D I S C U S S I O N P A P E R S E R I E S
zur Zukunft der Arbeit
Institute for the Study
Monopsony Power, Pay Structure and Training
IZA DP No. 5587
Stefan C. Wolter
Pay Structure and Training
University of Bern
King’s College Cambridge
Stefan C. Wolter
University of Bern,
CESifo and IZA
Discussion Paper No. 5587
P.O. Box 7240
Any opinions expressed here are those of the author(s) and not those of IZA. Research published in
this series may include views on policy, but the institute itself takes no institutional policy positions.
The Institute for the Study of Labor (IZA) in Bonn is a local and virtual international research center
and a place of communication between science, politics and business. IZA is an independent nonprofit
organization supported by Deutsche Post Foundation. The center is associated with the University of
Bonn and offers a stimulating research environment through its international network, workshops and
conferences, data service, project support, research visits and doctoral program. IZA engages in (i)
original and internationally competitive research in all fields of labor economics, (ii) development of
policy concepts, and (iii) dissemination of research results and concepts to the interested public.
IZA Discussion Papers often represent preliminary work and are circulated to encourage discussion.
Citation of such a paper should account for its provisional character. A revised version may be
available directly from the author.
IZA Discussion Paper No. 5587
Monopsony Power, Pay Structure and Training*
Although interest in monopsonistic influences on labour market outcomes has revived in
recent years, only a few empirical studies provide direct evidence on it. This paper analyses
empirically the effect of monopsony power on pay structure, using a direct measure of labour
market ‘thinness’. We find that having fewer competitors for skilled labour is associated at the
level of the establishment with lower pay for both skilled labour and trainees, but not for
unskilled labour. These findings have potentially important implications for the economic
theory of training, as most recent models assume that skilled pay is set monopsonistically but
both unskilled and trainee pay are determined competitively. Our results support those
assumptions for skilled pay and unskilled pay, but not for trainee pay.
JEL Classification: J24, J31, J42
Keywords: monopsony, wage differentials, firm-sponsored training
Forschungsstelle für Bildungsökonomie
* This research has partly been funded by the OPET through its Leading House on Economics of
Education: ‘Firm Behaviour and Training Policies’. The authors would like to thank Uschi Backes-
Gellner, Giorgio Brunello and Ludger Wössmann for helpful comments and discussions.
The recent renewal of interest in monopsony power in the labour market is indicated by the
various articles introduced by Ashenfelter, Farber and Ransom (2010).
Monopsony power is present when the firm faces an upward-sloping labour supply curve,
which gives it the power to set the wage, in contrast to its impotence if faced by the perfectly
elastic supply curve of perfect competition. Monopsony was initially thought of in literal
terms, as a single buyer in an isolated labour market, and judged correspondingly rare in
practice. However, an individual firm can still face an upward-sloping supply curve even
when other firms in the labour market compete with it for labour, whether because of fewness
of competitors (oligopsony), a dispersion of both jobs and workers in terms of location and
non-wage attributes (classic differentiation), or the costs of either geographical mobility or
search-related information (Boal and Ransom 1997).
Indeed, while the classic source of monopsony power – few competing firms and/or collusion
by employers – might not at first sight seem especially relevant to modern labour markets,
recent theories of monopsony power essentially involve the same assumption. Even when
many employers are present in a labour market, the other factors may mean that individual
workers face „few employment opportunities‟, because of the costs of locating vacancies or
travelling to distant locations, which means that labour markets are effectively „thin‟
(Manning, 2003a: 106-108).
Much of the evidence of monopsony power has been indirect, involving aspects of wage
dispersion, the finance of work-based training, pay differences by race and sex, and the
effects of changes in statutory minimum wages (Bhaskar, Manning and To 2002). Recent
research has focused on more direct evidence, in terms of the elasticity of labour supply to
individual firms (e.g., Ransom and Sims 2010; Staiger, Spetz and Phibbs 2010).
A second, longer-standing potential source of direct evidence is the association between pay
and buyer concentration across local labour markets: does the presence of fewer employers
mean lower pay? Boal and Ransom (1997) noted the absence of empirical consensus on the
issue, even in markets for skilled labour that might be expected to be monopsonistic, notably
school-teachers and hospital nurses. Some older studies, including Bunting (1962) and Hirsch
and Schumacher (1995), found little or no evidence that wages are lower in markets with
fewer employers; others, including Link and Landon (1976), evidence that they are. Boal and
Ransom (1997) attributed the weak pattern of results to the availability of only variable and
partial controls for heterogeneity across localities in other potential determinants of pay,
notably in workers‟ living costs and in the attributes of employers. In particular, they pointed
to the tendency for pay to be inversely associated with fewness of employers in unskilled
occupations as well. As such markets might be expected to behave competitively, the same
association suggests that the apparent effect of fewness of employers is actually caused by the
difficulty of delineating local labour markets empirically and controlling statistically for their
attributes. However, as that inference is itself not well established empirically, the issue
should remain open.
This paper adopts the latter approach, analysing the effects of fewness of firms across local
labour markets. We analyse the association between pay and fewness of employers across
local labour markets in a wide range of occupations, unskilled and skilled. We first establish
evidence of a fewness effect on pay in skilled labour markets but not in unskilled ones. Then
we analyse the association between fewness and the pay differentials between skilled
employees, unskilled employees, and trainees within an establishment. This specification
potentially controls for heterogeneity of firms and local living costs. We find that skill
differentials are higher in local labour markets with fewer employers, and, given that
uncontrolled differences in living costs and employer attributes are unlikely to matter as
much for intra-firm pay differentials as for absolute pay levels, we interpret that association
as evidence of monopsony power. Furthermore, we find that monopsony power is associated
more with the number of local competitors in the same sector rather than the number in all
sectors, which suggests that our results are not due to the overall level of economic activity in
An analysis of monopsony in relation to skill differentials in pay is economically interesting
in its own right as well. Contemporary models of the finance of work-based training interpret
evidence that employers share the cost of general training, which they could not do under
perfect competition, in terms of an assumption that firms have monopsony power over skilled
labour, but not over unskilled labour. The result is „wage compression‟ between skilled and
unskilled labour, i.e., a smaller growth in pay than in marginal product as an employee‟s skill
is increased by training. Wage compression in turn provides the incentive to and opportunity
for firms to bear the costs of training (Stevens 1994; Acemoglu and Pischke 1999a).
This central assumption has lacked direct evidence, and relied instead on indirect evidence
such as the distribution of training costs, and the relationships between pay, post-training
separations and recruitment in markets for skilled labour (Acemoglu and Pischke 1998). The
validity of that assumption may however be limited, particularly when unskilled employees
and trainees are distinguished, and the possibility of monopsony power over trainees is
allowed (Wolter and Ryan 2010).
Our evidence concerns Switzerland, a country whose training system closely resembles the
German one, from which influential conclusions have been drawn about monopsony power,
pay structure, and training volume (Acemoglu and Pischke 1998; Dustmann and Schönberg
2009). We analyse a representative data set for 3,500 establishments in Switzerland in 2004,
into which have been merged the attributes of the local labour markets in which those
establishments are located. We consider pay within the largest category of skilled
employment, at detailed occupation level, at the establishment, relative both to trainees in the
same occupation and to unskilled employees. The results represent a wide range of
occupations, not just the particular ones (e.g., nurses) or the broad ones (e.g., white collar
workers) that dominate the literature. We find evidence of significant monopsonistic
influence on the pay of both skilled workers and trainees, but not on that of unskilled
The next section discusses analytically issues concerning monopsony power in labour
markets. Section 3 outlines the sources of our data. In section 4 we outline our estimation
strategy. Section 5 presents the results for the effects of monopsony on, first, the firm‟s pay
level and pay structure in general, and, second, on training-related pay structure in firms that
provide formal training. The conclusions come in section 6.
2. Economics of monopsony
What generates buyer power in the labour market, and what are its effects on pay,
employment and training?
2.1 Pay setting and employment
The idea of monopsony may best be illustrated by the question asked in Manning (2003b : 3):
“What happens if an employer cuts the wage it pays its workers by one cent?”. Assuming
perfect competition, all workers immediately leave the firm, as implied by the law of one
wage. However, if not all would leave, the firm can create a rent by lowering wages below
the competitive level. The literal case of monopsony is a single buyer in a particular labour
market (Robinson 1933). The stereotypical example is a small coal mining town with no
employers beside the mining firm, which gives the latter wage-setting power similar to the
price-setting power a monopolist enjoys in a product market.
However, instances of literal monopsony are rare or non-existent. More plausibly, only a few
employers may compete for a particular type of labour in a locality, in which case each
oligopsonist enjoys monopsony power without being a strict monopsonist. Although
outcomes under oligopsony are a priori indeterminate, in the sense of being sensitive to
assumptions about behaviour in the face of strategic interdependence, in some models,
notably Cournot competition, an equilibrium exists in which the presence of fewer employers
is associated with lower pay. Similarly, even if collusion is more important in practice than
economists have tended to believe, the scope for its success, i.e., attaining the low wage of
literal monopsony, increases as the number of oligopsonists declines, thereby reducing the
benefit to the firm of breaking rank and increasing the visibility to other firms of its having
done so (Boal and Ransom 1997; Manning 2010).
Even interpreted in terms of oligopsony, the importance of monopsony power might be
doubted given the large numbers of employers that function in many modern labour markets.
Recent work emphasises however the potential relevance of monopsony power even in such
contexts (Manning 2003b; Ashenfelter et al. 2010). Inelasticity in the supply of labour to the
firm may be expected to result from frictions in the labour market – notably from the costs to
workers of information about job vacancies and the costs of geographical mobility, and also
from dispersion in job attributes and worker preferences concerning travel costs and other
non-wage amenities (type of work, social ties, etc.; Boal and Ransom 1997). For example,
workers who face substantial travel-to-work costs refuse a more distant job when the wage
gain from changing jobs is less than increased cost of getting to and from work (Bhaskar and
To 1999; Manning 2003a). Alternatively, if job search is taken to be costly, not least because
vacancies are scarce, workers may not be aware of alternative job opportunities in the first
place. Under such conditions, the labour market remains „thin‟ from the worker‟s perspective,
and the firm faces a rising supply price of labour, even when many employers formally
compete in the same labour market (Manning 2010).
Examined in more detail, the first class of models involves „classic differentiation‟: i.e., the
dispersion of job attributes and worker preferences in general, and those concerning
geographical location in particular. As a result of fixed costs of market entry, firms do not
exist at all places in geographical space and jobs are therefore differentiated in the eyes of
workers by their distance from the worker‟s home – a form of product differentiation
analysed by Salop (1979). Assuming non-trivial travel-to-work costs, the geographical
heterogeneity of jobs and workers then gives the firm market power over workers who live
nearby (Bhaskar and To, 1999; Hamilton, Thisse and Zenou 2000; Brueckner, Thisse and
The second class of model involves the economics of job search (Burdett and Mortenson
1998). In these models, jobs are assumed to be identical, and workers will now leave their
employers for another job if the wage exceeds their current wage. The problem is the
difficulty of finding such an employer, given the cost of job search. Monopsony power is
high when the arrival rate of job offers is low – as it is when search costs are high. The
supply of labour to the firm increases with its wage, in that firms that pay higher wages lose
fewer employees to their competitors. In equilibrium, where employment is constant, the
flow of new hires is equal to the flow of workers leaving the firm. The number of quits that a
firm avoids by offering higher pay is equal to the number of recruits it attracts. The firm‟s
elasticity of labour supply can be estimated from the association between its wage and its
separation and recruitment rates, with a potentially important difference between the short
run and the long run (Manning 2003b, Manning 2006). The low elasticities reported in recent
research2 require search costs so high as to suggest not that any higher wage job can be
1 Thus Staiger, Spetz and Phibbs (2010) find that deregulation-generated wage increases at particular hospitals
affected pay at other hospitals in the same locality, but had little effect on their employment, which suggests
considerable monopsony power in local labour markets for nurses.
2 A number of recent studies have adopted this very approach to obtain empirical estimates of the labour supply
elasticity. Ransom and Sims (2010) estimate the labour supply elasticity for school teachers in Missouri of 3.7,
which implies that school districts possess significant monopsony power. Hirsch, Schank and Schnabel (2010)
report elasticities ranging from 1.9 to 3.7 in the German labour market, with women‟s elasticity being always
lower compared to men‟s. Ransom and Oaxaca (2010) find an elasticity ranging from 2.4–3 for male and 1.5–
2.5 for female workers in a regional grocery retailer located in the United States.
obtained by any job searcher, but rather that vacancies themselves are scarce (Manning
2.2 Monopsony power and training
Monopsony power holds centre stage in the economics of work-based training. Mainstream
models3 address what had previously been seen as an anomaly in human capital theory
(Becker 1964): employers‟ investments in „general‟ skills (Dionisius et al. 2009), which
could not happen in perfect competition. The anomaly can be explained by assuming
imperfect competition, in the shape of monopsony power for employers over skilled labour,
and, in particular, of greater monopsony power over skilled than unskilled labour. That paired
assumption generates wage compression, defined as a growth with the amount of training
received in marginal product that exceeds that in pay. Given wage compression, employers
maximise profits by reducing unskilled employment in favour of skilled employment;
training provides the link between the two adjustments (Stevens 1994, 2001; Acemoglu and
Pischke 1998, 1999a, 1999b; Dustmann and Schönberg 2009).4
The assumption that monopsony power applies to skilled workers is typically justified in
terms of an underlying heterogeneity of employers. Stevens (1994) assumes that the content
of skill requirements in any given occupation differs across employers, and that the value to
other employers of a firm‟s skilled workers is therefore less than their value to the firm itself,
which has trained them for its own requirements exactly.
By contrast, the typical assumption about pay setting for unskilled workers and trainees,
whether implicit or explicit, is perfect competition, which equalises marginal products and
pay across employers. „Free entry at the start of period 1 makes that w1 [trainee pay] is set
such as to drive expected profits to zero‟ (Leuven 2005: 97).5 The assumption corresponds to
3 For a recent survey article see Leuven (2005).
4 Acemoglu and Pischke (1999a) note that even if the monopsony-based rate of exploitation (ratio of marginal
product to the wage) is the same for trainees as for skilled workers, the firm‟s absolute surplus is greater for the
skilled worker, which creates some incentive to provide training.
5 See also Winkelmann (1996), Chang and Wang (1996) and Stevens (1999: 21). Acemoglu and Pischke
(1999b) also analyse a „fully competitive‟ regime, in which the firm pays trainees only what is necessary to
obtain their services, while competition for trainees denies the firm any super-normal profits. They do however
also analyse two alternatives, a „constrained‟ and a „cooperative‟ regime, in which trainees cannot or will not
accept the competitive level of pay, though the source and feasibility of such non-competitive behaviour is not
an expectation that unskilled workers are equally valuable to employers across the entire
range of sectors and job tasks.
This pair of contrasting assumptions – monopsony for skilled labour, perfect competition for
unskilled labour and trainees – has become the norm in the training literature. That status may
be attributable partly to those models‟ central prediction, viz. that employers rationally incur
costs for „general‟ training, and partly to the analytical simplification that the assumptions
permit. In any case, the paired assumptions are widely viewed as having general applicability
(Manning 2003b: 304).
That view may however be unwarranted. Some evidence suggests monopsony in markets for
trainee labour. The surpluses earned during training by some German and many Swiss
employers suggest less than perfectly competitive markets for trainee labour (Dionisius et al.
2009). Similarly, the effectiveness of the episodic episodes of industrial action conducted by
apprentices in British metalworking industry in the last century is inconsistent with the
assumption of perfect competition for trainee labour, but consistent with monopsony (Ryan
Manning (2003b: 302-5) observes that monopsony power may not only be present over
unskilled workers and trainees, it may even be stronger than over skilled workers. He sees a
potential cause in a higher rate of recent labour market entry among young trainees, resulting
in lower wage elasticities of separations and accessions than among adults.
Greater monopsony power over trainee labour than over skilled labour may be expected from
other attributes of the two groups too. When trainees are young (teenagers or young adults)
and living with and subsidised by their parents, the costs of geographical mobility tend to be
proportionately higher for them than for adult employees, which increases their dependence
on local employers, and which, if local employers are few, generates greater monopsony
power. Secondly, although younger workers have the incentive (because of a longer expected
working life) to invest more in labour market search, matching and mobility than do older
ones, their inexperience may still mean a smaller stock of accumulated information, and with
it more dependence on their current employer.
Conversely, other factors may make monopsony power weaker over trainees than over other
workers. As trainees are typically young, they may have more alternatives to wage-earning –
including full-time education and family-financed leisure – than do skilled workers. They
may also have lower requirements for labour income, given fewer family responsibilities,
which may reduce their dependence on, not only on the labour market in general, but also any
particular employer. We therefore treat it as an empirical issue, whether trainee pay is
affected by monopsony power, and, if it is, whether it is more affected than is skilled pay.
The results are potentially important, as the predictions of economic models of training may
be sensitive to assumptions about pay-setting. In particular, the phenomenon of production-
oriented training, in which firms provide training more in order to reduce short-term
production costs than to increase future skill supply, lies outside the explanatory scope of
mainstream models of training, but it is potentially explicable in terms of what might be
termed wage decompression, based on greater monopsony power over trainee than over
skilled labour (Wolter and Ryan 2010).6
A further potential difficulty in mainstream models of training, implicit in the above
discussion, is the assumption that trainees and unskilled workers can be treated as a single
category, in terms of pay-setting. The assumption is consistent with the fact that any firm can
in principle make its unskilled employees into trainees and thus convert them into skilled
workers. Informal on-the-job training given to existing adult employees at the workplace can
plausibly be described in such terms. In particular, such training typically involves no
reduction in pay for trainees during training (Ryan 1984; Barron, Berger and Black 1999).
The same may not however be the case for formalised long-term training for intermediate
skills – a category that includes, in both Europe and North America, apprenticeship
programmes, a type of training on which much of the evidence of employers‟ investment in
general skills is based. In such programmes, trainees typically differ from unskilled
employees, in terms of educational attainment, age, experience, seniority and pay in
particular. Their lower pay is no surprise, as training costs are high and skills are certified and
therefore presumptively general. But it may also be affected by differences in average age
6 Manning (2003b) predicts from the assumption of greater monopsony power over trainees than over skilled
labour that firms will be unwilling to offer training, forcing workers to sponsor their own training apart from
employment ('education'). An alternative possibility may however be production-oriented training, and the
exploitation of trainees as production labour.
and length of service (lower) and educational attainments (higher), and also by the extent to
which training contracts are institutionally distinct from regular employment contracts
(Marsden and Ryan 1991; Ryan et al. 2010). For these reasons, pay setting for trainees may
differ from that for unskilled workers, and with it the extent of monopsony power.
Our data come from a recent (2004) administrative survey of the costs of apprenticeship
training to Swiss employers.7 The establishments involved are a random sample taken from
the Establishment Register, operated by the Swiss Federal Statistical Office, which contains
all non-agricultural establishments in the country. After classifying establishments as to
whether they provide apprenticeship training or not and sampling randomly within each
category, our data comprise 2,413 training establishments, with a total of 13,454 apprentices,
and (with a lower sampling frequency) 1,863 non-training firms.8
Our evidence for trainees concerns work-based training for skilled occupations in
Switzerland. The training is organised as apprenticeship, combining training and work
experience at workplaces with part-time vocational education at public colleges. It resembles
closely its German counterpart, a centre of interest in recent economic analyses of training,
not least in its clear distinction between the status of trainee and employee. Formally
recognised apprenticeship training programmes, lasting between two and four years, operate
in some 250 occupations, distributed across all sectors of the economy. More than three fifths
of young people enter this type of training, typically after completion of lower secondary
schooling (Hoeckel et al. 2009: 13; OPET 2009). The advantage of evidence for Switzerland
lies in the availability of data for both particular employers‟ training programmes and the
local labour markets in which they operate.
The survey collected information on the pay of skilled and unskilled employees, trainees, and
on such attributes of the establishment as its size, geographical location, sector and the
7 The series of surveys of the cost to employers of apprenticeship training in Germany in recent decades, which
began with the Edding Commission (1974), is reviewed in Dionisius et al. (2009). The Swiss survey used
broadly the same methodology as its German counterparts.
8 Most establishments do not provide training: in 2005 only 17.8 per cent of all establishments provided
apprenticeship in any occupation (FSO 2005), though the share rises to around one-third if one-person
businesses are excluded (Muehlemann et al. 2007: p. 144). The Swiss Federal Statistical Office used a sampling
rate of 1.0 per cent for non-training establishments and 3.4 per cent for those that train. The data represent the
entire national training system, excluding only a handful of agricultural apprentices.
occupations for which it trains. „Skilled‟ and „unskilled‟ employees are distinguished
according to whether they have completed an apprenticeship in the occupation in which they
are currently employed. The „unskilled‟ category therefore includes workers without any
Pay is measured as the annual pre-tax basic pay of full-time employees and trainees, averaged
across all individuals (apprentices or employees) in the occupation in the establishment.
Calculated from its monthly counterpart, it includes employee social security contributions,
but excludes 13th and 14th month additional payments and any other non-regular payments,
including performance bonuses.10
Summary statistics for pay and establishment attributes are reported in Table 1, for all firms
and for training firms only; full definitions of the variables are given in the Appendix (Table
A1). Skilled workers‟ pay exceeds unskilled pay by around 30 per cent on average. Unskilled
pay is 4.5 times as high as trainees‟ pay, and skilled pay 6.5 times higher. The pay of skilled
employees is 5.2 per cent higher in non-training than in training establishments. As unskilled
pay differs little between the two categories, wage compression is greater in establishments
with training than in those without it.
The data also include establishment size, the amount of workplace-based training provided
per trainee, and qualitative information on the profitability of and the technology used by the
9 Unskilled workers may however have some continuing or informal training but not enough for their skills to be
certified, even though Switzerland has a system for certifying informally acquired skills. The unskilled category
therefore comprises all workers whose skill levels are below those of certified skilled workers.
10 An alternative, wider definition of pay, including additional payments and employers‟ social security
contributions and therefore close to total labour cost, was also considered. Base pay amounts on average to 81.3
per cent of this measure of total labour cost (for skilled and unskilled employees combined). When pay is
defined instead as total labour cost, the results of the regressions in section 6 change only slightly, and the index
of monopsony power (number of firms per hectare) remains highly significant.
Table 1: Descriptive statistics
All firms (N=3592):
Annual pay of skilled worker („000 CHF) 65.579 9.716 36.000 120.00
Annual pay of unskilled worker in firm („000 CHF) 47.475 5.100 24.000 80.640
Skilled worker / unskilled worker pay ratio 1.378 0.177 1.000 1.750
Number of firms in area per hectare (sector) 0.035 0.038 0.001 0.145
Number of firms in area per hectare (all sectors) 0.256 0.158 0.010 0.562
Number of farms per hectare 0.023 0.008 0.002 0.043
Firm size 1-9 employees 0.706 - 0 1
Firm size 10-49 employees 0.234 - 0 1
Firm size 50-99 employees 0.032 - 0 1
Firm size 100+ employees 0.028 - 0 1
Profitability (1=very good /rather good, 0=medium/
rather poor /very poor)
Technical status of equipment (1=very good /rather
good, 0=medium/ rather poor /very poor)
Training firms (N=2243)
0.361 - 0 1
0.533 - 0 1
Annual pay of skilled worker („000 CHF) 63.352 10.834 36.000 120.000
Annual pay of unskilled worker in firm („000 CHF) 47.796 54.518 24.000 80.640
Annual average apprentice pay (‟000 CHF) 10.313 24.402 3.647 24.675
Skilled worker / unskilled worker pay ratio 1.322 0.188 1.000 1.750
Apprentice / unskilled worker pay ratio 0.218 0.056 0.092 0.513
Skilled worker / apprentice pay ratio 6.422 1.654 2.301 15.668
Number of firms in area per hectare (own sector) 0.031 0.034 0.001 0.145
Number of firms in area per hectare (other sectors) 0.213 0.135 0.001 0.562
Number of firms in area per hectare (all sectors) 0.244 0.153 0.010 0.562
Number of farms per hectare 0.023 0.008 0.002 0.043
Firm size 1-9 employees 0.557 - 0 1
Firm size 10-49 employees 0.321 - 0 1
Firm size 50-99 employees 0.053 - 0 1
Firm size 100+ employees 0.068 - 0 1
Hours of formal training per apprentice/week 6.022 3.337 0.500 20.000
Profitability (1=very good /rather good, 0=medium/
rather poor /very poor)
Technical status of equipment (1=very good /rather
good, 0=medium/ rather poor /very poor)
Source: 2004 Survey.
0.382 0.486 0 1
0.549 0.498 0 1
We enlarge the survey data with locational information, concerning the commune (i.e.,
municipality, local authority) in which the establishment is situated, taken from the Swiss
Federal Statistical Office. The key variable is the number of (non-agricultural) establishments
within the establishment‟s local labour market, differentiated by two-digit sectors, taken from
the Swiss establishment register, which is administered by the Swiss Federal Statistical
Office. We take from the census of establishments in agriculture the number of farms in the
locality, as a measure of the degree of rurality of the local labour market.
We take local labour markets to be based on particular towns, and include all establishments
that can be reached from the town centre within 30 minutes by car under normal traffic
conditions.11 This criterion is preferable to the two alternative criteria that dominate the
empirical literature on monopsony: the administrative district (e.g., town and city boundaries)
and travel distance (e.g., a 30 kilometre radius from the town centre).12 We define the 67
largest Swiss towns and cities as each constituting the centre of a local labour market, and
classify all establishments as belonging to one, and only one, of those markets.13 If the
establishment is situated in a town that is not large enough to be treated as the centre of a
labour market, it is classified as belonging to the closest one.
Taking monopsony power as arising from buyer concentration in the labour market, we
measure it as the number of establishments in an establishment‟s local labour market divided
by the area covered by the market: i.e., number of establishments per hectare.14 Other
scholars have used a two-dimensional conception of monopsony power in local labour
11 In our data, travel time to particular establishments was measured with the software Microsoft Autoroute
2005. The 30 minute criterion is chosen in the light of actual travel to work times: in 2000, 84 per cent of Swiss
employees spent no more than 30 minutes travelling to and from work (one way trip; FSO 2000). Duranton and
Overman (2005) find for the UK that the localization of employment takes place mostly within 50km, which
broadly corresponds to a travel time of 30 minutes. Similarly, Manning (2003a) reports for Britain that more
than 80 per cent of travel-to-work times last no more than 30 minutes.
12 Commuting cost is in principle a function of both time (foregone earnings or leisure) and distance (fares,
gasoline, etc.), with a positive correlation between time and distance. The correlation is however limited and in
some cases (e.g. when both home and job are located beside a freeway or railway line) a long distance does not
necessarily imply long time of travel. We implicitly treat the costs of time as more important than the costs of
distance in cases of divergence.
13 In densely populated areas, such as the region around Zurich, travel-to-work districts overlap, in the sense that
an establishment located at or near the intersection between two districts can potentially recruit workers from
14 An alternative to the number of employers is total employment in the local market. Neither measure is ideal,
and in the case of literal monopsony (a „company town‟), total employment would be distinctly inferior to
number of employers as a measure. Data for the ideal measure (number of vacancies) are not available,
however, and in any case the two measures are highly correlated and generate broadly similar statistical results,
but with slightly weaker associations when total employment is used.
markets: the size of the market and the (absolute) number of employers (in the same
occupation and sector; Hirsch and Schumacher 1995). Our use of travel-to-work areas rather
than local government administrative ones allows us to compress these two dimensions into
one by standardising the number of employers for the geographical size of the market.
We allow for potential sector-specificity in competition for labour by distinguishing between
establishments in the same (two-digit) sector and those in other sectors. To the extent that
skilled occupations are sector-specific, competition for skilled labour is expected to count
only within the sector of the establishment in question. Thus, in Silicon Valley, competition
may be expected to be fierce for IT software engineers, but not for pipeline welders, and this
favours the use of the in-sector rather than the all-sector measure of fewness of firms. By
contrast, for accountants, an occupation present in all sectors, the all-sector measure is
expected to be superior. As systematic data on the sector-specificity of skills across
occupations are not available, we measure buyer concentration on both the in-sector and the
all-sector bases and compare the results attained on each basis.15
Table A2 in the Appendix provides summary statistics for the number of establishments
within individual local labour markets by two-digit sector. For example (first row), the
number of establishments within mining and quarrying averages nearly 20 across the 63 local
labour markets, and ranges from a minimum of zero to a maximum of 50. Across all sectors
(penultimate row), the average is 416 establishments within the individual sector and local
labour market; the minimum number averages ten, the maximum number 1685. These data
suggest that competition for labour – that is, setting aside labour market frictions – is
potentially substantial overall, and highly variable across localities, both within particular
sectors and across sectors as a whole.
Finally, we need to control statistically for other local attributes that not only potentially
influence pay but also do so in association with fewness of employers. The notable factor is
rurality: living costs, and with them pay, are expected to be lower in the countryside and in
country towns than in large towns and cities. We therefore include a measure of the
importance of farming in the local labour market, in terms of the number of farms per
15 A two-way classification of employment by occupation and sector would be required to judge the extent of
sector specificity in skills, but no such data are available for Switzerland.
4. Estimation and results
This section discusses the issues that arise in estimating monopsony effects on pay from
cross-sectional data on establishments and local labour markets. It then presents the
estimates, distinguishing between results for all establishments and for those that offer
4.1 Empirical strategy
Our empirical approach contains two components: first, an analysis of standardised absolute
levels of pay within particular occupations; second, an analysis of occupational differentials
in pay between skilled employees, unskilled employees, and trainees. The latter results are of
particular interest for two reasons: first, they implicitly control for the various firm-specific
and location-specific factors that have been seen as distorting those previous studies of
monopsony power that exploit variation across local labour markets; second, they bear
directly on the economics of training, as noted in section 2 above.
In more detail, our first dependent variable is average pay in occupation k in establishment i,
relative to mean for pay in that occupation in the country as a whole:
where is average pay in occupational group k (skilled, unskilled16, trainee) in
establishment i, and k is average worker pay across establishments in the same occupational
group (at national level); Mki is our indicator of monopsony power (number of establishments
per hectare in the same local labour market as establishment i, distinguishing between
establishments in the same sector and in other sectors)17; Xi, a vector of establishment
attributes (size, technical status and profitability). Controls for occupation and sector are
16 We standardise unskilled pay as the ratio of unskilled pay observed in establishment i employing skilled
workers in occupational group k, divided by average pay of unskilled workers in all establishments in the whole
country employing skilled workers in occupational group k. (When we standardise instead by unskilled pay in
the whole country, our empirical results with respect to monopsony power remained qualitatively unaffected.)
17 Swiss Census data do not permit us to identify (as an establishment‟s local competitors) only other
establishments that have employees in the skilled occupation by which the establishment in question is
represented in the survey. We therefore take its competitors to be all local establishments that operate in the
same sector (product market). The closer the fit between principal occupation and sector across establishments,
the less is this discrepancy. Some occupations are highly sector-specific, such as retailing salesperson (to
retailing); others less so, such as administrator (Kaufmann/frau) to banking.
included. The occupational categories (k) for the skilled worker and trainee groups comprise
the 23 most important occupations in Switzerland, which accounted for roughly 70 per cent
of all apprenticeship contracts in 2004 (and thus cover the largest categories of skilled
occupations by employment, see Potterat 2006).18
Thus, the advantage of our empirical approach is that we analyse average pay in a single
occupational category in an establishment, rather than the average pay of skilled workers in
all occupations in that establishment, in order to reduce any uncontrolled occupation-specific
effects on pay.
Our second dependent variable is relative pay, i.e., the ratio of mean pay in one occupational
group (skilled, unskilled, trainee) relative to that in another group:
where is average pay in occupation group k, and in occupation group j, in
establishment i. When the two groups are skilled employees and unskilled employees, this
formulation tests for wage compression associated with monopsony power ( ).
We control for different establishment characteristics, represented by the vector . Pay may
be expected to vary with such employer characteristics as establishment size, technology, and
profitability. To some extent this is predicated on monopsony power: in particular, given an
upward sloping supply curve, larger establishments have to offer higher pay to maintain their
higher level of employment. But even in the absence of monopsony power, companies that
are more technologically advanced, productive and profitable are expected to pay more, as
part of widespread rent-sharing in labour markets (Manning 2010). We therefore include both
establishment size and a combined indicator of the technological sophistication and
profitability of the establishment.19 Finally, from the human capital standpoint, firms that
18 The remaining occupational categories are grouped by the length of the training required to obtain the
professional qualification (two, three and four years).
19 The survey provides only qualitative, self-assessment measures of technological level and profitability,
summarised in Table 1 above. Owing to a high (positive) correlation between the two measures, we combine
them into a single indicator that takes on the value 1 if either the technological level or profitability have a
reported value of “very good” or “rather good”, and the value 0 otherwise.
provide more training are expected to offer lower pay to trainees. We therefore control at
establishment level for the amount of training provided per trainee.
Weakness or absence of controls for other pay-relevant attributes of localities have impaired
much of the empirical literature on monopsony power in local markets (Boal and Ransom
1997). The major prospective concern is living costs: the prices of housing, food, fuel, etc.
vary considerably across localities, and are expected to cause pay to vary too. As it is normal
for local labour markets in rural areas to have both few employers and low living costs,
controls for living costs are necessary to avoid biased estimates of monopsony effects.
Lacking measures of living costs by locality, we address the problem in three ways. First, we
control statistically for rurality by including number of farms (per hectare) in the
regressions.20 Second, we use within-establishment pay differentials to control for locality
effects. If differences in living costs across districts are assumed to have the same
proportionate effect on the pay of skilled workers, unskilled workers, and trainees alike, any
association between relative pay and fewness of firms can be interpreted as evidence of
monopsony power. Third, while we analyse the effect on pay of the total number of
establishments within a local labour market, we also analyse the effect of the number of
establishments within the establishment‟s own sector. Thus, our empirical measure allows
monopsony power to exist in any local labour market, not just in localities with low levels of
Lacking employer-employee matched data, we cannot control directly for differences
between establishments and localities in employee attributes, notably schooling and ability.
Insofar as those attributes are associated with fewness of employers, our estimates of
monopsony effects on pay become biased. Although number of employers and labour quality
may not be associated across establishments within a sector, low pay may reflect low labour
quality rather than monopsony power, and rural labour markets may have both fewer
employers and lower labour quality.
Some reassurance is provided by two considerations. First, to the extent that establishments
with higher pay employ proportionally higher labour quality in all occupations, our results for
relative pay are unaffected by the lack of explicit controls for labour quality. Second, a
20 We also used in the regressions alternative measures of rurality: number of cows per hectare or farmland per
hectare. The three variables are highly inter-correlated and the results are not sensitive to which is chosen as
minimum educational attainment is present: first, all skilled employees possess a federal
apprenticeship certificate in upper secondary education21; second, trainees are required by
law to have obtained a certificate in lower secondary education before entering training.
Finally, as in all cross-sectional analysis, concerns arise about endogeneity. Can differences
in pay across establishments be taken as exogenous, or may they be jointly determined with
employment? A possible source of endogeneity might be the locational decisions of
employers and workers: to the extent that firms move to low wage labour markets and
workers to high wage ones, pay and employment are jointly determined. In particular, local
labour markets with low pay can attract in-migration by employers, reducing the monopsony
power of the existing employers. Such tendencies are however expected to be weak, given
that firms typically face mobility costs too, based e.g., on: distance from customers;
transportation costs; taxes; and loss of local externalities (LaFountain 2005, Devereux et al.
We first test for monopsony power in the pay of skilled and unskilled employees in the
sample of all establishments. The second step is to test for it for trainees as well, which
means restricting the sample to establishments that provide apprenticeship training. We
present results for both absolute pay and relative pay; in all equations are estimated by
ordinary least squares, using cluster-robust standard errors.
Starting with the full sample, we test for monopsony power in absolute pay using equation (1)
above. The standardised absolute pay of skilled workers proves positively and significantly
associated, as predicted, with the number of employers in the local labour market. The
coefficient of 0.114 in column 1 of Table 2 means that a one standard-deviation increase in
21 The Eidgenössisches Fähigkeitszeugnis (EFZ) is a certified federal professional qualification at upper
secondary education level (OPET 2009).
22 Instrumental variables regression would in principle address directly the issue of endogeneity. To be valid, an
instrument would have to be associated with the number of competitors in the locality without itself having any
any direct impact on pay levels or differentials. Our data do not however contain information that could be used
as an instrument.
the number of establishments in the local labour market is associated with an estimated
increase in the pay of skilled workers in the establishment of 1.8 per cent. The same does not
however apply to unskilled employees: changes in the number of local establishments are not
significantly associated with their pay (column 3).
When the number of local establishments is divided into those in the same sector and those in
other sectors, skilled pay proves positively associated with both (column 2). A one standard-
deviation increase in the number of establishments in the same sector is associated with an
increase in relative pay of 1.1 per cent; in other sectors, by the same amount. Neither variable
is significant for unskilled employees (column 4).
These results align with the standard assumption of many economic analyses that pay is set
by monopsony power for skilled workers and by perfect competition for unskilled ones. They
are however potentially impaired by the difficulty of controlling for fixed effects by
establishment and locality. The difficulty is in principle removed when pay is analysed in
relative terms – i.e., using differences in skilled and unskilled pay within an establishment to
account for unobserved heterogeneity in the establishment and the locality.
The pay of an establishment‟s skilled employees remains positively associated with the
number of local establishments when expressed in relative terms, using unskilled pay as the
base. A one standard-deviation increase in the number of firms is associated with an increase
in the skilled-unskilled pay differential of 1.6 per cent (column 5). When the sectoral
affiliation of establishments is considered, a one standard-deviation increase in the number of
establishments is associated with increases of 1.0 and 0.6 per cent in the relative pay of
skilled employees, for „same sector‟ and „other sector‟ variants, respectively (column 6). This
pattern suggests that competition for skilled labour has a somewhat stronger effect on relative
pay when it occurs within a sector than across sectors, but that – consistent with cross-
sectoral employment in some skilled occupations – competition across sectors is also
Concerning other influences on pay, we find the standard positive relationship between
establishment size and pay that is the predicted consequence of rising supply price,
presumptively caused by monopsony power. The relative pay of skilled workers in
establishments with 100 or more employees is eight per cent higher than that in those with
less than ten employees. The size effect on pay is also significantly positive for unskilled
workers, but somewhat weaker than for skilled ones.
The possibility that competition for labour has different effects on pay according to the size
of establishment – e.g., affecting pay less strongly in large than in small plants – is examined
by interacting establishment size and number of establishments.23 No significant interactions
Rurality, a proxy for local attributes potentially related to living costs, also matters: the
number of farms in the local labour market – interpreted as an inverse proxy for living costs –
is negatively and significantly associated with pay in all equations. A one standard-deviation
increase in the number of farms is associated with a reduction of 1.2 per cent in skilled pay,
one of 0.5 per cent in unskilled pay, and one of between 0.6 and 0.8 per cent in the skilled-
unskilled pay differential, depending on whether the sectoral affiliation of local
establishments is distinguished or not.
Our measure of the condition of the establishment, based on managers‟ reports of equipment
quality and profitability, is positively associated with pay for skilled employees, but not for
unskilled ones. Firms that use more modern technology or are more profitable have around
two per cent higher pay for skilled workers but no pay premium for unskilled ones. All
regressions involve controls for unobserved fixed effects by sector.
We noted in section 2 above that pay setting for trainees may differ from that for skilled and
unskilled workers, and that employers may possess market power over trainees, particularly
apprentices, even when they do not over unskilled workers.
We therefore widen the scope to include pay determination for trainees, which means
restricting the sample to establishments that offer apprenticeship training. We first re-estimate
the absolute pay and the relative pay equations (1 and 2, above, respectively) for training
firms only, using the equivalent equations for trainees. We also control for the amount of
training provided by including a measure of average training time provided per trainee, as
23 Large establishments might enjoy more monopsony power than small ones, for a give number of competitors
– so that, where one very large plant competes with many very small ones, number of establishments might not
capture well the monopsony power of the large establishment.
measured by the weekly hours that skilled workers are removed from productive work on
average in order to provide training.
The results show that the number of local establishments is associated with the absolute level
of pay across training establishments, as for all establishments. A one standard-deviation
increase in the number of local establishments is associated with an increase of 1.5 per cent in
skilled pay among training firms, as compared to 1.8 per cent in the full sample (Table 3, col.
1).24 When local competitors for labour are distinguished according to „own sector‟ or „other
sector‟ status, the coefficients on both variables are correctly signed, but neither is
statistically significant (p=.05; loc cit., column 2).25 Nor does any significant effect emerge
for the number of establishments, however the variable is defined, in the case of unskilled
pay. The results for the control variables, notably the degree of rurality, are similar to those
for the „all establishments‟ sample, except that the technology-profitability index is now
Including trainees, as distinct from unskilled employees, we find monopsony effects on
absolute pay. A one standard-deviation increase in the number of local establishments is
associated with an increase of trainee pay of 1.9 per cent (Table 3, col. 5). When the sectoral
affiliation of establishments is brought in, the effect is found to involve the „same sector‟
measure, a one standard-deviation increase in which is associated with an increase in trainee
pay of 3.3 per cent. The number of establishments in other sectors, by contrast, has no
significant association with trainee pay.
24 As economic theories of training predict that training firms have more monopsony power than non-training
ones, the difference between the coefficients might be viewed as anomalous. Our result – quite apart from its
small size, statistically speaking – bears however on a different issue: whether an increase in monopsony power
has more effect on pay among training firms than among other ones.
25 The coefficient on the „other sectors‟ variable is significant at p=.10. Statistical insignificance in the training
establishments sub-sample, in contrast to the results for all establishments, is associated with higher standard
Table 2: Regression results: skilled pay, unskilled pay and pay differentials
Number of establishments in locality per hectare (all sectors)
Notes: Cluster-robust standard errors in parentheses; n = 3562
Less skilled pay
Number of establishments in locality per hectare (own sector)
Number of establishments in locality per hectare (other sectors)
Number of farms per hectare
Establishment size 10-49 employees
Establishment size 50-99 employees
Establishment size 100+ employees
Technical status of equipment / profitability
0.105 0.105 0.449 0.449
This evidence of monopsony power in apprentice pay setting is inconsistent with the
assumptions of contemporary models of training, which for the most part treat trainees and
„unskilled‟ workers as identical and assume competitive pay setting for both. Also striking is
the evidence that non-competitive effects on trainee pay involve the number of
establishments in the same sector (i.e., ex post, the sector in which training is occurring), not
the number in other sectors: trainees might be expected to lack sector affiliation before they
start training. If so, competition should therefore be measured, and matter statistically, on an
„all sector‟ basis rather than an „own sector‟ one. The result may be interpreted as evidence
that potential trainees develop affiliations (preferences), by sector and implicitly by
occupation (e.g. hairdressing), before they start training, combined with a high success rate
for would-be trainees in terms of finding a training place in their preferred sector and
occupation.26 Our result suggests that offers of training by employers in sectors other than the
desired one are not valued highly, even when well paid.
We now analyse pay differentials between the three occupational categories (skilled,
unskilled, and trainees), as representing the specification for which uncontrolled
establishment and locality effects may be expected to be weak. We consider first the pay
differential between skilled and unskilled workers. The number of establishments in the local
labour market again proves positively and significantly associated with pay structure. A one-
standard deviation increase in the number of establishments (all sectors) is associated with an
increase in the skilled-unskilled pay differential of 1.6 per cent (Table 4, col. 1) – a finding
that similar to that for all firms (Table 2, col. 5).
Separation of number of establishments according to „own‟ and „other‟ status produces, as for
absolute skilled pay, positive coefficients, but no significant association with either variable
(Table 4, col. 2).27
26 There is empirical evidence in favour of this hypothesis, as 70 per cent of Swiss apprentices report that they
get their first choice of training occupation (SKBF, 2010).
27 The number of firms becomes insignificant when the sample is restricted to training firms partly because of
the increase in standard errors, which may in turn reflect a reduction in identifying variation caused by non-
randomness in the selection of firms according to training status: training firms are expected to have more
monopsony power on average.
Table 3: Skilled, unskilled and trainee pay, training establishments only
Less skilled pay
(2) (4) (6)
Number of establishments in locality
Number of establishments in locality
Number of establishments in locality
Number of farms per hectare
Establishment size 10-49 employees
Establishment size 50-99 employees
Establishment size 100+ employees
Technical status of equipment / profitability
Hours of weekly training time x 10-1
Notes: N= 2243; Cluster-robust standard errors in parentheses.
(0.018) (0.018) (0.011) (0.011) (0.027) (0.027)
0.101 0.101 0.240 0.240 0.089 0.097
The number of farms, our measure of the rurality of localities, does not have a significant
association with relative pay among training establishments – a result consistent with the
similarity of its association with absolute pay in both employee categories (Table 3).
Similarly, the indicator used for technology and profitability is not significantly associated
with relative pay.
By contrast, plant size is positively related to the skilled-unskilled wage differential, which is
about 10 per cent higher in firms with 100 or more employees compared to firms with less
than 10 employees. The significance of plant size is consistent with its greater importance for
the absolute pay of skilled than for that of unskilled workers (Table 2, above).
Table 4: Within-establishment pay differentials in training firms
Number of establishments in locality
Number of establishments in locality
Number of establishments in locality
Number of farms per hectare
Establishment size 10-49 employees
Establishment size 50-99 employees
Establishment size 100+ employees
Technical status of equipment / profitability 0.005
Hours of weekly training time (x 10-1)
Skilled / unskilled Trainee / unskilled Skilled / trainee
(0.047) (0.047) (0.012) (0.012) (0.311) (0.313)
0.192 0.193 0.472 0.477 0.331 0.335
Notes: Cluster-robust standard errors in parentheses.
2243 2243 2243 2243 2243 2243
Turning to the pay differential between trainees and unskilled employees, we find a positive
association with number of establishments (all sectors). A one standard-deviation increase in
the number of establishments is associated with a 1.4 per cent increase in the relative pay of
trainees (Table 4, col. 3). This suggests that reductions in competition in a locality depress
trainee pay proportionately more strongly than they do skilled pay.
Breaking down competition by the two sector categories suggests – as for absolute trainee
pay (Table 3) – that it is competition in the same sector, not other sectors, that matters. A one
standard-deviation increase in the number of establishments in the same sectors is associated
with an increase in trainees‟ relative pay of 4.4 per cent (Table 4, col. 4), whereas an increase
in the number in the same sector is not associated with any significant difference in relative
Rurality, viz. the number of farms in the local labour market, has a significant association
with trainees‟ relative pay: a one-standard deviation increase in the number of farms is
associated with a decrease in relative pay of 1.4 per cent (Table 4, cols 3, 4). This reflects the
stronger negative association of absolute pay with rurality for trainees than for unskilled
employees.28 In addition, the relative pay of trainees and unskilled workers in establishments
with more than 100 employees is about 6 per cent (or 0.12 percentage points) higher than in
firms with less than 10 employees.
The number of weekly training hours provided by the employer has a significantly negative
association with relative pay, consistent with the predictions of human capital theory, and
reflecting lower absolute pay for trainees in establishments that provide more training. A one-
standard deviation increase in weekly training hours decreases trainee relative pay by 1.6 per
Finally, having tested for monopsony power over skilled employees and trainees, using
unskilled employees as an implicit control group, the question arises: is monopsony power
greater over skilled workers than over trainees? Influential models of training assume that
this is the case, and that it generates „wage compression‟. The issue is however a priori
ambiguous: some factors point to greater monopsony power for trainees than for skilled
workers, others to less (section 2.2).
Our data exceptionally permit us to test for monopsony power in terms of the pay differential
between skilled workers and trainees within occupations. The skilled-trainee pay differential
is not associated with number of local establishments when the latter is measured on an „all
sector‟ basis, but it is when establishments in the same sector (only) are considered (Table 4,
cols. 5, 6). A one-standard deviation increase in the number of establishments in the same
sector is associated with a 3.4 per cent decrease in the pay of skilled workers relative to that
of trainees (col. 6). As noted above, the pattern may be explicable in terms of cross-sector
competition for skilled labour in particular occupations, combined with strong ex ante
28 A result not shown in Table 4 is that the coefficient on number of establishment is affected only marginally by
the removal of the rurality variable.
preferences for sectors and occupations among potential trainees. It suggests that monopsony
power is associated with wage decompression, not wage compression, when trainees replace
unskilled employees as the numeraire.
The extent of wage compression is however not significantly associated with any of the
control variables – rurality, establishment size, technology/profitability, nor with training
intensity – apart, that is, from fixed effects by sector and occupation.
In sum, the results suggest that monopsony power, at least in the spatial dimension that our
data capture, is indeed associated with wage compression between skilled and unskilled
employees, as is widely assumed in the economics of training. The evidence also suggests, by
contrast, that monopsony power leads, under these circumstances at least, to wage
decompression between skilled employees and trainees. This result, which does not align
with the assumptions of standard economic models of training, points to the need, first, to
distinguish between trainees and unskilled workers, and, second, to consider production-
oriented motives in order to explain the supply of training by the employer (Wolter and Ryan
Using a direct measure of the „thinness‟ of local labour markets, we test for the presence of
monopsony power in markets for skilled labour, unskilled labour, and trainees in the
contemporary Swiss labour market. We control for establishment-specific and locality-
specific determinants of pay by focusing on pay differentials between skilled workers and
unskilled employees, on the one side, and trainees in the same establishment, on the other.
The results suggest that firms possess market power over skilled workers, and that differences
in that power are associated with differences in pay. The within-establishment pay
differential between skilled and unskilled labour increases with the number of competitors in
the local labour market. The finding contrasts to some earlier studies that found no effect for
employer concentration on pay when controls are imposed for other locality-specific factors.
A key difference from earlier studies is our definition of local labour markets, as dependent
on travel-to-work time rather than political borders or travel distance, which provides us with
a superior measure of the commuting costs that underlie monopsony power, as captured here.
Our evidence also covers a wider range of skilled occupations than in the previous literature,
which focused primarily on single occupations, such as nurse or school-teacher.
The pay effects we estimate are moderately large. A two standard-deviation increase in both
directions in the local number of establishments in the same two-digit sector, which embraces
the vast majority of cases, is associated, in the results for all establishments, with an increase
in the relative pay of skilled to unskilled labour of 4.0 per cent; the same increase in the
number of establishments in other sectors, with an additional 2.4 per cent. These magnitudes
are consistent with the view of monopsony power as „widespread but small on average‟ (Boal
and Ransom 1997: 110). At the same time, our evidence captures only differences across
establishments and only one source of monopsony power. More comprehensive measures,
were they available, could be expected to show bigger effects on pay.
Our finding of monopsony power for employers in the Swiss labour market opens new
perspectives on the economics of work-based training. We conclude that more monopsony
power is associated with lower absolute pay for skilled workers, but not for unskilled
workers. These results not only align well with the assumptions of mainstream models of
training, they provide unusually direct evidence in support thereof. In addition, however, we
find that firms have significant monopsony power over trainees, and that the effect on pay is
proportionately stronger for trainees than for skilled workers: a one standard-deviation
increase in the number of local employers in the same sector reduces the ratio of skilled pay
to trainee pay by 3.4 percent. Monopsony is associated with wage decompression, not
compression, in this dimension of training-related pay structure.
The stronger effect of monopsony power on pay for trainees than for skilled workers
indicates that the „action‟ when it comes to its stimulative effect on training involves trainees
as least as much as skilled workers. This suggests that economic models of training should be
extended to allow for greater monopsony power over trainees than over skilled workers. The
result may be an incentive to employers to supply higher levels of training than predicted by
some models of training under wage compression. The anticipated results are both a firmer
analytical basis for the economics of production-oriented training, and the prospect of an
economic theory of training that has greater generality than those that currently dominate the
Acemoglu, D. and J.-S. Pischke (1998), „Why do firms train? Theory and evidence‟, Quarterly
Journal of Economics, 63(1), 79-110.
Acemoglu, D. and J.-S. Pischke (1999a), „Beyond Becker: training in imperfect labour markets‟,
Economic Journal, 109, F112-42.
Acemoglu, D. and J.-S. Pischke (1999b), „The structure of wages and investment in general training‟,
Journal of Political Economy, 107(3), 539-72.
Ashenfelter, O., H. Farber and M.R. Ransom (2010), „Modern models of monopsony in labor
markets: A Brief Survey‟, Journal of Labor Economics 28(2), 203-210.
Barron, J.M., M.C. Berger and D.A. Black (1999), „Do workers pay for on-the-job training?‟, Journal
of Human Resources, 34(2), 235-52.
Becker, G. (1964), Human Capital. New York: Columbia University Press.
Bhaskar, V., and T. To (1999), „Minimum wages for Ronald McDonald monopsonies: a theory of
monopsonistic competition‟, The Economic Journal 109 (April), 190-203.
Bhaskar, V., A. Manning and T. To (2002), „Oligopsony and monopsonistic competition in labor
markets?‟, Journal of Economic Perspectives 16(2), 155-174.
Becker, G.S. (1964), Human Capital. New York: NBER.
Beicht, U., G. Walden and H. Herget (2004), Kosten und Nutzen der betrieblichen Berufsausbildung
in Deutschland. Bundesinstitut für Berufsbildung. Berichte zur beruflichen Bildung.
Boal, W.M. and M.R. Ransom (1997), „Monopsony in the labour market‟, Journal of Economic
Literature, March; 35(1), 86-112.
Brueckner, J.K., J.-F. Thisse and Y. Zenou (2002), „Local labor markets, job matching and urban
location‟, International Economic Review 43, 91-107.
Bunting, R. L. (1962), „Employee concentration in local labour markets‟, Chapel Hill: University of
North Carolina Press.
Burdett, K. and Mortensen, D.T. (1998), „Wage differentials, employer size, and unemployment‟.
International Economic Review 39(2), 257-273.
Chang, C. and Y. Wang (1996), „Human capital investment under asymmetric information: the
Pigovian conjecture revisited‟, Journal of Labour Economics, 4(3), 505-19.
Devereux, M. P., R. Griffith and H. Simpson (2007), „Firm location decisions, regional grants and
agglomeration externalities‟. Journal of Public Economics, 91(3-4): 413-435.
Dionisius, R., S. Muehlemann, H. Pfeifer, G. Walden, F. Wenzelmann and S.C. Wolter (2009), „Costs
and benefits of apprenticeship training: a comparison of Germany and Switzerland‟, Applied
Economics Quarterly, 55(1), 7-37.
Duranton, G. and Overman, H.G. (2005), „Testing for localization using micro-geographic data‟,
Review of Economic Studies, 72: 1077–1106
Dustmann, C. and U. Schönberg (2009), „Training and union wages‟, Review of Economics and
Statistics, 91(2): 363–376.
Dustmann, C. and U. Schönberg (2010), „What makes firm-based apprenticeship schemes work? The
role of commitment to training provision‟. Working paper, Department of Economics,
University College London.
Edding Commission (1974), Kosten und Finanzierung der auβerschulischen beruflichen Bildung
(Costs and Financing of Non-School Vocational Education and Training), Final Report,
Sachverständigenkommission. Bielefeld: W. Bertelsman Verlag.
Field, S., K. Hoeckel, V. Kis and M. Kuczera (2009), Learning for jobs: OECD policy review of
vocational education and training: initial report, Paris: OECD.
FSO (1998), Mikrozensus Familie in der Schweiz 1994/95. Neuchâtel: Swiss Federal Statistical
FSO (2000), Swiss Census 2000, Neuchâtel: Swiss Federal Statistical Office.
FSO (2004), Die Schweizerische Arbeitskräfteerhebung (SAKE) 2004, Neuchâtel: Swiss Federal
FSO (2005), Betriebszählung 2005, Neuchâtel: Swiss Federal Statistical Office.
Hamilton, B., J.F. Thisse and Y. Zenou (2000), „Wage competition with heterogeneous workers and
firms‟, Journal of Labor Economics, 18 (3), 453-472.
Hirsch, B. T. and E.J. Schumacher (1995), „Monopsony power and relative wages in the labor market
for nurses‟, Journal of Health Economics, 14, 443-476.
Hirsch, B., T. Schank and C. Schnabel (2010), „Differences in labor supply to monopsonistic firms
and the gender pay gap: an empirical analysis using linked employer-employee data from
Germany‟, Journal of Labor Economics 28(2), 291-330.
Hoeckel, K., S. Field and W.N. Grubb (2009), Learning for jobs. OECD reviews of vocational
education and training – Switzerland, Paris: OECD.
LaFountain, C. (2005), „Where do firms locate? Testing competing models of agglomeration‟,
Journal of Urban Economics, 58(2), 338-366.
Leuven, E. (2005), „The economics of private sector training: a survey of the literature‟, Journal of
Economic Surveys, 19(1), 91-111.
Link, C.R. and J.H. Landon (1976), „Market structure, nonpecuniary factors and professional salaries:
registered nurses‟, Journal of Economics and Business, 28(2), 151-55.
Manning, A. (2003a), „The real thin theory: monopsony in modern labour markets‟, Labour
Economics, 10(2003), 105-131.
Manning, A. (2003b), Monopsony in motion: imperfect competition in labour markets, Princeton:
Princeton University Press.
Manning, A. (2006), „A generalised model of monopsony‟, Economic Journal, 116, 84-100
Manning, A. (2010), „Imperfect competition in the labour market‟, Discussion Paper 981, Centre for
Economic Performance, London School of Economics.
Marsden, D.W. and P. Ryan (1991), „The structuring of youth pay and employment in six European
economies‟, in P. Ryan, P. Garonna and R.C. Edwards (eds), The Problem of Youth. London:
Muehlemann, S., S.C. Wolter, M. Fuhrer and A. Wüest (2007), „Lehrlingsausbildung – ökonomisch
betrachtet (Apprentice Training Considered from the Economic Standpoint)‟, Zurich: Rüegger
Muehlemann, S., H. Pfeifer, G. Walden, F. Wenzelmann and S. C. Wolter (2010), „The financing of
apprenticeship training in the light of labour market regulations‟, Labour Economics, 17(5),
OPET (2009), Vocational education and training in Switzerland 2009 – facts and figures, Berne:
Federal Office for Professional Education and Technology.
Potterat, J. (2006), Kosten und Nutzen der Berufsbildung aus Sicht der Betriebe im Jahr 2004,
Neuchâtel: Swiss Federal Statistical Office.
Ransom, M.R. and R.L. Oaxaca (2010), „New Market Power Models and Sex Differences in Pay‟,
Journal of Labor Economics, 28(2), 267-289.
Ransom, M.R. and D.P. Sims (2010), „Estimating the firm‟s labor supply curve in a “New
Monopsony” framework: schoolteachers in Missouri‟, Journal of Labor Economics 28(2), 331-
Robinson, J.F. (1933), The economics of imperfect competition, London: St. Martin‟s.
Ryan, P. (1984), „Job training, employment practices and the large enterprise: the case of costly
transferable skills‟, in P. Osterman (ed), Internal Labour Markets, Cambridge, Mass.: MIT Press,
Ryan, P. (2010), „Apprentice strikes, pay structure, and training in twentieth century UK
metalworking industry‟, in C. Brown, B. Eichengreen, and M. Reich (eds), Labor in the Era of
Globalization, Cambridge: CUP, 317-53.
Ryan, P., K. Wagner, S. Teuber and U. Backes-Gellner (2010), „Trainee Pay in Britain, Germany and
Switzerland: Markets and Institutions‟, Research Paper 96, Oxford: SKOPE.
Sachverständigenkommission (1974), Kosten und Finanzierung der ausserschulischen beruflichen
Bildung, Bielefeld: Bertelsmann.
Salop, S.C. (1979), „Monopolistic competition with outside goods‟, Bell Journal of Economics, 10,
SKBF (2010), Bildungsbericht Schweiz 2010, Aarau: Schweizerische Koordinationsstelle für
Staiger, D.O., J. Spetz and C.S. Phibbs (2010), „Is there monopsony in the labor market? evidence
from a natural experiment‟, Journal of Labor Economics 28(2), 211-236.
Stevens, M. (1994), „A theoretical model of on-the-job training with imperfect competition‟, Oxford
Economic Papers, 46(4), 537-62.
Stevens, M. (1999), „Human capital theory and UK vocational training policy‟, Oxford Review of
Economic Policy, 15(1), 16-32.
Stevens, M. (2001), „Should firms be required to pay for vocational training?‟, Economic Journal,
Winkelmann, R. (1996), „Less skilled labor and wage determination: An empirical investigation for
Germany‟, Journal of Population Economics, 9(2), 159-171.
Wolter, S.C. and P. Ryan (2010), „Apprenticeship‟, in R. Hanushek, S. Machin and L. Wössmann
(eds), Handbook of the Economics of Education, Volume 3. Amsterdam: Elsevier, 522-76.
Table A1: Definitions of variables
Mean annual pre-tax base pay in CHF (of apprentices, averaged across all
years of training, or unskilleda employees, or skilledb employees) in the
occupation and establishment, calculated from its monthly counterpart,
including employees‟ (and apprentices‟) social security contributions, but not
including 13th and/or 14th monthly pay, other non-regular payments
(performance bonuses, etc.) and employers‟ social security contributions
Ratio of mean pay in two of the three categories (apprentices, skilled
employees and unskilled employees)
Number of employees (apprentices excluded)
Total hours per week that skilled workers cannot work productively because
they are instructing apprentices
Number of establishments in local labour market in the same two-digit sector
/ Area of local labour market in hectares
Total number of establishments in local labour market / Area of local labour
market in hectares
Hours of formal training per trainee per
Number of establishments per hectare
Number of establishments per hectare
Notes: a. Employees who have not completed any type of post-compulsory schooling.
b. Employees who have completed apprenticeship training (i.e., a vocational qualification at upper secondary
level) for the occupation in which they are working.
Table A2: Number of establishments within local labour markets by two-digit sector
Mining and quarrying
Manufacture of food products and beverages
140.7 120 95.7 3 376
Manufacture of tobacco products
0.6 0 1.3 0 6
Manufacture of textiles and textile products
40.9 28 38.4 0 178
Manufacture of apparel; dressing and dyeing of fur
57.5 41 55.2 0 246
Tanning and dressing of leather; manufacture of luggage,
handbags, saddlery, harness and footwear
Manufacture of wood, and wood and cork products, except
furniture; manufacture of straw and plaiting materials
Manufacture of pulp, paper and paper products
14.4 15 9.0 0 34
314.2 278 191.7 22 913
15.0 10 14.1 0 52
Publishing, printing and reproduction of recorded media
295.4 195 301.8 1 1390
Manufacture of coke, refined petroleum products and
Manufacture of chemicals and chemical products
0.6 0 1.0 0 5
55.6 42 49.9 0 231
Manufacture of rubber and plastic products
47.7 39 39.3 0 170
Manufacture of other non-metallic mineral products
84.9 70 59.8 3 277
Manufacture of basic metals
16.1 13 12.3 0 52
Manufacture of fabricated metal products, except machinery
Manufacture of machinery and equipment n.e.c.
431.1 369 321.8 3 1440
196.6 159 153.0 1 691
Manufacture of office machinery, data processing devices
8.2 7 7.8 0 32
Manufacture of electrical machinery and apparatus n.e.c.
63.9 50 59.7 0 287
Manufacture of radio, television and communication
equipment and apparatus
Manufacture of medical, precision and optical instruments,
watches and clocks
Manufacture of motor vehicles, trailers and semi-trailers
39.7 30 38.0 0 205
192.2 146 166.1 0 759
10.6 8 9.7 0 39
Manufacture of other transport equipment
22.5 21 17.3 0 75
Manufacture of furniture, jewelry, musical instruments,
sports goods, games and toys and other goods
214.8 174 169.7 7 718
21.3 17 15.9 0 67
Electricity, gas, steam and hot water supply
35.1 33 18.8 3 91
Collection, purification and distribution of water
13.7 12 10.3 0 55
1906.0 1662 1350.6 60 6660
Sale, maintenance and repair of motor vehicles and
motorcycles; retail sale of automotive fuel
830.6 695 614.5 15 3062
33 Download full-text
Wholesale trade and commission trade, except of motor
vehicles and motorcycles
Retail trade, except of motor vehicles and motorcycles;
repair of personal and household goods
Hotels and restaurants
1193.6 898 1166.5 11 5906
2829.9 2057 2147.1 79 10205
1363.7 1172 836.5 102 4161
Land transport; transport via pipelines
445.7 353 353.2 14 1691
5.1 3 8.2 0 39
13.2 4 25.4 0 101
Supporting and auxiliary transport activities; travel agencies 239.5
163 239.5 6 1092
Post and telecommunications
274.5 233 180.8 20 828
245.0 176 246.3 9 1115
Insurance (except compulsory social security)
165.8 128 119.1 6 554
Activities auxiliary to financial intermediation
214.9 137 252.0 3 1128
Real estate activities
262.8 173 263.6 4 1292
Rental of machinery and equipment without operators and
of personal and household goods
Computing services and related activities
56.7 44 51.4 0 237
721.5 449 874.4 6 4310
Research and development
31.8 20 32.4 0 114
Other business activities
3533.1 2368 3767.8 47 18305
Public administration and defense; compulsory social
498.4 421 331.0 24 1483
865.4 686 679.2 26 3230
Health, veterinary and social work
1487.5 979 1274.7 15 5680
Sewage and refuse disposal, sanitation and similar activities 102.9
82 75.0 6 365
Activities of membership organizations n.e.c.
326.1 221 262.4 11 1087
Recreational, cultural and sporting activities
432.7 275 403.2 11 1844
Other service activities
804.5 592 652.7 13 3012
Mean across sectors (in-sector establishments only)
415.6 311.5 354.4 10.4 1685.1
Mean number of establishments (all sectors) per local
Notes: a. rounded up to the nearest integer where a tie occurs.
22,175 17,196 17,845.7 1188 85,746