flexible contracts and subjective well-being contracts and... · flexible contracts and subjective...
TRANSCRIPT
Flexible Contracts and Subjective Well-being
Colin P. Green
Department of Economics, Lancaster University
Lancaster LA1 4YT, UK
John S. Heywood
Department of Economics, University of Wisconsin-Milwaukee and Birmingham
Business School, University of Birmingham
Abstract
If workers choose between permanent and flexible contracts, utility should equalize.
Estimates of job satisfaction using the British Household Panel Survey reveal the
critical role played by unmeasured worker heterogeneity. Accounting for such
heterogeneity, shows that flexible contracts are a strong negative determinant of
satisfaction with job security but are often a positive determinant of other dimensions
of job satisfaction. As a consequence, flexible contracts have either a weak negative
influence or no influence at all on overall job satisfaction, a finding broadly consistent
with equalizing differences. This consistency emerges even more dramatically in
estimates of overall life satisfaction in which flexible contracts have generally no
influence.
KEYWORDS: Flexible Contracts, Job Satisfaction
JEL CODES: J28, J41
Corresponding Author: Colin P. Green
Lancaster University
Phone: +44 (0) 1524 594667
Fax: +44 (0)1524 59244
1
Flexible Contracts and Subjective Well-being
1. INTRODUCTION
This paper estimates the influence of flexible -- more contingent -- employment
contracts on job satisfaction and life satisfaction. The hedonic theory of the labor
market provides no a priori prediction for the influence of such contracts on job
satisfaction (on-the-job utility). Workers sort across contract types maximizing their
utility and to the extent that flexible contracts contain undesirable characteristics this
process generates either compensating earnings differentials or other offsetting
desirable characteristics (Rosen 1974). As a single illustration, flexible contracts
exhibit greater job insecurity (Blanchard and Landier 2002) but job insecurity
generates higher compensating earnings (Abowd and Ashenfelter 1981). Thus, the
net influence of flexible employment contracts remains an empirical question.
We estimate the determinants of job satisfaction to summarize these
potentially conflicting outcomes. Hamermesh (2004) emphasizes that economists
studying job satisfaction should attempt to test theoretical predictions about worker
behaviour and/or labour market functioning. As Hamermesh (2001, p. 2) puts it, job
satisfaction is the only measure "that might be viewed as reflecting how (workers)
react to the entire panoply of job characteristics" and as such "it can be viewed as a
single metric that allows the worker to compare the current job to other labour market
opportunities." Thus job satisfaction measures allow a summary worker evaluation of
the consequences of flexible employment contracts.1 If such contracts consistently
associate with large negative job satisfaction differentials, it becomes more difficult to
argue for a hedonic or equalizing differences view of this aspect of the labour market.
Yet, even in the presence of large job satisfactions differentials, it may be the case that
2
workers in flexible contracts have traded off lower job satisfaction for higher
satisfaction in other aspects of life such as fulfilling family responsibilities,
undertaking education or arranging leisure activities. Examining overall life
satisfaction allows testing the influence of flexible contracts on well-being in a
framework that accounts for these potential trade-offs as well as the equalizing
differences that happen within the labour market.
This paper provides evidence on the relationship between a variety of flexible
employment contracts and job satisfaction. This is provided across a range of
dimensions of job satisfaction: job security, pay, hours and the nature of the work
itself. Moreover, by estimating a fixed effects ordered probit on our panel, we provide
the first estimates of flexible work’s influence on job satisfaction that are robust to
sorting across a variety of employment contracts. By also bringing this estimation to
the determinants of life satisfaction, we provide evidence on whether flexible
contracts may be associated with off-setting benefits outside the realm of work.
In sharp contrast to the broad trend in the existing literature (for example,
Booth et al. 2002; D'addio et. al 2004; de Graaf-Zijl 2005a), we find that when
controlling for worker heterogeneity, flexible contracts play at best a minor role in
diminishing job satisfaction and this role is limited to diminishing satisfaction with
job security. Moreover, we find that when controlling for worker heterogeneity,
flexible contracts play no role in diminishing overall life satisfaction among the
employed. In essence, our results are broadly supportive of labour markets generating
equalizing differences for flexible contracts.
In what follows, the next section assumes that both flexible and traditional
contracts are offered in the same labor market as a device to discuss predictions about
the relationship between flexible contracts and job satisfaction. The third section
3
presents our data and testing methodology. The fourth section presents the initial
results followed by more detail and further estimates in the fifth section. The sixth
section concludes.
2. FLEXIBLE CONTRACTS AND JOB SATISFACTION
Flexible employment contracts have become increasingly prevalent among OECD
countries (Mangan 2000).2 This increasing prevalence reflects changes in labour
market regulation, technological change and increasing female labour force
participation. While some observers may characterize flexible contracts as wholly
beneficial or detrimental for employees, a balanced reading presents mixed
consequences. For instance, flexible employment contracts are associated with lower
levels of employer provided training (Arulampalam and Booth, 1998, Draca and
Green, 2004), higher risk of social exclusion for msen (Addio and Rosholm, 2005),
lower wages in the UK (Booth et al, 2002) and increased job insecurity (Blanchard
and Landier, 2002). At the same time, flexible employment contracts may present a
means to gain entry into permanent employment (when compared to unemployed job
search) both in general (Guell and Petrongolo, 2001; Addison and Surfield 2006), and
via promotion within the firm (Green and Leeves, 2004). In addition, increased
employment flexibility may lead to overall higher employment and participation rates
(Lazear, 1990) and serve as a critical bridge for those displaced by large-scale plant
closing or layoffs (Farber 1999). Finally, flexible contracts may reflect the appeal of
more contingent employment to workers who desire flexible schedules (Morris and
Vekker 2001).
The basic economics of flexibility focuses on the coordination of worker
effort. Deardorff and Stafford (1976) make clear that both firms and workers have
4
preferences over the direction of this effort. The firm prefers that workers be flexible
allowing it to coordinate effort across workers in the cheapest fashion (for example
having all workers show up for the same scheduled shift with on call workers to fill
absences). On the other hand, workers prefer that the firm be flexible allowing them
to work when it best fits their schedules (yields the most utility). Indeed, the term
flexibility has actually been used to characterize both of these extremes in the
contract. Thus, when the firm is being flexible it is often identified as providing a
family friendly work practice such as "flextime" (Heywood et al. 2007). Yet, when
the worker is being flexible as happens with short intensive hires or on call and
agency relationships it is identified as a "flexible staffing arrangement" (Gramm and
Schnell 2001; Houseman 2001). The probability that the firm's cost minimizing work
arrangement and the workers' utility maximizing work arrangement coincide is
essentially zero.3 Thus, competition in product and labor markets should generate a
wage that is higher if the agreed upon arrangement more closely mirrors that desired
by the firm and that is lower if the agreed upon arrangement more closely mirrors that
desired by the workers.
Given the natural heterogeneity in the cost for firms to provide flexibility to
workers, a hedonic equilibrium should develop in which employers with higher costs
in providing workers flexibility retain flexibility for themselves (flexible staffing
arrangements) and pay higher wages (Duncan and Stafford 1980). Those employers
with lower costs give up their flexible staffing arrangements, providing staffing
arrangements that are more beneficial to workers and pay lower wages. Thus,
economic theory predicts that compensating earnings differences should emerge that
offset the disadvantage of a flexible staffing contracts such as seasonal work, having a
fixed term or doing agency work.
5
This prediction receives support from a variety of empirical studies. Both
Moretti (2000) and Del Bono and Weber (2008) show that otherwise equal workers
earn significantly higher wages when working on seasonal jobs compared to similar
permanent jobs. De Graaf-Zijl (2005b) confirms a positive compensating wage
differential for on-call workers in the Netherlands. Such findings fit with the detailed
examination of worker valuations confirming a marginal willingness to pay for a
reduced risk of unemployment (Van Ommeren and Hazans 2007). They also fit with
evidence from the provision of family friendly work practices. Baughman et al.
(2003) show that US employers that are unable to provide scheduling freedom to
employees pay significantly higher entry wages. Using representative UK data,
Heywood et al. (2007) demonstrate the existence of sizable negative wage
differentials both for more generous leave policies and for providing employees
choice over working hours. Thus, when the firm retains flexibility in assigning work
effort, wages are higher, and when workers gain favourable flexible arrangements
wages are lower.
Yet, the fundamental concern of workers may be that the firm's flexibility
includes greater termination rights. In short, workers enjoy far less job security and it
is known that workers value job security all else equal (Theodossiou and Vasileio
2007). However, the basic insight made by Deardorff and Stafford (1976) remains. If
the firm retains the flexibility to terminate more easily, the earnings required to
encourage a worker into such an arrangement must be greater. This receives
empirical support from Abowd and Ashenfelter (1981), Li (1986), Heywood (1989)
and Assaad and Tunali (2002) who each show that in representative samples higher
unemployment risk is compensated for with higher earnings.
6
While such empirical evidence does not suggest that these relationships should
be taken for granted, it does suggest that both firms and workers value flexibility and
that the possibility of a hedonic market in which this flexibility is exchanged should
be considered. Thus, our initial suggestion from theory is that there is no a priori
prediction that job satisfaction, as a measure of on-the-job utility, should differ
between otherwise equal workers in traditional and in "flexible staffing" contracts.
While workers may not like such contracts, they should be compensated either by
earnings or other work dimensions for taking them.
Having drawn this suggestion we are quick to emphasize that it follows from
an assumption of active competition between workers and firms across different types
of contracts. Certainly, Polivka (1996) reports that many contingent workers in the
US prefer such positions. Yet, if workers in flexible staffing contracts are not able to
find alternative contracts and are somehow crowded into a limited labour market that
consists of only flexible contracts, this suggestion is unlikely to hold. Instead, it may
emerge that flexible staffing contracts are associated with lower job satisfaction.
Beyond informing the theory, this would be of interest in its own right as consistently
lower job satisfaction is associated with the intention to quit a worker's current
employment situation (Antecol and Cobb-Clark 2006, 2008 and Clark 2001).
Researchers in management have explored this issue but have not used either
representative surveys of the workforce or the established set of job satisfaction
determinants common in work by economists. Moreover, the conclusions from this
work are mixed with flexible staffing arrangements associated with greater satisfaction
in some studies and reduced satisfaction in others (Connelly and Gallagher 2004).
What does emerge, and what fits with our emphasis on the role played by active
competition, is the importance of worker volition. Simply put, workers express greater
7
job satisfaction when they view their contingent work as a choice rather than resulting
from a lack of alternatives (Krausz et al 1995; Ellingson et al 1998).
Finally, there exists within some policy circles a sense that economics has been
too fixated on the benefits of flexible contracts to firms without considering the position
of workers. Thus, Guest et al. (2007) state "in the UK much of the research and policy
on new forms of employment has been dominated by labour economists who tend to
under-emphasise the subjective expectations and the experiences of employees." We
hope to redress this imbalance while striving for the standard suggested by Hamermesh
(2004) using representative job satisfaction data to test the general prediction that labor
markets are sufficiently competitive and generate sufficient worker choices that the net
employee benefits associated with flexible staffing arrangements are similar to those
associated with more traditional contracts.
3. DATA AND METHODOLOGY
The data used in this paper are drawn from the British Household Panel Survey
(BHPS). The BHPS is a nationally representative sample that each year interviews
approximately 10,000 individuals from roughly 5,500 households. We use the waves
of the BHPS corresponding to 1999-2004, as earlier waves do not contain detailed
information on types of flexible employment contracts.4 Within these waves, we
identify those with a fixed term contract, those who work for a temporary agency, and
those with other flexible contracts including casual workers, seasonal workers and
others. Table A1 presents summary statistics for males and females separately.
Briefly, women are more likely to be employed under flexible working arrangements,
although the gender differences for some contract types, such as agency-based work,
are not large.
8
In Table A2, we provide summary statistics for selected covariates split
according to contract type. Across a number of dimensions workers on fixed term
contracts are broadly comparable to permanent employees. Other flexible workers
have markedly lower average weekly wages and are less likely to be unionised, hold a
position with a managerial or supervisory role or work in a large firm when compared
to permanent employees. Agency workers are the youngest.
All job satisfaction questions in the BHPS are reported on a 7 value Likert
scale, 1 being the least satisfied, 7 the most satisfied. At different times a variety of
job satisfaction questions have been included in the BHPS but for the period in which
the more detailed flexible contract type information is available, five job satisfaction
questions are available. These include overall job satisfaction, satisfaction with pay,
satisfaction with hours worked, satisfaction with job security, and satisfaction with the
work itself. We restrict our sample to those individuals aged 20 to 65 and exclude the
self-employed and those with missing data. This yields an unbalanced panel of 11,433
individuals.
Table 1 presents the mean satisfaction level for overall satisfaction and each of
the four dimensions of job satisfaction. It does so for workers on permanent contracts
and each of the flexible contracts. As the averages make clear, the workers on
permanent contracts do not routinely report the greatest satisfaction. Out of the four
contract types, they rank second in satisfaction with pay, third in satisfaction with
hours and third in satisfaction with the work itself. Workers on permanent contract do
rank first in satisfaction with job security and their advantage in satisfaction in this
dimension is enormous.
Following past research, the values of job satisfaction are fitted to the
cumulative normal distribution through ordered probit estimates (see Clark and
9
Oswald 1996 and Clark 1997 among many others). The ordered probit estimation
follows appropriately when the dependent variable has a natural ordering, such as
from least to most satisfied (see McKelvey and Zavoina 1975 for details) and can be
used to predict the probability of reporting each value for job satisfaction for variation
in the values of the independent variables.
4. INITIAL EVIDENCE
This section presents the initial estimates of the relationships between contract type
and job satisfaction. Throughout the paper estimates are presented separately for
males and females. Separating the estimates follows from both the routine finding of
separate job satisfaction regimes for men and women (Clark 1997) and from tests
within our own sample rejecting the hypothesis of a common set of coefficients.
INSERT TABLE 2
Table 2 presents pooled ordered probit estimates of the impact of flexible employment
contracts on job satisfaction split by gender with the errors clustered by individual.
Only standard controls for personal characteristics are included allowing all job
characteristics including earnings to vary. Fixed term contracts provide insignificantly
different job satisfaction compared to permanent contracts. Agency contracts provide
significantly less satisfaction than do permanent contracts. Other flexible contracts are
associated with lower job satisfaction for men but not for women. Thus, in a simple
specification that holds constant only worker characteristics, the evidence on the role
of flexible contracts is mixed suggesting that in at least some cases other job
10
characteristics including earnings may vary to compensate if there are undesirable
characteristics.
INSERT TABLE 3
We next introduce a full range of job characteristics of the type typically
included in the estimation of job satisfaction. We do this in two waves as indicated in
Table 3. The variation in the estimates between model 2, with a partial set of job
characteristics and model 3 with our full set of job characteristics is fairly minimal.
The critical point is that more significant negative coefficients emerge on flexible
contracts when compared with the parsimonious specification. First, note that all
twelve coefficients on contract type emerge as negative across models 2 and 3.
Moreover, instead of only the coefficient on agency work being significant for
women, all three types of flexible contracts take significant coefficients. Such a result
would be anticipated if other characteristics compensate for undesirable aspects of
flexible contracts. In other words, once hours, earnings and a host of job
characteristics are held constant, flexible contracts appear less desirable but allowing
them to vary they are not less desirable. As a note, it is clear both the worker and
workplace characteristics broadly conform to the signs and significance reported in
numerous other studies using the BHPS (see for instance Clark and Oswald, 1996;
Green and Heywood, 2008).
Critically, workers are likely to sort into flexible contracts in a non-random
manner. Hence, the relationship observed between job satisfaction and flexible
working arrangements may be biased by unobservable characteristics that influence
both the propensity to be in flexible working contracts and the level of job
11
satisfaction. At an extreme, inherently unsatisfied workers may sort into flexible
contracts rather than permanent contracts lowering the apparent job satisfaction.
Alternatively, unmeasured worker characteristics such as lower initiative or ability
may be correlated both with lower job satisfaction and with working in flexible
contracts. 5
We investigate these possibilities by re-estimating individual fixed effects
ordered probit versions of the models reported in Table 4. These estimates follow
from a routine dedicated to this purpose in Limdep 8.0 and the full results are
available from the authors upon request. We have estimated the results for all three of
our models: controlling only for worker characteristics, model 1, and controlling for
job characteristics, models 2 and 3.
INSERT TABLE 4.
The estimates depend only on the variance within worker across years. In other
words, they follow from observing a given worker changing status into and out of
flexible contracts and examining the change in job satisfaction. For males, these
estimates suggest that the negative effects of flexible contracts are largely eliminated.
The failure to find statistical significance (with one exception) results from both the
anticipated increase in standard errors and substantial declines in the point estimates.
Similarly, the earlier suggestion that females on fixed term contracts and agency work
may have lower job satisfaction is not supported by the fixed effects estimates.
However, those on other flexible contracts emerge with statistically significant lower
satisfaction. In short, looking at Table 4 there is only weak evidence, at best, of the
strong negative effect of flexible contracts on job satisfaction that is often claimed.
12
The pattern shown in Table 4 is not the result of a selected sample being used
to obtain the fixed effect estimates. Critically, the same pattern of statistical
significance emerges as did in Tables 2 and 3 if one limits the sample to workers
showing variation in contract status and so determining the fixed effect estimates. A
variation on this concern is that there are unmeasured factors correlated with changing
contract type that also correlate with changes in job satisfaction. Thus, it may be that
those leaving permanent jobs have unusually poor permanent jobs and that those
leaving more flexible jobs have unusual good flexible jobs (those with a chance for
advancement). This would make results based on movers potentially biased
downward. While the sample of movers looks similar in a cross-section estimate, we
tried to directly investigate this particular conjecture. In a sample of only workers in
permanent jobs we found modestly lower job satisfaction among those who would
move to temporary jobs. Yet, in a sample of workers only in temporary jobs, those
who would move to permanent jobs were no more satisfied. These results are
available from the authors on request. While not unequivocal, the sum of our
investigations, leave us reasonably confident in the fixed-effect estimates.
One might also be concerned that due to the small numbers of movers between
contract types that we simply have imprecise estimates and large standard errors. We
investigate this by re-estimating our three models using one aggregated variable,
flexible contracts. The resulting fixed effects estimates (again, available from the
authors) continue to show that flexible contracts have no effect on male’s job
satisfaction (point estimates between -0.03 to 0.02). They also continue to show some
support for a negative effect of flexible contracts on female job satisfaction (from -
0.133 and just barely significant at the 10 percent level in model 1 to -0.119 and not
significant in model 3).
13
We now exploit the dimensions of job satisfaction controlling for individual
fixed effects. These results are shown in Table 5 for model 3 that includes the full set
of controls. In three of the dimensions of job satisfaction, the influence of flexible
contracts appears to be broadly positive. In estimating satisfaction with pay, all six
coefficients are positive and two are statistically significant. In estimating satisfaction
with hours five of the six coefficients are positive and four of the five are statistically
significant. In estimating satisfaction with the work itself, again, five of six
coefficients are positive with two of the five are statistically significant. Across all
eighteen estimates for these three dimensions, the only negative and significant
coefficient is that for other flexible contracts in the female satisfaction with hours
estimate. We emphasize that this pattern emerges almost entirely because of the fixed
effects estimate and so emerges across all three specifications. Indeed, if one returns
to our parsimonious first model, the signs and significance for satisfaction with pay
and with hours remain as presented. In the estimates for satisfaction with the work
itself, two additional coefficients emerge as statistically significant, one positive and
one negative. In short, there remains a broad pattern suggesting, if anything, a general
positive influence of flexible contracts on these dimensions of jobs satisfaction.
INSERT TABLE 5
This broad pattern contrasts dramatically with the routinely large and negative
coefficients on all six estimates of satisfaction with job security. Each coefficient
emerges as statistically significant in all three specifications and both with and
without controlling for fixed effects. It seems very clear that flexible contracts of any
type act as associated with reduced satisfaction with job security. Thus, a critical
14
question becomes how important is job security as a component of overall job
satisfaction.
5. EXTENSIONS
This section provides two major extensions to the results already provided. First, we
examine in detail the relationship between flexible contracts and satisfaction with job
security. We present estimates suggesting that if flexible contract jobs had the same
security as permanent jobs, they would be largely preferred to permanent jobs. This
finding is consistent with the suggestion of equalizing differences. Second, we
explore the issue of whether some of the potential benefits workers see to flexible
contracts accrue outside the realm of work by examining overall life satisfaction.
These estimates suggest that flexible contracts do not diminish life satisfaction among
the group of employed workers.
5.1 Flexible Contracts and Satisfaction with Security
A typical way to determine the importance of various job dimensions in overall job
satisfaction includes the dimensions of satisfaction into an estimate of overall job
satisfaction. Clark (1997) used roughly such a methodology to show that men and
women differed in their valuation of job dimensions. In examining Dutch workers, de
Graaf-Zijl (2005a) follows work by Van Praag et al. (2003) estimating overall job
satisfaction as a composite of various dimensions of job satisfaction. In her estimates
she identifies job content (the work itself) as contributing the most to overall job
satisfaction and finds that job security contributes the least to overall job satisfaction.
Intermediate positions were held by satisfaction with pay, hours, working conditions,
15
working times and commuting distance. Importantly, the relative unimportance of
satisfaction with job security remains even for Dutch workers on flexible contracts.
We return to our estimates of overall job satisfaction to explore which
dimensions by sequentially adding a single dimension of satisfaction (satisfaction
with pay, with hours etc) as a control. This generates four estimates each with a
different dimension added. We emphasize that the object is to control for satisfaction
with that dimension to see what independent variation remains to be explained by the
critical flexible contract variables. We, like those who have previously estimated
similar models, recognize that the dimension of satisfaction and overall satisfaction
are surely endogenous and both may be determined by other controls. We are not
interested in the point estimate on the dimensions of job satisfaction but rather what
their inclusion does to the role of contract types.
We summarize the fundamental result with Table 6. Satisfaction with job
security is included in the full specification without fixed effects. This specification
generated the most negative and significant coefficient as shown in Table 2. Yet, the
inclusion of the job security measure causes a dramatic reversal. In short, if one holds
constant satisfaction with job security, each of the three types of flexible contracts
emerge as significant positive determinants of overall job satisfaction for both
genders. Moreover, this result emerges from any of the three specifications we
estimate and either with or without worker fixed effects. Equally important, this result
is not simply a function of including any dimension of job satisfaction. It is unique to
job security. Including any of the other three dimensions of job satisfaction does not
change the pattern shown in Table 2 and 3. There remains a predominance of
negative coefficients on the flexible contract variables several of which are
statistically significant for each gender.6
16
INSERT TABLE 6
The fundamental point is that lower satisfaction with job security generates the
lower overall satisfaction observed in the estimates that do not control for fixed
effects. This dramatically differs from the Dutch results discussed earlier in which
satisfaction with job security played little or no role in generating the lower
satisfaction observed in flexible contracts. Similarly, there is no evidence in our UK
data that dissatisfaction with the work itself generates overall dissatisfaction.
Including this dimension of satisfaction as a determinant of overall satisfaction does
not dislodge the negative influence of flexible contracts as would be implied by the
Dutch results.
Thus, the fixed effects results suggest an intriguing possibility. If it is lower
satisfaction with job security that drives the appearance of lower overall job
satisfaction, the absence of strongly lower overall job satisfaction in the fixed effects
estimate (Table 4) suggests sorting on job security. Workers with lower inherent job
security, greater likelihood of separation, may be sorted into flexible contracts. While
the extent of this sorting is not sufficient to eliminate the influence of flexible
contracts on satisfaction with job security, it is sufficient to eliminate their influence
on overall job satisfaction. This empirical result remains consistent with the theory of
equalizing differences. Flexible contracts lower satisfaction with job security but tend
to have a broadly positive influence on other dimensions of satisfaction (Table 5).
The net effect is that overall satisfaction does not routinely differ for flexible contracts
(Table 4). This pattern is confirmed by recognizing that when satisfaction with job
17
security it held constant, flexible contracts have a positive influence on overall job
satisfaction (Table 6).
5.2 Flexible Contracts and Life Satisfaction
The consequences of flexible contracts can extend beyond their influence in changing
on-the-job utility. Workers may trade off lower job satisfaction for higher satisfaction
in other aspects of life such as fulfilling family responsibilities, undertaking education
or arranging leisure activities. We examine this by estimating the impact of flexible
employment contracts on individuals’ overall life satisfaction. The life satisfaction
question in the BHPS is `How satisfied/dissatisfied are you with your life overall?’ to
which respondents provide a 1 to 7 Likert scale response, where 7 is the highest
satisfaction. This variable has been used extensively in the literature on happiness and
life satisfaction (for instance Clark 2006; Oswald and Powdthavee 2008) to capture
the consequences of major labor market changes such as unemployment, illness and
retirement.
We estimate two broad groups of specifications. First, we estimate ordered probits of
life satisfaction for all workers aged 20 to 65 in the BHPS, where the control variables
are those in Table 2 with the addition of a control for whether the individual was
employed. Second, we estimate ordered probits of life satisfaction for the sample of
all respondents (including the unemployed and those out of the labour market), where
the controls are the same as in model 3 (see table 3). We estimate these for males and
females, with and without controls for individual fixed effects.
INSERT TABLE 7
18
As the top panel in table 7 shows, the cross-sectional estimates reveal a wide variety
of negative partial correlations between flexible contracts and life satisfaction. This
might suggest that having a flexible contract lowers life satisfaction. Yet, controlling
for the worker fixed effects causes every significant coefficient to vanish and reveals
more positive coefficients than negative. This suggests that the workers moving
between contracts are largely unaffected by contract type.
The results in the second panel show that for the sample of all respondents,
employment is associated with a large increase in life satisfaction, an increase that is
equally evident in both the cross-section and fixed effect estimates. The cross-section
results show conditional upon being employed both men and women in agency
contracts report lower life satisfaction. The fixed effects estimate causes the male
coefficient to fall in half and no longer be significant while it causes the female
coefficient to switch sign revealing a large positive increment associated with agency
work. The female coefficient for other flexible contracts is now also positive and
statistically significant. While the fixed effect estimates show a significant negative
coefficient for males on other flexible contracts, the overall pattern continues to
suggest the role of equalizing differences.
6. CONCLUSION AND DISCUSSION
We began with a suggestion that if most workers have a choice between permanent
and flexible contracts, wage differentials and other equalizing differences will create
roughly similar on-the-job utility between the two types of contracts. We tested this
hypothesis using data from the BHPS and parsimonious pooled cross-section
specification that included only basic worker characteristics. The results emerge as
19
mixed with agency work associated with lower overall satisfaction for both genders
and other flexible contracts associated with lower overall satisfaction for men. As
more controls were added to these estimates, the negative role of flexible contracts
appeared stronger.
A second set of specifications controlled for individual worker fixed effects.
This caused two changes. First, there was far less evidence of a negative influence on
overall job satisfaction from flexible contracts. Only other flexible contracts had a
routinely negative influence and this was only for women. Some coefficients were
positive and the overwhelming majority of all coefficients were insignificant. It might
be tempting to blame the insignificance on the imprecision associated with such
estimates, but the point estimates were typically much smaller. Moreover, the fixed
effects estimates run separately by dimension also suggest imprecision is not the
issue. Those estimates show that flexible contracts had broadly positive influences on
three of the four dimensions of job satisfaction and sharply negative influence on
satisfaction on job security. In other words, the lack of a persuasive result in the
overall satisfaction resulted from offsetting influences on the disaggregate
dimensions. This was further supported by evidence that holding constant satisfaction
with job security, all three types of flexible contracts had positive influences on
overall job satisfaction for each gender. Finally, we showed that flexible contracts
are generally not associated with diminished life satisfaction in estimates that control
for worker or respondent fixed effects.
Thus, at a minimum, we think it is too early to dismiss the notion that
equalizing differences may exist between permanent and flexible contracts.
Obviously, more work could be done. First, it is surely the case that the type of
contract represents a choice both by firms to offer them and by workers to accept
20
them. Accounting for this choice proves difficult. In our case we would have a
choice across four sectors (permanent, fixed term, agency and other) that would need
to be imbedded in an ordered probit estimation. One might still be worried about
worker fixed effects and one would, even then, still have a reduced form of the worker
and firm choice processes. We are aware of only one study that attempts to model
selection (Beckmann et al 2007). To accomplish this they linearize the ordinal
satisfaction variable and limit the choice modelling to two sectors, permanent or fixed
term. Nonetheless, it is interesting to note that they find that, if anything, fixed term
contracts are associated with greater satisfaction. Second, we have not identified
which characteristics might play an equalizing role. Simple experiments with changes
in specification do not identify immediate candidates. We removed wages from many
of our later estimates and the results did not markedly change. While more detailed
experiments may be useful, we emphasize that much of the statistical traction comes
from accounting for worker fixed effects. Third, while we have suggested that our
results may differ from those in more regulated labor markets, further comparative
studies are needed to draw a firm conclusion. Despite these limitations, it remains
the case that in our preferred fixed effect estimates, offsetting influences on the
dimensions of job satisfaction emerge and overall satisfaction (either job or life) is not
routinely lower in flexible contracts.
21
REFERENCES
Abowd, J. and O. Ashenfelter, O. 1981. Temporary Layoffs, Anticipated
Unemployment and Compensating Wage Differentials” in S. Rosen ed. Studies in
Labor Markets. Chicago: University of Chicago Press, 141 - 70.
Addio, C., Rosholm, M., 2005. Exits from temporary jobs in Europe: A competing
risks analysis. Labour Economics 12, 449-468
Addison, J.T. and Surfield, C. 2005. Atypical work and Compensation, IZA
discussion paper, 1477, Bonn Germany.
Addison, J.T. and Surfield, C. 2006. The Use of Alternative work Arrangements by
the Jobless: Evidence from the CAEAS/CPS. Journal of Labor Research 27, 149 –
62.
Antecol, H., Cobb-Clark, D., 2008. Racial Harassment, Job Satisfaction and
Intentions to Remain in the Military. Journal of Population Economics,
Forthcoming.
Antecol, H. and Cobb-Clark, D., 2006. The Sexual Harassment of Female Active-
Duty Personnel: Effects on Job Satisfaction and Intentions to Remain in the
Military. Journal of Economic Behavior and Organization 61, 55 – 80.
Arulampalam, W. and Booth, A., 1998. Training and labour market flexibility: Is
there a trade-off. British Journal of Industrial Relations 36, 521–36.
Assaad, R. and Tunali, I., 2002 Wage Formation and Recurrent Unemployment,
Labour Economics, 9, 17-61.
Baughman, R., DiNardi, D. and Holtz-Eakin D., 2003. Productivity and Wage Effects
of "Family-Friendly" Fringe Benefits. International Journal of Manpower 24, 247-
59.
22
Beckmann, M, Binz, A. and Schauenberg., 2007. Fixed-term Employment and Job
Satisfaction: Evidence from Individual-level Data Accounting for Selectivity Bias,
WWZ Working Paper No. 03/07.
Blanchard, O. and Landier, A., 2002. The Perverse Effects of Partial Labour Market
Reform: Fixed–Term Contracts in France. Economic Journal 112, F214-244.
Booth, A., Francesconi, M., Frank, J., 2002. Temporary Jobs: Stepping Stones or
Dead Ends? Economic Journal 112, F189–F213.
Clark A. 2006. "A Note on Unhappiness and Unemployment Duration”. IZA
Discussion Paper No. 2406.
Clark A. 2001. "What Really Matters in a Job? Hedonic Measurement Using Quit
Data. Labour Economics 8, 223 – 42.
Clark, A., 1997. Job Satisfaction: Why are Women So Happy at Work? Labour
Economics 4, 341 – 72.
Clark, A., Oswald, A., 1996. Satisfaction and comparison income. Journal of Public
Economics 61, 359 – 81.
Connelly, C.D., Gallagher, D.G., 2004. Emerging Trends in Contingent Work
Research. Journal of Management 30, 959 – 83.
D'Addio, A.C., Eriksson, T. and Frijters, P. 2007. An Analysis of Job Satisfaction
when Individuals' Baseline Satisfaction Levels Differ. Applied Economics 39:
2413 – 23.
De Graf-Zijl, M. 2005a. The Anatomy of Job Satisfaction and the Role of Contingent
Employment Contracts. Tinbergen Institute Discussion Paper 119/3, University of
Amsterdam.
23
De Graf-Zijl, M., 2005b. Compensation of On-Call and Fixed-Term Employment:
The Role of Uncertainty. Tinbergen Institute Discussion Paper 120/3, University
of Amsterdam.
Deardorff, A. and Stafford, F. P., 1976. Compensation of Cooperating Factors.
Economica 44, 671 – 84.
Del Bono, E. and Weber, A., 2008. Do Wages Compensate for Anticipated Working
Time Restrictions? Evidence from Seasonal Employment in Austria. Journal of
Labor Economics 26, 181-221
Draca, M. and Green, C., 2004. The Incidence and Intensity of Employer Funded
Training: Australian Evidence on the Impact of Flexible Work. Scottish Journal of
Political Economy 51, 609-625.
Duncan, G. J., Stafford, F. P., 1980. Do Union Members Receive Compensating
Differentials? American Economic Review 70, 355–81.
Ellingson, J.E., Gruys, M.L. and Sackett, P.R., 1998. Factors Related to Satisfaction
and Performance of Temporary Employees. Journal of Applied Psychology 83,
913 – 21.
Farber, H.S., 1999. Alternative and Part-Time Employment Arrangements as a
Response to Job Loss. Journal of Labor Economics 17, 142-69.
Forde, C., Slater, C., 2005. Agency Working in Britain: Character, Consequences and
Regulation. British Journal of Industrial Relations 43, 249 – 71.
Gramm, C. and Schnell, J., 2001. The Use of Flexible Staffing Arrangements in Core
Production Jobs. Industrial and Labor Relations Review, 55, 245-258.
Green, C. and Heywood, J.S., 2008. Does Performance Pay Increase Job Satisfaction?
Economica, 75, 710-728.
24
Green, C. and Leeves, G., 2004 Casual Employment and Internal Labour Markets
Manchester School 72, 658-676.
Güell, M. and Petrongolo, B., 2001. Worker Transition From Temporary to
Permanent Employment: The Spanish Case. CEP Discussion Paper 438, Centre
forEconomic Performance, London School of Economics.
Guest, D., MacKenzie Davey, K., and Smewing, C., 2007. Innovative Employment
Contracts: A Flexible Friend?, Working Paper, Department of Organizational
Psychology, Birkbeck College, University of London.
Hamermesh, D.S., 2004. Subjective Outcomes in Economics. Southern Economic
Journal 71, 2-11.
Hamermesh, D.S., 2001. The Changing Distribution of Job Satisfaction. Journal of
Human Resources 36, 1–30.
Heywood, J.S., 1989. Do Union Members Receive Compensating Differentials? The
Case of Employment Security. Journal of Labor Research 10, 271–85.
Heywood, J.S., Siebert, W.S. and Wei, X., 2007. The Implicit Wage Cost of Family
Friendly Work Practices. Oxford Economic Papers 59, 275–300.
Houseman, S. N., 2001. Why Employers use Flexible Staffing Arrangements.
Industrial and Labor Relations Review 55, 149–70.
Krausz, M., Brandwein,T. and Fox, S., 1995. Work Attitudes and Emotional
Responses to Permanent, Voluntary and Involuntary Temporary-Help Employees:
An Exploratory Study, Applied Psychology 44, 217–32.
Lazear, E. P., 1990., Job security provisions and employment. Quarterly Journal of
Economics 105, 699– 726.
25
Li, E. H., 1986. Compensating Differentials for Cyclical and Non-cyclical
Unemployment: The Interaction between Investors' and Employees' Risk
Aversion. Journal of Labor Economics 4, 277–300.
Mangan, J., (2000). Workers Without Traditional Employment: An International
Study of Non-Standard Work. Cheltenham: Edward Elgar.
McKelvey, R. and Zavoina, W. 1975. A Statistical Model for the Analysis of Ordinal
Level Dependent Variables, Journal of Mathematical Sociology, 4, 103-120.
Moretti, E., 2000. Do Wages Compensate for Risk of Unemployment? Parametric and
Semiparametric Evidence from Seasonal Jobs. Journal of Risk and Uncertainty 20,
45–66.
Morris, M.D. and Vekker, A., 2001. An Alternative Look at Temporary Workers,
Their Choices and the Growth of Temporary Employment. Journal of Labor
Research 22, 373–90.
Oswald, A. and Powdthavee, N. 2008. Does Happiness Adapt? A Longitudinal Study
of Disability with Implications for Economists and Judges, Journal of Public
Economics, 92, 1061-1077
Polivka, A.E., 1996. Into Contingent and Alternative Employment: By Choice?
Monthly Labor Review 119, 55 – 74.
Rosen, S. 1974. Hedonic Prices and Implicit Markets: Product Differentiation in Pure
Competition. Journal of Political Economy 82, 34 – 55.
Theosdossiou, I. and Vasileio. E., 2007. Making the Risk of Job Loss a Way of Life:
Does it Affect Job Satisfaction? Research in Economics 61, 71–83.
Van Ommeren, J. and Hazans. M., 2007. Workers' Valuation of the Remaining
Employment Contract Duration. Economica 74, 1–24.
26
Van Praag, B.M.S., Frijters, P. and Ferrer-i-Carbonell, A., 2003. The Anatomy of
Subjective Well-Being. Journal of Economic Behavior and Organization 51, 29 –
49.
27
TABLE 1– Job Satisfaction and Temporary Contracts Type, 1999 BHPS.
Overall Pay Hours Work Security
Permanent (41902) 5.364 4.935 5.207 5.436 5.555
Fixed Term Contract (1028) 5.293 5.064 5.367 5.514 3.656
Agency Temp (378) 4.682 4.513 5.126 4.845 3.207
Other Flexible Contract (603) 5.298 4.860 5.349 5.513 3.778
Observations 43911
28
TABLE 2 – Job Satisfaction and Flexible Working Arrangements, Ordered Probit
Estimates Males and Females, 1999-2004 BHPS.
Male Female
Model (I) Model (I)
Fixed Term Contract -0.068 [0.064] -0.062 [0.051]
Agency Work -0.581* [0.093] -0.397* [0.089]
Other Flexible Contract -0.189 **[0.095] -0.107 [0.071]
Age -0.034* [0.007] -0.026* [0.008]
Age2 0.0004* [0.0001] 0.0004* [0.0001]
Married 0.066* [0.027] 0.137* [0.024]
Dependent Child 0.139 [0.169] 0.138* [0.025]
A Level -0.093* [0.028] -0.086* [0.029]
Diploma -0.031 [0.039] -0.133* [0.042]
Degree -0.032 [0.034] -0.152* [0.032]
Higher Degree 0.026 [0.052] -0.271* [0.058]
Log Likelihood -28645.504 -27951.942
Observations 19048 19782
Controls included but not reported; region and year dummies. Robust standard errors
clustered at the individual level. *,**, *** indicate statistical significance at 1%, 5%
and 10% level, respectively.
29
TABLE 3 – Work Characteristics, Conditions, Flexible Work and Job Satisfaction:
Ordered Probit Estimates, 1999-2004, BHPS.
Male Female
Model 2 Model 3 Model 2 Model 3
Fixed Term Contract -0.053 [0.066] -0.031 [0.066] -0.152* [0.052] -0.146* [0.053]
Agency Work -0.418* [0.094] -0.375* [0.095] -0.346* [0.091] -0.319* [0.090]
Other Flexible Contract -0.163*** [0.096] -0.133 [0.097] -0.232* [0.072] -0.218* [0.073]
Tenure -0.001 [0.001] -0.004* [0.001] -0.002** [0.001] -0.004* [0.001]
Log Wage 0.065* [0.013] 0.055* [0.014] 0.019 [0.014] 0.049** [0.026]
Bonus/Profit Share 0.079* [0.022] 0.074* [0.023] 0.051** [0.026] -0.055 [0.037]
Performance Pay 0.075** [0.035] 0.057*** [0.035] -0.053 [0.037] -0.018 [0.040]
Bonus/Profit Share &
Performance Pay
0.047 [0.032] 0.031 [0.032] -0.012 [0.040] -0.010* [0.001]
Usual Hours -0.003*** [0.002] -0.003 [0.002] -0.010* [0.001] -0.003 [0.002]
Usual Overtime 0.004* [0.001] 0.004* [0.001] -0.003 [0.002] -0.156* [0.026]
Union Member -0.130* [0.028] -0.141* [0.028] -0.142* [0.025] 0.076** [0.032]
Public Sector 0.129* [0.036] 0.074** [0.036] 0.116* [0.030] 0.018 [0.024]
Manager/Supervisor 0.047** [0.024] 0.036 [0.024] 0.027 [0.023] -0.083* [0.025]
Firm Size
50-99 Employees -0.096* [0.028] -0.105* [0.028] -0.078* [0.025] -0.135* [0.029]
100-499 Employees -0.145* [0.029] -0.158* [0.029] -0.129* [0.028] -0.162* [0.030]
500+ Employees -0.122* [0.033] -0.142* [0.033] -0.150* [0.030] -0.158* [0.029]
Employer Pension 0.018 [0.027] -0.028 [0.026]
Employer Training 0.060* [0.022] 0.097* [0.021]
Annual Pay Increment 0.173* [0.021] 0.145* [0.020]
Night Shift -0.112 [0.080] -0.064 [0.078]
Shift Work -0.064**[0.034] 0.044 [0.039]
Other Non-Standard Hours -0.033 [0.035] -0.064** [0.034]
Flexitime 0.060* [0.029] 0.004 [0.026]
Annualised Hours -0.086* [0.048] -0.056 [0.052]
Term Time Work Only 0.135 [0.149] 0.079 [0.056]
Job Sharing -0.148 [0.267] -0.179**[0.093]
Log Likelihood -28425.514 -28348.044 -27739.950 -27678.834
Observations 19048 19782
Note: Controls included but not reported; industry, occupation, region and year
dummies, all other controls included in Table 2. Robust standard errors clustered at
the individual level in parentheses. *, **, *** indicate statistical significance at the
1%, 5% and 10% level, respectively.
30
Table 4 Fixed Effects Ordered Probit Estimates, Employees Aged 20-65
Males Females
Model 1 Model 2 Model 3 Model 1 Model 2 Model 3
Fixed Term Contract 0.061 [0.098] 0.051 [0.099] 0.071 [0.099] -0.080 [0.080] -0.102 [0.079] -0.089 [0.080]
Agency Work -0.334** [0.139] -0.191 [0.142] -0.154 [0.142] -0.184 [0.137] -0.084 [0.140] -0.048 [0.141]
Other Flexible Contract 0.112 [0.159] -0.099 [0.127] 0.128 [0.128] -0.221*** [0.133] -0.271** [0.133] -0.263** [0.133]
Log Likelihood -19914.340 -19817.220 -19784.210 -19159.360 -19068.22o -19037.900
Observations 19048 19782
Note: All controls as in table 3. Standard errors in parentheses. *,**, *** indicate statistical significance at 1%, 5% and 10% level, respectively.
31
TABLE 5 - Dimensions of Job Satisfaction, Fixed Effect Ordered Probit Estimates, Employees Aged 20-65.
Note: Controls included but not reported; industry, occupation, region and year dummies, all other controls included in Table 4,
model 3. Standard errors in parentheses. *, **, *** indicate statistical significance at the 1%, 5% and 10% level, respectively.
Pay Hours
Male Female Male Female
Fixed Term Contract 0.383* [0.095] 0.400* [0.095] 0.296* [0.096] 0.149** [0.077]
Agency Work 0.120 [0.139] 0.168 [0.140] 0.374** [0.143] 0.222 [0.140]
Other Flexible Contract 0.095 [0.123] 0.126 [0.152] 0.400** [0.156] -0.267** [0.127]
Log Likelihood -22033.990 -22039.93 -21813.670 -21313.730
Security Work Itself
Male Female Male Female
Fixed Term Contract -1.242* [0.095] -1.331* [0.076] 0.222* [0.098] 0.121 [0.081]
Agency Work -1.676* [0.142] -1.664* [0.144] 0.132 [0.143] -0.126 [0.144]
Other Flexible Contract -1.672* [0.159] -1.622* [0.125] 0.547* [0.158] 0.001 [0.132]
Log Likelihood -21024.91 -20096.40 -20044.230 -19837.530
Observations 19048 19782
32
TABLE 6 Job Satisfaction Controlling for Satisfaction with Job Security, Ordered Probit Estimates, 1999-2004, BHPS.
Pooled Fixed Effects
Male Female Male Female
Fixed Term Contracts 0.585* [0.066] 0.517* [0.057] 0.637* [0.103] 0.432* [0.084]
Agency Work 0.394* [0.109] 0.311* [0.093] 0.647* [0.145] 0.562* [0.147]
Other Flexible Contracts 0.509* [0.105] 0.324* [0.081] 0.989* [0.167] 0.435* [0.136]
Log Likelihood -26253.595 -26144.749 -18580.53 -18100.05
Observations 19007 19782
Note: All other controls as per table 4. Robust standard errors clustered at the individual in parentheses. *,**, *** indicate statistical significance
at 1%, 5% and 10% level, respectively.
33
TABLE 7 Life Satisfaction and Flexible Employment Contracts, Ordered Probit Estimates, Respondents Aged 20-65
Employees Only
Male Female
Pooled Fixed Effects Pooled Fixed Effects
Fixed Term Contract -0.095 [0.067] 0.149 [0.122] -0.157** [0.063] 0.108 [0.116]
Agency Work -0.229** [0.104] 0.218 [0.188] -0.369* [0.099] 0.002 [0.177]
Other Flexible Contract -0.227** [0.097] -0.130 [0.151] -0.220* [0.080] -0.208 [0.131]
Log Likelihood -21637.13 -12536.75 -22168.83 -12885.11
Observations 15200 15506
All Respondents
Male Female
Pooled Fixed Effects Pooled Fixed Effects
Fixed Term Contract -0.004 [0.045] -0.023 [0.072] -0.049 [0.042] 0.090 [0.063]
Agency Work -0.264* [0.086] -0.130 [0.127] -0.132*** [0.073] 0.220** [0.112]
Other Flexible Contract -0.070 [0.052] -0.225** [0.075] 0.041[0.043] 0.166* [0.063]
Employed 0.216* [0.015] 0.288* [0.036] 0.186* [0.013] 0.071** [0.028]
Log Likelihood -39487.07 -24159.40 -48284.30 -30561.32
Observations 25796 30400
Note: Standard errors in parentheses. *,**, *** indicate statistical significance at 1%, 5% and 10% level, respectively.
34
TABLE A1 – Summary Statistics, BHPS, 1999-2004.
Variable Male Female
Fixed Term Contract 0.019 0.027
Agency Work 0.007 0.009
Other Flexible Contract 0.010 0.018
Age (yrs) 39.196 39.443
Tenure (yrs) 10.200 10.382
Married 0.596 0.577
Dependant Child 0.005 0.317
A Level 0.220 0.191
Diploma 0.092 0.076
Degree 0.134 0.142
Higher Degree 0.042 0.032
Log Pay 6.667 6.371
Normal Hours 39.646 30.173
Overtime Hours 3.990 2.164
Union Member 0.285 0.321
Public Sector 0.232 0.450
Manager/Supervisor 0.415 0.320
Firm Size
50-99 0.256 0.269
100-499 0.249 0.186
500+ 0.175 0.158
Observations 19007 19782
35
TABLE A2 – Selected Summary Statistics by Contract Type, BHPS, 1999-2004.
Permanent Fixed Term Contract Agency Work Other Flexible
Male Female Male Female Male Female Male Female
Age 38.845 39.189 36.976 37.069 35.431 35.783 39.187 38.537
Tenure 10.743 10.722 7.346 8.211 5.723 6.075 9.293 8.627
Log Weekly Pay 6.674 6.385 6.484 6.295 5.473 5.434 5.662 5.463
Usual Hours 39.542 31.182 36.788 27.917 38.810 32.491 34.120 22.753
Union Member 0.317 0.363 0.185 0.299 0.080 0.161 0.160 0.205
Public Sector 0.232 0.452 0.424 0.764 0.153 0.255 0.267 0.541
Manager/Supervisor 0.426 0.349 0.224 0.175 0.058 0.075 0.240 0.131
Firm Size
0-49 Workers 0.295 0.355 0.296 0.330 0.190 0.248 0.527 0.474
50 to 100 Workers 0.260 0.276 0.203 0.294 0.372 0.286 0.173 0.328
100 to 499 Workers 0.261 0.200 0.257 0.146 0.277 0.211 0.207 0.138
500+ Workers 0.184 0.170 0.245 0.230 0.161 0.255 0.093 0.060
Observations 18568 18998 335 521 137 161 150 268
36
Endnotes
1 Satisfaction with the job has also been seen as one of the major "domains" that
together with satisfaction with leisure, health and others aggregate to an overall
measure of subjective well-being (van Praag et al. 2003).
2 In the UK, the growth in flexible contracts has been concentrated in the rise of
agency work (Forde and Slater 2005).
3 This results from the differences in the distribution of desires across firms and
workers and by search costs that make sorting imperfect. See Duncan and Stafford
(1980) for more detail.
4 Prior to 1999, flexible contracts were overly aggregated not allowing variations
within sub-groups of flexible contracts. For instance fixed term contracts and agency
workers were grouped together.
5 Existing US evidence suggests that unmeasured worker heterogeneity largely
accounts for the apparent lower earnings of flexible contract workers (Addison and
Surfield 2005).
6 These three additional estimates are available from the authors upon request.