implications for career mobility can be obtained from these models if there are (i)
differences in male and female human capital acquisition that make men or women less highly
valued by a firm (Blau and Kahn, 1981); or (ii) gender differences in job search costs
(Meitzen, 1986); or (iii) differences in employers’ monopsony power, which arise if there is
gender discrimination in hiring so that women must search for a gender match as well as the
usual job match (Neumark, 1988).
From the demand side, firms will terminate a job if the profits from so doing exceed
expected profits from continuation, less any redundancy costs. Permanent exogenous shocks
(through, for example, technical and structural change rendering skills obsolete) reduce
expected profits by lowering marginal productivity. Nominal wages in Britain are typically
downwardly rigid, and hence demand shocks may induce layoffs. To the extent that there is
occupational or industrial segregation or sex-typing, women and men may be subject to
differential demand shocks. Moreover, employers will want to dismiss a worker if the match
quality is poor. Although the scope for dismissal may be circumscribed by labour laws,
4
dismissals may be packaged as redundancies for which there is a statutory procedure in
Britain. If employers follow different strategies in their attempts to retain men and women
(through promotions, wages and/or bonuses, targeted redundancy pay), there may be gender
differences in involuntary job separation behaviour.
The last fifteen years have witnessed a remarkable growth of studies of the
organisation of labour within firms (for recent surveys, see Gibbons, 1996, and Gibbons and
Waldman 1999). However, we currently have little systematic knowledge of the contribution
of promotions to career mobility, and even less knowledge of sources of gender differences in
career mobility after accounting for promotions.2 Firms’ promotion policies are a means of
increasing productivity within an organisation by increasing human capital acquisition,
increasing effort, of inducing separating equilibria in terms of worker types (Chang and Wang,
1995), or even of constraining favouritism (Prendergast and Topel, 1996). Firms may
backload compensation to elicit higher levels of effort, where effort may be proxied by hours
of overtime work (Landers et al., 1996). If women are constrained by family factors from
working long hours, this may lead to gender differences in promotion rates. There may also be
gender differences in the way family responsibilities affect promotion and mobility; if women
are more likely to quit, firms will be less likely to train and promote them. On the other hand,
if women view promotion as unlikely due to discriminatory promotion practices, they may be
less prone to put themselves forward for training programmes at the firm.
Gender differences in job mobility rates are not only interesting in their own right, but
may also suggest explanations of the gender pay gap. If, for example, women are more likely
to quit their jobs than men, firms will be less willing to invest in their training, resulting in
lower accumulation of human capital and ultimately lower rates of pay. Alternatively, if
women are more likely to receive a promotion, and large wage jumps are observed upon
5
promotion, we may expect promotions to play an important role in reducing gender wage
differentials.3
3. The data
The data are from the first six waves of the British Household Panel Survey (BHPS), a
nationally representative survey collected annually since 1991. The BHPS provides
information on the timing and type of job changes, including job changes at the same
employer. For all jobs ending during the 12-month period between interviews, workers give
the reason for stopping a job. Therefore we can identify job changes involving promotion (a
change of duties or different job spell at the one employer), movement across employers, and
other forms of job termination. We define a firm-initiated separation or layoff as when a
worker is either made redundant or dismissed, or when a temporary job is terminated. All
other movements across employers are defined as worker-initiated separations.4
Our estimating sample comprises white men and women who: (i) were born after
1936; (ii) reported full interviews; (iii) have at least two years of labour market data; (iv) were
in full-time employment at the time of the survey; and (v) were not self-employed, farmers, or
in the armed forces. These restrictions primarily imposed to narrow the sample to those with
a reasonably strong attachment to the labour market yield an unbalanced panel comprising
2,135 men and 1,475 women, with 9,697 and 6,210 person-year observations, respectively.5
At the bottom of Tables 1 and 2, we show the distributions of the male and female
samples by career states (the omitted state is staying with the same employer without
promotion). For this sample of workers, the mobility rates by gender are very similar, with
women being slightly (but significantly) more mobile over the sample period. About 12% of
female person-year observations and 10.4% of male person-year observations were promoted
6
within the firm, while another 15-16% were observed to move across firms. Some 9.5% of
women and just less than 9% of men quit voluntarily, while 7% and 6.3% of women and men
respectively were laid-off layoffs. Thus we find evidence of considerable job mobility over the
sample period, but gender differences are quantitatively small.
The last columns of Tables 1 and 2 show the sample means of the variables used in the
multivariate analysis to follow. Men have greater work experience and job tenure than
women, work longer hours, are more concentrated in skilled-manual and managerial
occupations, and in the private sector, and have more dependent children. But a larger fraction
of women than men with dependent children are out of the labour force or in part-time jobs,
and thus would not be included in our sample.
4. Results
The first three columns of Tables 1 and 2 report the coefficients (and robust standard
errors)6 of a multinomial logit (MNL) regression for men and women, respectively. The
Tables report the results for the mutually exclusive states of “promotion”, “quit”, and “layoff”,
relative to the base of “staying” in the same job with the same employer. The explanatory
variables include: tenure in the current job and its square, labour market experience and its
square, highest educational qualification (4 dummy variables), usual hours of overtime work,7
living in London, marital status, number of children by three age groups (aged 0 to 4; 5 to 11;
and 12 to 16), union coverage, working in the public sector,8 establishment size (2 dummy
variables), occupation (5), occupation of origin (5), cohort of entry into the labour market (4),
travel-to-work time (3), and local unemployment rate. The base for highest educational
qualification is qualification below an ‘Ordinary’ level.9 The base for establishment size is
more than 200 employees, while the base occupational group is semi-skilled and unskilled
7
workers. The base for date of labour market entry is the cohort entering by 1960, while for
travel-to-work time the base is ‘less than 20 minutes’.
TABLES 1 AND 2 ABOUT HERE
We computed the test that Cramer and Ridder (1991) suggest for pooling career states
over the entire sample of men and women, after including a gender dummy and all the
interaction terms of gender with the initial regressors.10 The test for pooling quits and layoffs
yields a likelihood-ratio test statistic of 178.3, which is asymptotically distributed as χ2 with
59 degrees of freedom. After distinguishing quits from layoffs, the Cramer-Ridder test for
pooling promotions and stays yields a value of 963.9, and this is again χ2(59). These tests
suggest that it is inappropriate to pool inter-firm transitions (quits and layoffs), and that
promotion within a firm must be kept separate from staying in the same firm without
promotion.
4.1 Similarities by gender
We find some striking similarities by gender. The probability of being promoted is
significantly higher for male and female workers who are married or cohabiting, and are in
managerial occupations. It is also increasing in the number of hours worked overtime, ceteris
paribus. These findings are consistent with theories viewing promotion as a reward for higher
human capital embedded in higher occupational levels (Sicherman and Galor, 1990), or as a
reward for effort or longer hours of work (Landers, et al., 1996),11 or responsibility in higher-
paying occupational positions (Manove, 1997).
Tenure and experience have a statistically significant negative impact on promotion for
both men and women. The fact that tenure has a significant negative effect on promotion for
both men and women is not surprising (and does not contradict the predictions of human
8
capital theory), because tenure here measures tenure in the job rather than tenure with the
employer.12 The negative relationship between experience and promotion is probably
primarily an age effect: the older (or more experienced) the worker, the less likely is he or she
to be promoted, perhaps reflecting skills obsolescence (see Harper, 1995: Table 6, for a
similar result).
For both male and female workers, highest educational qualifications have no
statistically significant effect on the promotion probability (with the exception of the
vocational qualification for men, which significantly increases the male promotion
probability). However, when we estimated the model excluding current occupation (because it
might produce a spurious association with promotion), our main results remained virtually
unaffected.13 The only notable exception involved the estimated coefficients of the highest
educational qualification variables, which become positively and significantly related to the
promotion rate.14
The quitting probability for both men and women is significantly reduced by the
number of very young children, union coverage, longer job tenure, and being single. The
result that union coverage reduces the likelihood of quitting is consistent with exit-voice
theories, and is a common finding in the union literature (Booth, 1995:199). The negative
correlation between quitting and job tenure is consistent with job matching theories and is also
well established in the empirical literature (McLaughlin, 1991); while the negative
relationship between quitting and number of young children aged up to 4 years may capture
time constraints in job search when young children are present.
Finally, for both men and women, the layoff probability is significantly reduced with
longer job tenure, and managerial occupations. The finding about job tenure suggests that
labour market institutions (e.g., firing costs) are important in regulating firm-initiated
9
separations. For example, many firms adopt a “last-in-first-out” rule. On the other hand,
ceteris paribus, firms tend to keep on workers with greater levels of embodied human capital,
such as men and women in managerial occupations.
4.2 Differences by gender
Despite these similarities, there are some substantial gender differences in the way in which
such variables affect individuals’ job mobility. To illustrate this point more starkly, we
perform the same analysis for the pooled male and female subsamples, and include in the
multinomal logit regression a gender dummy along with the interaction terms of gender with
all the regressors of Tables 1 and 2. To stress the principal differences, Table 3 presents
predicted probabilities of job mobility, which compare the overall distribution of job mobility
at alternative values for selected variables evaluated at the sample distribution of the sample
of male and female workers.15 The third column of Table 3 shows the difference between
female and male mobility rates as a percentage of the corresponding male rate.16 Notice that,
after controlling for all our explanatory variables, the baseline promotion and quit rates do not
differ by gender, but women are still significantly more likely to be laid off than men.
TABLE 3 ABOUT HERE
An extension of union coverage to the entire population of workers is associated with
an increase in men’s and women’s promotion probabilites from 11% to 12.7% and from
10.9% to 11.6% respectively. The increase in the promotion probability for men is 9.4
percentage points higher than that of women, and may offer some support for the long-
standing hypothesis that unions typically look after men better than women (Glucksmann,
1990). Male and female promotion probabilities also increase if they work 5 more hours of
overtime per week, but the female increase is far larger than the male, at about 19 percentage
10
points. Men work longer hours (on average 2.5 overtime hours more than women each week,
see Tables 1 and 2), and a larger proportion of men work overtime (60% of men vs 45% of
women). Thus, women may have a relative advantage in using overtime hours (a signal for
effort) to elicit promotion at both the intensive and extensive margins. Alternatively, this
gender difference may be picking up some unobserved heterogeneity.17 A similar effect is
found for an occupational shift into managerial jobs.
The largest gender gap in promotion probabilities occurs when we modify the
distribution of workers by firm size: moving workers into the smallest establishments would
lead women’s promotion probabilities to be 32% higher than men’s. Larger organisations may
be more likely to operate a well-defined internal labour market with institutionalised career
ladders than smaller firms (Chang, 1996). The differential response by gender is primarily
generated by a substantial reduction in the male promotion probability of about one-fifth of
the baseline probability. This gap may then be explained by the fact that more than 70% of
women in the smallest establishments are in higher occupational positions (professionals,
managers and, particularly, skilled non-manual workers), while only 40% of men in the
smallest establishments are in such occupations.
The most sizable gender differences in quitting probabilities are associated with an
additional child aged 0 to 4 years, and a change in the occupational composition of the labour
force. The first change decreases women’s quitting probability by 11 percentage points more
than men’s. As noted above, the presence of young children is likely to increase the costs of
job search (because of time constraints) or reduce the moticvation to search for a new job
(because of financial benefits such as maternity pay embedded in the current contract):
mothers seem to be more affected than fathers. Moving all workers to the managerial
11
occupations decreases the firm-to-firm mobility of both men and women, but the decrease for
female managers is 14% less than their male counterparts.
Male and female layoff rates respond most differently to recent labour market entry,
the presence of young children, and public sector employment. While a hypothetical shift of
all workers into the public sector reduces the male layoff probability, it leaves the female
virtually unchanged, making women 40 percentage points more likely to be laid-off than men,
a very large effect. An additional child leaves the female layoff probability almost unaltered as
compared to the baseline, but reduces the layoff probability of men by approximately one-
fourth: this generates the 30 percent gender differential shown in Table 3. A larger number of
young children may induce a stronger job attachment or may simply provide better job
protection for fathers.
5. Conclusions
We have examined a special longitudinal sample of full-time British workers observed
between 1991 and 1996 to study gender differences in job mobility, distinguishing between
promotions, quits and layoffs. Our results show that, although women’s promotion and quit
rates are higher than men’s in the raw data, such differences vanish once we control for
standard individual and job characteristics. This contrasts with the popular view that women
quit more often and are promoted less frequently than men. However, women’s layoff rates
remain significantly higher than men’s. Our results also demonstrate that, although average
job mobility rates of men and women in the sample are similar, the rates do respond
differently to specific changes in their socio-economic environment.
These findings emphasise the importance of distinguishing between different forms of
job mobility. The turnover effects of certain variables, such as the presence of young children,
12
union coverage and occupation, differ significantly by gender across all forms of job mobility.
Thus, even with a reasonably homogenous sample of workers, as long as women’s and men’s
career patterns differ in terms of their response to changes in individual or job-specific
characteristics, analyses that focus simply on the overall separation rate and neglect intra-firm
mobility may provide an incomplete picture of workers’ career development and reach
potentially misleading conclusions.
13
Endnotes
1 To our knowledge the only exception is Harper (1995), who uses the National Training Survey (NTS) to
analyse male occupational mobility over the period 1974-195. Harper’s study is not directly comparable to ours
because it does not refer to female mobility, applies to a substantially different time period, and examines
occupational mobility (defined to occur when individuals move from their 3-digit occupation to one of the other
395 occupations recorded in the NTS data) rather than job mobility, which is our primary interest.
2 While most of the literature on promotion incidence uses data from individual firms (e.g., Baker, Gibbs and
Holmstrom, 1994), there are only a few studies of job mobility that use representative samples of workers
(Sicherman and Galor, 1990; McCue, 1996; Groot and van den Brink 1996).
3 However, Booth, Francesconi, and Frank (1998) find a significant gender pay gap for promoted workers over
the period 1991-95, using BHPS data. They conclude that the promotion process does not systematically mitigate
the general disadvantage women face in the labour market.
4 Booth, Francesconi, and Garcia-Serrano (1999) use the retrospective work history data from the BHPS to
examine gender differences in job tenure and job mobility up to 1990. The work history data distinguish between
voluntary and involuntary separations, but do not include any information on promotion.
5 Part-time work occurs when an individual claims to have usually worked less than 30 hours per week on his/her
primary job in the past 12 months. A substantial proportion of British women works part-time. With restriction
(iv), we exclude 797 women (for a total of 4,088 person-periods) and only 37 men (227 person-periods).
However, part-time workers are likely to face different career profiles, with fewer promotion prospects.
Therefore, our restriction to full-time employees is designed to make the male and female samples as comparable
as possible.
6 Given the panel nature of our data, the estimated standard errors account for multiple observations on the same
individual. Consequently, they are robust to arbitrary forms of correlation within each ‘cluster’ or individual.
7 This variable is obtained from the following question: “How many hours overtime do you usually work in a
normal week?”. It refers to normal circumstances in a usual week of work (rather than temporary or exceptional
circumstances), and it includes both paid and unpaid hours of overtime work.
14
8 The definition of public sector includes civil servants and central government employees, local government and
town hall employees, workers in the National Health Service, nationalised industries, higher education and non-
profit organisations.
9 ‘O level’ takes the value unity if the individual’s highest educational qualification was one or more ‘Ordinary’-
level qualifications (later replaced by GCSE), usually taken at the end of compulsory schooling at age 16 years.
‘A-level’ takes the value unity if the individual’s highest educational qualification was one or more ‘Advanced’-
level qualifications, representing university entrance-level qualification typically taken at age 18 years.
‘Vocational qualification’ takes the value unity if the individual’s highest educational qualification was a
vocational qualification (such as Higher National Diploma (HND), Higher Natioanl Certificate (NHC), nursing,
and teaching qualifications), while ‘higher qualification’ takes the value unity if the individual obtained a
university degree or above.
10 In the interests of brevity, we do not report the estimates for this model, but its principal findings are
summarised in Table 3 below.
11 To provide further evidence on the finding that promotion is a reward for effort, we re-estimated the model
with the same regressors as reported in Tables 1 and 2 but excluding current overtime hours, and including
lagged overtime hours of work. For both men and women, the estimated coefficients of this new variable remain
positive, at values of 0.008 and 0.027, respectively. The male estimate, however, is much less precisely measured
yielding a t-ratio of 1.364, while the female estimate remains highly significant with a t-ratio of 3.449. These
estimates should be interpreted with caution, because our sample sizes are reduced after lagging overtime hours
one period. For the regressions with lagged overtime hours, most of the results for the other variables are
unaltered.
12 Furthermore, specific human capital and job matching theories predict a negative effect of tenure on inter-firm
mobility after an initial positive duration effect (see Mortensen, 1986). If different jobs with the same employer
are viewed as different “inspection-goods”, standard job-matching models yield monotone-decreasing job-
separation hazard rates (Mortensen, 1978; Jovanovic, 1979 and 1984). If there is, however, gradual learning
about the quality of the worker-firm match, such models would predict a non-monotonic separation hazard (that
is, an initial rise followed by a decline in the hazard rate). Given the annual frequency of our observations, it is
not surprising that our data can only detect a monotonically declining relationship of job tenure with promotion
rates.
15
13 Of those who have received a promotion from one year to the next, the vast majority remained in the same
occupational group. The proportions of promoted workers who moved across occupations are 23 percent for men
and 16 percent for women.
14 For example, the promotion coefficients (and t-ratios) associated to “Higher qualification” are 0.482 (t-
ratio=2.967) and 0.430 (t-ratio-3.125) for men and women respectively; while those associated to “Vocational
degree” are 0.684 (t-ratio=3.938) and 0.215 (t-ratio=2.841) for men and women, respectively.
15 In particular, let pij denote the predicted probability for individual i (i=1,…N) in outcome j (j=1,2,3,4, our four
mobility states). Because of the MNL assumption, pij=(exp(bjXij))/(1+Σj(exp(bjXij))), where bj is the MNL
parameter estimate and Xij denotes the sample value of the corresponding regressor. Let pj=(1/N)Σipij be the
average predicted probability over individuals. The first two columns of Table 3 report pjm and pjw, that is, the
average predicted probability for men and women respectively, evaluated at sample values of the variables X
used in the estimation.
16 Formally, using the notation of the previous footnote, the values reported in this column are computed using
the following expression (pjw-pjm)×100/pjw. For example, the gender difference in the promotion rate due to union
coverage is given by (0.116-0.127)×100/0.116=-0.011×100/0.116=-9.4 (as Table 3 shows). Furthermore, from
the pooled men-women regression that includes the entire set of gender interactions, the standard errors of each
gender difference are given by the standard errors of each gender-interaction term.
17 For example, women who work longer hours have some unobserved characteristics (e.g., motivation) that make
them more likely to receive a promotion than men working the same number of overtime hours. This should also
be related to the finding that higher levels of past effort (that is, more lagged overtime hours of work)
significantly increase the promotion rate for women but not for men.
16
References
Baker, G., Gibbs, M., Holmstrom, B.1994. “The Internal Economics of the Firm: Evidence
from Personnel Data.” Quarterly Journal of Economics, 109(4), pp. 881-919.
Blau, F.D., Kahn, L.M. 1981. “Race and Sex Differences in Quits by Young Workers.”
Industrial and Labor Relations Review, 34, pp.563-577.
Booth, A.L. 1995. The Economics of Trade Unions. Cambridge University Press, Cambridge.
Booth, A.L., Francesconi, M., Frank, J. 1998. “Glass Ceilings or Sticky Floors?” Centre for
Economic Policy Research, Discussion Paper No. 1965, September.
Booth, A.L., Francesconi, M., Garcia-Serrano, C. “Job Tenure and Job Mobility in Britain”,
Industrial and Labor Relations Review, 53(1), October, pp43-70..
Chang, C., Wang, Y. 1995. “A Framework for Understanding Differences in Labor Turnover
and Human Capital Investment.” Journal of Economic Behavior and Organization,
28(1), pp. 91-105.
Cramer, J.S., Ridder, G. 1991. “Pooling States in the Multinomial Logit Model.” Journal of
Econometrics, 47, pp. 267-272.
Gibbons, R. 1998. “Incentives in Organisations.” Journal of Economic Perspectives, 12(4),
pp. 115-132.
Gibbons, R., Waldman, M. 1999, “Careers in Organizations: Theory and Evidence.” In
Ashenfelter, O. and D. Card (eds.), Handbook of Labor Economics, vol. IIIB,
Amsterdam: New Holland.
Glucksmann, M. 1990. Women Assemble: Women Workers and the New Industries in Inter-
war Britain. Routledge, London.
Groot, W., van den Brink, H.M. 1996. “Glass Ceilings or Dead Ends: Job Promotion of Men
and Women Compared.” Economics Letters, 53, pp. 221-226.
Hall, R.E., Lazear, E.P. 1984. “The Excess Sensitivity of Layoffs and Quits to Demand.”
Journal of Labor Economics, 1984, 2(2), pp. 233-257.
Harper, B. 1995. “Male Occupational Mobility in Britain.” Oxford Bulletin of Economics and
Statistics, 57(3), pp. 349-369.
17
Jovanovic, B. (1979). “Firm-Specific Capital and Turnover.” Journal of Political Economy,
87(6), pp. 1246-1260.
Jovanovic B. (1984). “Wages and Turnover: A Parametrization of the Job-matching Model.”
In G.R. Neumann and N.C. Westergard-Nielsen (eds.), Studies in Labor Market
Dynamics. Berlin: Springer-Verlag.
Landers, R.M., Rebitzer, J.B., Taylor, L.J. 1996. “Rat Rate Redux: Adverse Selection in the
Determination of Work Hours in Law Firms.” American Economic Review, 86(3), pp.
329-348.
Manove, M. 1997. “Job Responsibility, Pay and Promotion.” Economic Journal, 107(1), pp.
85-103.
McCue, K. 1996. “Promotions and Wage Growth.” Journal of Labor Economics, 14(2), pp.
175-209.
McLaughlin, K.J. 1991. “A Theory of Quits and Layoffs with Efficient Turnover.” Journal of
Political Economy, 99(1), pp. 1-29.
Meitzen, M.E. 1986. “Differences in Male and Female Job-quitting Behavior." Journal of
Labor Economics, 4(2), pp. 151-167.
Mincer, J., Ofek, H. 1982. “Interrupted Work Careers: Depreciation and Restoration of
Human Capital.” Journal of Human Resources, 17(1), pp. 3-24.
Mortensen, D.T. 1978. “Specific Capital and Labor Turnover.” Bell Journal of Economics,
9(2), pp. 572-586.
Mortensen , D.T. 1986. “Job Search and Labor Market Analysis.” In O. Ashenfelter and R.
Layard (eds.), Handobook of Labor Economics, vol. II, Amsterdam: New Holland.
Neumark, D. 1988. “Employers’ Discriminatory Behavior and the Estimation of Wage
Discrimination.” Journal of Human Resources, 23(3), pp. 279-295.
Oi, Walter Y (1962) “Labor as a Quasi-fixed Factor”, Journal of Political Economy, 70, 538-
55.
Prendergast, C., Topel, R.H. 1996. “Favoritism in Organizations.” Journal of Political
Economy, 104(5), pp. 958-978.
18
Royalty, A.B. 1998. “Job-to-Job and Job-to-Nonemployment Turnover by Gender and
Education Level.” Journal of Labor Economics, 16 (2), pp. 392-443.
Sicherman, N., Galor, O. 1990. “A Theory of Career Mobility.” Journal of Political Economy,
98(1), pp.169-192.
19
Table 1: Multinomial Logit Estimates of Job Mobility, Men
Types of Job Mobility
Variables Promotion Quit Layoff Means
O level 0.100 0.071 -0.177 0.328(0.136) (0.136) (0.137)
A level 0.207 0.095 -0.176 0.236(0.140) (0.144) (0.150)
Vocational degree 0.399** 0.238 -0.310 0.078(0.177) (0.181) (0.226)
Higher qualification 0.177 0.116 -0.096 0.149(0.168) (0.172) (0.194)
Experience (years) -0.133*** -0.050*** -0.037* 17.805(0.016) (0.016) (0.019)
Experience squared 0.003*** 0.001** 0.001 445.178(0.0004) (0.0004) (0.001)
Tenure (years) -0.094*** -0.576*** -0.399*** 5.617(0.019) (0.045) (0.041)
Tenure squared 0.002** 0.015*** 0.011*** 68.565(0.0007) (0.002) (0.002)
Living in London -0.069 -0.113 0.030 0.092(0.136) (0.147) (0.173)
Number of children:Aged 0-4 -0.148* -0.261*** -0.326*** 0.200
(0.082) (0.089) (0.123)Aged 5-11 -0.123* -0.114 -0.162** 0.321
(0.065) (0.070) (0.082)Aged 12-15 -0.117 -0.039 0.061 0.170
(0.105) (0.103) (0.105)Married or cohabiting 0.597*** 0.422*** 0.245** 0.765
(0.104) (0.102) (0.121)Firm size:
< 50 workers -0.298*** 0.067 0.045 0.395(0.093) (0.100) (0.112)
50-200 workers -0.007 -0.117 -0.223* 0.256(0.093) (0.111) (0.128)
Overtime (weekly hours) 0.007*** 0.001 -0.003 5.970(0.002) (0.005) (0.007)
Union coverage 0.268*** -0.220** 0.011 0.550(0.088) (0.099) (0.103)
Travel-to-work time:20-40 minutes 0.076 0.173* 0.068 0.239
(0.091) (0.095) (0.108)40-60 minutes 0.074 0.195 -0.022 0.115
(0.119) (0.125) (0.146)> 60 minutes -0.121 0.374** 0.071 0.047
(0.187) (0.157) (0.217)Public sector -0.097 -0.147 -0.374*** 0.246
(0.100) (0.116) (0.133)
20
Table 1: (continued)
Current occupation:Professional 0.848*** -0.176 -0.360 0.107
(0.189) (0.195) (0.224)Managerial 1.275*** -0.330** -0.499*** 0.186
(0.156) (0.156) (0.177)Skilled non-manual 0.657*** -0.177 -0.225 0.195
(0.153) (0.142) (0.155)Skilled manual 0.449*** 0.004 -0.051 0.304
(0.144) (0.128) (0.129)Occupational of origin:
Professional -0.222 -0.045 -0.298 0.088(0.188) (0.195) (0.249)
Managerial -0.431*** -0.200 -0.299 0.065(0.184) (0.199) (0.249)
Skilled non-manual -0.051 -0.015 0.001 0.222(0.130) (0.142) (0.151)
Skilled manual -0.063 0.006 -0.126 0.422(0.117) (0.127) (0.124)
Entry cohort in labour market:1961-70 0.124 0.317** 0.070 0.242
(0.148) (0.172) (0.170)1971-80 0.065 0.363** 0.058 0.342
(0.164) (0.180) (0.195)1981-90 0.074 0.376** -0.453** 0.223
(0.147) (0.173) (0.191)After 1990 -0.245 -0.232 -0.592** 0.051
(0.189) (0.197) (0.232)Local unemployment rate 3.203** -3.376** 0.535 0.083
(1.572) (1.691) (1.924)Constant -1.895*** -0.404 -0.425*
(0.247) (0.258) (0.282)
Log likelihood -7192.080Psuedo R2 0.1241
Observed proportion by state (%) 10.37 8.88 6.31Number of observations 9697Note: Estimated coefficients are relative to the state of staying with the same employer without promotion.
Standard errors are shown in parentheses.* significant at 0.1 level** significant at 0.05 level*** significant at 0.01 level
21
Table 2: Multinomial Logit Estimates of Job Mobility, Women
Types of Job Mobility
Variables Promotion Quit Layoff Means
O level -0.259 0.040 -0.019 0.421(0.160) (0.161) (0.169)
A level 0.038 0.305 -0.227 0.176(0.178) (0.184) (0.211)
Vocational degree -0.045 0.125 0.095 0.079(0.220) (0.221) (0.286)
Higher qualification -0.031 -0.298 -0.170 0.140(0.210) (0.234) (0.273)
Experience (years) -0.153*** -0.039* -0.006 14.376(0.019) (0.021) (0.025)
Experience squared 0.003*** 0.001 0.0003 296.678(0.0005) 0.001) (0.001)
Tenure (years) -0.064** -0.735*** -0.527*** 4.431(0.028) (0.064) (0.051)
Tenure squared 0.001 0.022*** 0.017*** 41.598(0.001) (0.003) (0.002)
Living in London -0.031 0.394** 0.190 0.108(0.143) (0.145) (0.178)
Number of children:Aged 0-4 -0.271 -0.487*** -0.195 0.068
(0.186) (0.193) (0.185)Aged 5-11 -0.236** 0.065 -0.066 0.155
(0.125) (0.108) (0.124)Aged 12-15 -0.113 -0.017 -0.056 0.141
(0.118) (0.120) (0.137)Married or cohabiting 0.635*** 0.458*** 0.287** 0.703
(0.107) (0.107) (0.124)Firm size:
< 50 workers 0.107 0.093 0.076 0.436(0.109) (0.127) (0.142)
50-200 workers -0.043 0.028 0.264* 0.263(0.117) (0.138) (0.146)
Overtime (weekly hours) 0.032*** 0.010 (-0.005 3.440(0.007) (0.010) (0.010)
Union coverage 0.100 -0.545*** -0.223* 0.553(0.111) (0.125) (0.130)
Travel-to-work time:20-40 minutes -0.172 0.031 -0.033 0.231
(0.106) (0.111) (0.129)40-60 minutes 0.038 0.224 0.055 0.111
(0.135) (0.149) (0.177)> 60 minutes -0.008 -0.082 0.464* 0.032
(0.214) (0.268) (0.262)Public sector 0.040 0.157 0.010 0.385
(0.118) (0.137) (0.142)
22
Table 2: (continued)
Current occupation:Professional 0.351 -0.125 -0.560* 0.115
(0.309) (0.304) (0.335)Managerial 1.283*** -0.349 -0.737*** 0.141
(0.264) (0.239) (0.273)Skilled non-manual 0.636*** -0.192 -0.314 0.476
(0.249) (0.220) (0.225)Skilled manual 0.070 -0.311 -0.271 0.193
(0.261) (0.227) (0.234)Occupational of origin:
Professional 0.095 0.224 0.488 0.075(0.294) (0.312) (0.365)
Managerial -0.413 -0.220 0.064 0.042(0.292) (0.305) (0.360)
Skilled non-manual 0.372* 0.215 0.142 0.529(0.224) (0.208) (0.218)
Skilled manual 0.261 0.253 0.046 0.276(0.223) (0.205) (0.224)
Entry cohort in labour market:1961-70 0.534*** -0.062 0.280 0.259
(0.187) (0.208) (0.218)1971-80 0.337* -0.205 0.028 0.238
(0.194) (0.215) (0.242)1981-90 0.213 0.062 0.239 0.303
(0.178) (0.193) (0.208)After 1990 -0.248 0.133 -0.260 0.070
(0.215) (0.212) (0.264)Local unemployment rate 2.226 3.575* 3.404 0.083
(1.886) (1.985) (2.131)Constant -1.939*** 0.005 -1.323***
(0.361) (0.357) (0.385)
Log likelihood -4892.738Psuedo R2 0.1254
Observed proportion by state (%) 11.87 9.47 6.99Number of observations 6210Note: Estimated coefficients are relative to the state of staying with the same employer without promotion.
Standard errors are shown in parentheses.* significant at 0.1 level** significant at 0.05 level*** significant at 0.01 level
23
Table 3: Predicted Probability Distributions of Job Mobility, by Gender and SelectedCharacteristics
Mobility Status and Variables Men Women ∆ (%)
PromotionBaseline 0.110 0.109 -1.8Changes:
Number of children 0-4 (+1) 0.096 0.100 4.8**Union coverage 0.127 0.116 -9.4***Public sector 0.100 0.115 14.4Weekly overtime hours (+5) 0.113 0.135 18.5***Experience (+1 year) 0.093 0.104 11.8***Job tenure (+1 year) 0.088 0.102 15.9***Managerial occupation 0.179 0.213 19.1***Entry after 1990 0.108 0.107 -1.9Firm size: < 50 workers 0.087 0.115 31.6***Local unemployment rate (+1%) 0.121 0.112 7.5*
QuitBaseline 0.092 0.089 -3.9Changes:
Number of children 0-4 (+1) 0.075 0.067 -10.8***Union coverage 0.079 0.076 -4.1***Public sector 0.084 0.103 23.5Weekly overtime hours (+5) 0.092 0.090 -2.1Experience (+1 year) 0.079 0.084 6.3***Job tenure (+1 year) 0.061 0.058 -3.9***Managerial occupation 0.076 0.087 14.4**Entry after 1990 0.090 0.092 2.2Firm size: < 50 workers 0.095 0.091 -4.5Local unemployment rate (+1%) 0.086 0.091 6.3**
LayoffBaseline 0.064 0.068 -4.9**Changes:
Number of children 0-4 (+1) 0.049 0.064 30.5***Union coverage 0.064 0.066 3.2*Public sector 0.049 0.069 40.3***Weekly overtime hours (+5) 0.063 0.067 4.7Experience (+1 year) 0.060 0.068 12.7*Job tenure (+1 year) 0.050 0.052 2.2***Managerial occupation 0.044 0.045 2.5***Entry after 1990 0.056 0.067 19.6***Firm size: < 50 workers 0.065 0.069 5.1Local unemployment rate (+1%) 0.064 0.072 6.7
Note: Predicted probabilities are obtained from pooled (men and women) MNL estimates with full set ofinteractions, equivalent to those reported in Tables 1 and 2. ∆(%) is the percentage change of thefemale predicted probability with respect to the male predicted probability relative to the female rate.See text for detailed explanation and example. Baseline values are computed at sample values.* 0.05 < p < 0.10** 0.01 < p < 0.05*** p < 0.01