University of Groningen
Risk aversion and job mobility
van Huizen, Thomas; Alessie, Rob
Published in:
Journal of Economic Behavior & Organization
DOI:
10.1016/j.jebo.2019.01.021
IMPORTANT NOTE: You are advised to consult the publisher's version (publisher's PDF) if you wish to cite from
it. Please check the document version below.
Document Version
Publisher's PDF, also known as Version of record
Publication date:
2019
Link to publication in University of Groningen/UMCG research database
Citation for published version (APA):
van Huizen, T., & Alessie, R. (2019). Risk aversion and job mobility.
Journal of Economic Behavior &
Organization
,
164
, 91-106. https://doi.org/10.1016/j.jebo.2019.01.021
Copyright
Other than for strictly personal use, it is not permitted to download or to forward/distribute the text or part of it without the consent of the
author(s) and/or copyright holder(s), unless the work is under an open content license (like Creative Commons).
The publication may also be distributed here under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license.
More information can be found on the University of Groningen website: https://www.rug.nl/library/open-access/self-archiving-pure/taverne-
amendment.
Take-down policy
If you believe that this document breaches copyright please contact us providing details, and we will remove access to the work immediately
and investigate your claim.
Downloaded from the University of Groningen/UMCG research database (Pure): http://www.rug.nl/research/portal. For technical reasons the
number of authors shown on this cover page is limited to 10 maximum.
Download date: 09-08-2022
Journal of Economic Behavior and Organization 164 (2019) 91–106
Contents lists available at ScienceDirect
Journal of Economic Behavior and Organization
journal homepage: www.elsevier.com/locate/jebo
Risk aversion and job mobility
Thomas van Huizen
a , ∗
, Rob Alessie
b , c
a
Utrecht School of Economics, Utrecht University, Kriekenpitplein 21-22, 3584 TC, Utrecht, The Netherlands
b
Faculty of Economics and Business, University of Groningen, Nettelbosje 2, 9747 AE, Groningen, The Netherlands
c
Netspar, Warandelaan 2, 5037 AB, Tilburg, The Netherlands
a r t i c l e i n f o
Article history:
Received 19 August 2017
Revised 18 January 2019
Accepted 21 January 2019
Available online 4 June 2019
JEL classification:
C90
D03
D81
J62
Keywords:
Job mobility
Risk aversion
Job search
Risk preferences
a b s t r a c t
Job mobility is inherently risky as workers have limited ex ante information about the
quality of outside jobs. Using a large longitudinal Dutch dataset, which includes data on
risk preferences elicited through an (incentivized) lottery-choice experiment, we examine
the relation between risk aversion and job mobility. The evidence shows that risk averse
workers are less likely to move to other jobs. The results are stronger for male workers
and for workers who hold a permanent contract. Our empirical findings indicate that the
negative relation between risk aversion and job mobility is driven by the job acceptance
rather than the search effort decision.
©2019 Elsevier B.V. All rights reserved.
1.
Introduction
Although most decisions involve risk, this is particularly true in the labor market. This study focuses on a risky decision
that is relevant to almost all workers and that may have major consequences for the individual’s career path: the decision
to quit and move to another job. According to canonical models on job mobility, uncertainty plays a crucial role in ex-
plaining mobility processes: there is uncertainty about whether the worker will be able to find a better job ( Burdett, 1978;
Mortensen, 1986 ) and about the quality of outside jobs ( Jovanovic, 1979 ). Because workers have limited information about
many aspects of the new job, they may realize that they ended up in a poor match after accepting an outside job offer.
The premise that job mobility is risky is also consistent with the empirical literature, showing that job mobility can be an
important source of wage growth ( le Grand and Tåhlin, 2002; Schmelzer, 2012; Topel and Ward, 1992 ), but may also lead
to wage losses or lower wage growth ( Borjas, 1981; Light and McGarry, 1998; Tjaden and Wellschmied, 2014 ). Because job
mobility is inherently risky, it can be expected that risk preferences affect the decision to move to another job.
This paper examines the relation between risk aversion and job mobility. Given that uncertainty plays a central role
in canonical models for the analysis of turnover, it is surprising that the role of risk aversion in turnover decisions has
been largely ignored in the labor economics literature. Existing theoretical models generally assume risk neutral individuals
or homogeneous risk preferences. Allowing for heterogeneity in risk preferences, we demonstrate theoretically that risk
∗
Corresponding author.
E-mail address: t.m.vanhuizen@uu.nl (T. van Huizen).
https://doi.org/10.1016/j.jebo.2019.01.021
0167-2681/© 2019 Elsevier B.V. All rights reserved.
92 T. van Huizen and R. Alessie / Journal of Economic Behavior and Organization 164 (2019) 91–106
Table 1
Choices in the lottery task.
All Incentivized Hypothetical
Men Women Men Women Men Women
Panel A: Fraction choosing the certain payoff
Game # ( € certain payoff)
Game 1 (20) 32.80 46.92 29.30 43.09 37.58 51.63
Game 2 (25) 41.67 60.70 37.21 61.70 47.77 59.48
Game 3 (30) 58.33 69.50 57.21 70.74 59.87 67.97
Game 4 (35) 73.92 80.94 73.49 81.38 74.52 80.39
Game 5 (40)
81.18 87.68 79.07 87.77 84.08 87.58
Panel B: Average number of safe choices
Nr of safe choices 2.88 3.46 2.76 3.45 3.04 3.47
Notes: In all five games, when the risky option is chosen the subject receives either
€ 5or€ 65 (both outcomes occur with a 50% chance).
aversion is negatively related to job mobility. There are two potential channels through which job mobility is affected by risk
aversion: the job acceptance channel and the job search channel. First, given the uncertainty associated with outside jobs,
risk averse workers are more critical about outside job offers. Hence, conditional on receiving an offer risk averse workers
are more likely to decline an offer. Second, due to the higher probability of rejecting outside offers, the marginal gains from
search are lower for risk averse workers and they will therefore search less intensively. Moreover, risk averse workers will
invest less in search because this is a risky investment activity involving short-run costs and uncertain benefits.
We examine the relation between risk aversion and job mobility using the LISS, a longitudinal panel from the Nether-
lands. Given the longitudinal nature of the data we are able to follow individual labor market trajectories over time, which
allows us to analyze job mobility behavior. Moreover, as the data contains information about (on-the-)job search behavior,
we are able to test one of the channels through which risk aversion is related job mobility. A unique feature of the data is
that in addition to information on labor market behavior and a wide range of background characteristics, the LISS panel in-
cludes various measures of risk preferences. We exploit a measure of risk aversion that is elicited through an (incentivized)
lottery-choice experiment as well as several survey questions on the respondent’s willingness to take risks.
The evidence indicates that workers who are more risk averse are less likely to be mobile on the labor market. This
finding appears not to be crucially dependent on the risk aversion measure: both the results based on the lottery-choice
experiment data and the results based on the survey questions on risk aversion point out a negative relation between risk
aversion and job mobility. Overall, this relation appears to be stronger for men and when job mobility involves more uncer-
tainties. The evidence from the lottery-choice experiment does not indicate that risk aversion decreases on-the-job search
effort, which suggests that the negative relation between risk aversion and job mobility is driven by the job acceptance deci-
sion: risk averse workers are more likely to reject outside offers. Surprisingly, some of the estimation results show that more
risk averse workers search more intensively on-the-job. An explanation for this finding is that on-the-job search is not only
used to generate outside job offers (as assumed in on-the-job search models), but also to obtain more information about the
quality of potential job offers and the individual’s labor market position. Search may thereby decrease uncertainties involved
in job mobility.
Our study contributes to a growing literature testing the effects of risk aversion on labor market outcomes. Existing
studies have examined how risk aversion is related to educational and occupational choice ( Bonin et al., 2007; Falco, 2014;
Fouarge et al., 2014 ), migration decisions ( Bauernschuster et al., 2014; Dustmann et al., 2017; Goldbach and Schlüter, 2018;
Heitmueller, 2005; Jaeger et al., 2010 ), reservation wages of unemployed job seekers ( Feinberg, 1977; Pannenberg, 2010 ),
wage growth ( Budria et al., 2013; Shaw, 1996 ) and the decision to become (and remain) an entrepreneur ( Caliendo et al.,
2010; 2009; Koudstaal et al., 2015; Skriabikova et al., 2014 ).
1
Empirical evidence on the relation between risk aversion and
job mobility is virtually non-existent. To our knowledge, Maier et al. (2016) is the only other study that tests the relation
between risk attitudes and job mobility empirically.
2
Using the German Socio-Economic Panel Survey, Maier et al. (2016) find
that more risk-tolerant individuals move more often from one job to another. However, they rely on a survey question on
risk attitudes rather than experimental data to capture heterogeneity in risk aversion and do not test the relation between
risk aversion and on-the-job search effort. In general, most studies on the relation between risk aversion and labor market
behavior rely on survey-based questions on risk attitudes to capture variation in risk preferences ( Koudstaal et al. (2015)
and Goldbach and Schlüter (2018) are exceptions). In addition to examining the predictive value of risk aversion in a new
domain, one of the strengths of our approach is that we use an experimentally elicited measure of risk aversion for a
relatively large sample of field subjects.
1
A related strand of literature examines how risk aversion affects technology adoption decisions (e.g. Liu, 2013 ).
2
The idea that job mobility is a risky decision is mentioned in several economic studies but never examined explicitly. For instance,
Tom et al. (2007) state: “Many decisions, such as ... to accept a new job, involve the possibility of gaining or losing relative to the status quo. When
faced
with such decisions, most people are markedly risk averse.” Outside the field of economics, the study of Allen et al. (2007) discusses the role of risk
attitudes and derives several propositions drawing (mainly) on the psychological literature. However, the study does not test these propositions empirically.
T. van Huizen and R. Alessie / Journal of Economic Behavior and Organization 164 (2019) 91–106 93
Furthermore, this study contributes to the literature on turnover and labor market dynamics by providing new insights
in the determinants of turnover. Turnover is a relevant economic variable, as it affects wages and careers ( Blau and DeVaro,
2007; Dustmann and Pereira, 2008; Topel and Ward, 1992 ): the results may therefore provide a new explanation for wage
inequality. Given the evidence indicating a significant intergenerational correlation of risk attitudes ( Dohmen et al., 2012 ),
the study sheds light on a new mechanism explaining (low levels of) intergenerational income mobility. In addition, turnover
has an impact on firm productivity ( Ilmakunnas et al., 2005; Jackson, 2013; Siebert and Zubanov, 2009 ) and is relevant for
the (allocative) efficiency of the labor market ( Mortensen, 2011 ). The paper demonstrates that heterogeneity in preferences
are important, and that (policy) evaluations assuming risk neutrality or a single risk aversion parameter (representative
agent models) produce incomplete results.
The paper is structured as follows. The next section discusses the theoretical mechanisms through which risk aversion
affects job mobility. Next, the data is discussed and the empirical results are presented. The final section concludes.
2. Theory
2.1. Theoretical models on job mobility
The benchmark theoretical models of turnover in economics are based on imperfect information. Borjas and Gold-
berg (1978) already pointed out the relevance of uncertainty in the job mobility process: “it is likely that uncertainty both
before and after search about firms and workers and the on-the-job learning process which reduces this uncertainty is an
important characteristic of the labor market” ( Borjas and Goldberg, 1978 , p. 124). In the current theoretical literature, we
can distinguish between two general models: on-the-job search models, where jobs are search goods (e.g. Burdett, 1978;
Mortensen, 1986 ), and learning models, where jobs are experience goods ( Johnson, 1978; Jovanovic, 1979; 1984 ). According
to the first type of models, workers search for other jobs and when an offer is located they accept it if the value (wage)
of the alternative job is higher than the value of the current job. On-the-job search models assume that workers have im-
perfect information before search takes place but perfect information about the job once it has been located. Hence, there
is no ex ante uncertainty about the value of an offered job. In contrast, learning models are based on the assumption that
workers have no or limited ex ante information about the value of the job: the worker learns about the value of the new job
(quality of the job match) while on the job. As pointed out by Jovanovic (1979 , p.973), the fundamental difference between
these models is that in models where jobs are pure search goods job mobility is due to the arrival of new information about
alternative job opportunities, whereas in models where jobs are pure experience goods turnover is the result of obtaining
new (negative) information about the current job.
In existing theoretical models on job mobility heterogeneity in risk preferences plays no role since risk neutrality or a
single risk aversion parameter (representative agent models) is assumed. However, given the uncertainty involved in job
mobility it is likely that heterogeneity in risk preferences explains turnover. Based on the two main theoretical models, we
can distinguish between two channels through which risk aversion affects job mobility. First, in on-the-job search models
search can be considered as an investment involving costs (mainly in terms of time and effort) and uncertain rewards (in
terms of locating a good job offer). More risk averse workers are less willing to engage in investment activities such as on-
the-job search and are therefore less likely to receive outside offers. Second, in learning models where jobs are considered
as experience goods, the job acceptance rather than the job search decision is affected by the individual’s degree of risk
aversion. Because it is impossible to completely evaluate the value of the job before accepting the offer, the worker faces
the risk of accepting a poor job match. The more certain option would be to stay and reject the uncertain job offer. Hence,
risk averse workers are less likely to quit and move to another job as they invest less in job search activities and are more
critical about alternative job offers. However, it is plausible that search costs are quantitatively small and that the potential
losses associated with unsuccessful search are limited. In contrast, the potential losses of accepting a ‘lemon’ are substantial.
Hence, we argue that job mobility is risky mainly because it is impossible to completely evaluate the value of the job before
accepting an offer and that the role of risk aversion is more relevant in models where jobs are considered as experience
goods. In the spirit of these models, we present below a model in a rather stylized form that shows how risk aversion
affects job mobility. In Section 2.3 we discuss the relation between risk aversion and on-the-job search more extensively.
2.2. Risk av ersion and the job acceptance decision
The model follows the central premise of the model of Jovanovic (1979) : the individual has more information about the
current job than about outside job opportunities and new information arrives while on the job. To capture the idea of ex
ante uncertainty about match quality, we assume that each period the worker receives an alternative offer y , which indicates
the match-specific value of the job drawn from the cumulative distribution function F ( y ). Since many uncertainties about a
job involve non-pecuniary characteristics, the job has a particular value in terms of utility and is not simplified to the wage
level (which is generally observed before accepting the offer). The value of the match contains all aspects of the job that
generate (dis)utility for holding the job, such as income, working hours, work atmosphere and commuting time. The true
value of the job offer is not observed when the worker receives an offer. Instead, workers receive a noisy signal
ˆ
y = y + ε.
94 T. van Huizen and R. Alessie / Journal of Economic Behavior and Organization 164 (2019) 91–106
The noise term ε is stochastically independent of y and has mean zero and variance σ
2
ε
.
3
For simplicity, we assume a zero
discount rate. Furthermore, we assume that the worker has perfect information about the current job match and that the
value of the job is immediately revealed when the job is accepted.
Conditional on receiving an outside offer, a worker accepts the job only if the observed signal of the job
ˆ
y is higher than
the reservation match quality
ˆ
y
∗
. The worker is indifferent between rejecting and accepting the offer if:
V (y
0
) = E
V (y ) |
ˆ
y =
ˆ
y
∗
(1)
where V ( y
0
) is the utility value of the current job match y
0
and E[ V (y ) |
ˆ
y =
ˆ
y
∗
] represents the expected utility value of the
reservation match quality
ˆ
y
∗
. We assume that workers are risk averse, i.e. V
(.) > 0, V
(.) < 0. Risk averse workers have a
positive risk premium which is implicitly defined in the following equation:
E
V (y ) |
ˆ
y =
ˆ
y
∗
= V
E
y |
ˆ
y =
ˆ
y
∗
−
(2)
Eqs. (1) and (2) imply:
E
y |
ˆ
y =
ˆ
y
∗
= y
0
+ (3)
Note that E(y |
ˆ
y =
ˆ
y
∗
) is a (possibly non-linear) function of
ˆ
y
∗
: E(y |
ˆ
y =
ˆ
y
∗
) = g(
ˆ
y
∗
) . Under the assumption that g (.) is a
monotonically increasing function,
4
we can write down the following expression for the reservation match quality
ˆ
y
∗
:
ˆ
y
∗
= h
(
y
0
+
)
(4)
where h (. ) = g
−1
(. ) and h
(.) > 0. If we assume that y and ε are normally distributed, we can rewrite Eq. (4) as Eq. (A.4) (see
Appendix A ). Eq. (4) implies that the reservation match quality increases with the individual’s risk premium. Obviously,
more risk averse workers ask a higher premium to accept the new job than a less risk averse worker.
5
Consequently,
more risk averse workers are more selective about job offers and are less mobile between jobs. Moreover, it can be shown
that (for small risks) the relevance of risk aversion in the job acceptance decision positively depends on the noise of the
signal σ
2
ε
(see also Eq. (A.4) ).
6
The interaction between risk aversion and noise is intuitive: if the quality of the match can
be perfectly observed (i.e. when jobs are pure search goods, σ
2
ε
= 0 ), then job mobility involves no uncertainty (
ˆ
y = y ) and,
according to Eq. (1) ,
ˆ
y
∗
= y
0
and = 0 . In that case, risk aversion does not affect the job acceptance decision.
2.3. Risk av ersion and the job search decision
The model so far describes the worker’s choice conditional on receiving an alternative offer. In the model the effect of
risk aversion on job mobility operates through the job acceptance decision. The analysis implicitly assumes that the job
offer arrival rate is exogenous. However, in on-the-job models the intensity of search effort plays an important role as it
determines the probability of receiving an offer. On-the-job search includes all kinds of activities that increase the chances
of locating and receiving an offer: the worker has to search for available vacancies and prepare for and actively attend job
interviews. These activities involve time and effort and are often experienced as stressful.
Theoretically, on-the-job search involves costs and uncertain future rewards: search is unsuccessful when no job offer
is received or when an offer is below the worker’s reservation match quality. Given that search is an investment activity
with uncertain rewards, risk averse workers are more reluctant to make such investments. So even if jobs are pure search
goods, one may expect that risk averse workers invest less in on-the-job search and for that reason are less mobile between
jobs. When risk averse workers set a higher reservation match quality (as discussed above), they are more likely to reject
a job offer conditional on receiving one. This implies that the marginal gains from search decline with the worker’s degree
of risk aversion. Hence, risk aversion may affect on-the-job search through two channels: risk averse workers are less likely
to invest in activities with uncertain rewards and have lower expected gains from search as they are more likely to reject
potential offers (see Appendix A for a more formal discussion).
2.4. Discussion
Several factors are not explicit in the basic theoretical model but may be relevant in actual job mobility processes. First,
we have assumed that staying is less risky than moving. Although it is likely that the worker knows more about his current
job than about others jobs, this assumption may not always hold in reality. In general, we may expect that the current
match offers more employment protection than the alternative match as firing costs increase with tenure and workers
may have obtained a permanent contract. This is especially relevant in the European context, where often large differences
in employment protection exist between workers on permanent contracts and their temporary counterparts. Permanent
workers may have to sacrifice their employment protection if they move to another job, so for these workers quitting is
3
See also Dustmann et al. (2016) for an interesting application of uncertainty in the job mobility process: their model allows productivity to be match
specific and as in our model this is observed with a noisy signal before the worker accepts the job. However, the model assumes risk neutrality.
4
This seems to us a plausible assumption because y and
ˆ
y are positively correlated ( co
v (y,
ˆ
y ) = cov (y, y + ε) = σ
2
y
> 0 ).
5
See e.g. proposition 1.5 of Eeckhoudt, Gollier and Schlesinger (2005).
6
See e.g. proposition 1.3 of Eeckhoudt, Gollier and Schlesinger (2005).
