Starting Wages Respond to Employer's Risk
| Date | 01 July 2014 |
| Published date | 01 July 2014 |
| DOI | http://doi.org/10.1111/sjpe.12043 |
STARTING WAGES RESPOND
TO EMPLOYER’S RISK
Peter Berkhout*, Joop Hartog** and Hans van Ophem**
ABSTRACT
Firms hiring new graduates face uncertainty on the future productivity of work-
ers. Theory suggests that starting wages reflect this, with lower pay for greater
uncertainty. We use the dispersion of exam grades within a field of education as
an indicator of the unobserved heterogeneity that employers face. We find solid
evidence that starting wages are lower if the variance of exam grades is higher
and higher if the skew is higher: employers shift the cost of productivity risk to
new hires, but pay for the opportunity to catch a really good worker. Estimating
the extent of risk cost sharing between firm and worker shows that shifting to
workers is larger in the market sector than in the public sector and diminishes
with experience.
1I
NTRODUCTION
In labour economics, as in other fields of economics, the role of uncertainty
has increasingly been acknowledged. Labour contracts are concluded under
conditions of imperfect information. Future market conditions are obviously
unknown, and this raises the question who will bear the risk of market shocks
in a relationship that is commonly intended to last for more than just one per-
iod and in which parties must make specific investments. Once the worker has
been engaged, an employer would like to reward her for her output, but as
output will depend both on effort and on external shocks beyond the control
of the worker, the question arises how a contract can stimulate effort and at
the same time efficiently allocate risk between employer and employee. Mal-
comson (1999) has surveyed this literature. The present paper is focused on a
particular aspect in this general setting of uncertainty: unknown productive
qualities of a worker who enters the labour market after finishing an educa-
tion. While we provide an analytical model for this problem, we consider the
empirical results our main contribution. We show how starting wages are
affected by the allocation of the risk associated with hiring a worker of imper-
fectly known quality. Our emphasis on starting wages is particularly relevant,
as labour market entrants cannot provide evidence regarding their workplace
performance.
*RIGO Amsterdam
**University of Amsterdam
Scottish Journal of Political Economy, DOI: 10.1111/sjpe.12043, Vol. 61, No. 3, July 2014
©2014 Scottish Economic Society.
229
An employer hiring a new employee fresh from school has no more than
imperfect information on the worker’s qualities. The diploma itself, some
information on school grades, extracurricular activities, a job interview and
perhaps a psychological test cannot fully resolve the uncertainty about future
productivity. Firms may be expected to bill the workers for the cost of dealing
with this uncertainty. Workers fresh from school have no successes yet to sup-
port a strong bargaining position and will have to accept that employers put
a discount on starting wages in accordance with the risk they face. Thus, we
predict that starting wages will be lower in fields where employers face more
uncertainty on any individual’s productivity. However, we also predict that
starting wages will be higher if employers perceive more skew in the produc-
tivity distribution: they appreciate the chance to catch an individual with very
high productivity. We find clear support for these predictions.
We use the distribution of exam grades within a field
1
to measure uncer-
tainty. Exam grades differ among individuals in a given field because of differ-
ences in abilities, efforts and no doubt other factors. We assume that the
heterogeneity that is reflected in the variance of exam grades is correlated with
the heterogeneity in productivity that is relevant for employers. If the variance
of exam grades across graduates in economics is larger than across graduates
in physics, we assume that employers can make less accurate predictions on
the productivity of an individual economist than on the productivity of an
individual physicist. We do not require that employers consult grade variance
data. Employers have their own ways of learning about the productivity risks
among graduates in different fields. We only assume that these risks are ade-
quately proxied by school grade variance (and skew). Our core hypothesis is
that wages will respond negatively to the variance of exam grades in a field
(workers pay a risk premium) and positively to the skew of the exam grades
in a field (firms appreciate the upside risk of hitting upon a very good
worker). This hypothesis is supported by a large sample of starting salaries
for graduates from tertiary education in the Netherlands. We scrutinize our
interpretation by adding robustness checks and by considering alternative
explanations. Our stance is that we do not see a consistent alternative expla-
nation for our findings.
We present formal modelling to derive our core prediction that wages of
starting graduates will reflect the productivity risk that employers face.
2
We
will also specify conditions for productivity risk to be indicated by the distri-
bution of exam grades. Worker payment for productivity risk has been high-
lighted in some earlier papers. Freeman (1977) introduced the idea that risk
neutral firms are willing to insure starting workers against wage drops as
information develops on their productivity. Harris and Holmstrom (1982)
1
In the Netherlands, studies are not organized by major: first-year students right away
specialize in an academic discipline (economics, physics, etc.).
2
However, we have not developed a full model of the labour market for new graduates.
Harris and Holmstrom (1982) provide such a model and come up with essentially the same
prediction as we do; they only consider the effect of the variance of unobserved productivity,
not of skew.
230 P. BERKHOUT, J. HARTOG AND H. VAN OPHEM
Scottish Journal of Political Economy
©2014 Scottish Economic Society
further developed this model to a market with risk neutral firms, risk averse
workers and symmetric incomplete information on individual worker produc-
tivity. They show that wages are reduced by an insurance premium that
diminishes with work experience (as information accumulates) and that is
increasing in the (perceived) variance of productivity. Also, the variance of
the wage increases with experience as wages come to reflect individual produc-
tivity. Harris and Holmstrom demonstrate that their model is in line with sev-
eral stylized facts on wages, but offer no new direct testing.
3
Rothschild and Stiglitz (1982) and Aigner and Cain (1977) also predict that
wages will respond negatively to higher perceived variance in unobserved ability.
The former argue on the basis of mismatch between imperfectly observable indi-
vidual ability and the optimal ability level for a given job, the latter on the basis
of noise in an imperfect indicator of productivity. Neither of these papers offers
any empirical evidence. To the best of our knowledge, there is no empirical work
comparable to ours (cf. Waldman, 2007). Our paper is the first that directly tests
the prediction that wages are lower if the variance in unobserved productivity is
higher with an indicator of productivity variance.
Allowing for skew is not routinely considered in labour market applications,
but is well known in the lifecycle consumption-savings literature (as ‘prudence’).
We include skew because we want to mirror our treatment of worker compensa-
tion for risk in educational choice (Hartog, 2011). An education gives access to
a distribution of wages, not to a single wage rate, and there is substantial evi-
dence that workers are compensated for a high variance of wages and accept a
wage cut for a high skew. These results provide a good backdrop for the present
analysis: in a sense it is the mirror image of the case we study here. Complemen-
tarity of these two cases means that the results of the tests reinforce each other.
We consider the empirical results as the prime contribution of this paper:
wages respond to the distribution of exam grades in a field of education. We
show that this result is quite robust. We offer an analytical interpretation that
requires two steps. The first step is to show that firms want compensation for
the risk they face when hiring a new graduate with imperfectly known produc-
tivity. This, essentially, requires the assumption that firms are risk averse. It is
routinely assumed that workers are risk averse and firms are risk neutral. This
is probably pushing the case too far. It is quite likely that on average workers
are more risk averse than firms, but no doubt firms are also risk averse. Small
firms may have every reason to behave as risk averters, as they often lack the
resources to survive bad draws. But large firms are also observed to engage in
buying all kinds of insurances, for failing debtors, worker safety hazards, cur-
rency fluctuations, etc.
4
We know of no direct evidence that firms are risk
neutral; direct measurement of entrepreneurs’ risk attitudes points to lower
risk aversion than among employees but not to risk neutrality as the domi-
nant outcome (Cramer et al., 2002). We think there is sufficient evidence to
3
Waldman (2007) gives a good survey of the literature on wages under imperfect informa-
tion.
4
To witness: Dutch electronics multinational Philips sells its chips division because sales
and profits vary too much over the business cycle (NRC, 4 August 2006).
STARTING WAGES AND RISK 231
Scottish Journal of Political Economy
©2014 Scottish Economic Society
To continue reading
Request your trial