firm-specific training and competition

8/18/2019 Firm-specific Training and Competition

1/16

Full Terms & Conditions of access and use can be found athttp://www.tandfonline.com/action/journalInformation?journalCode=raec20

Download by: [University of New Brunswick] Date: 03 February 2016, At: 08:5

Applied Economics

ISSN: 0003-6846 (Print) 1466-4283 (Online) Journal homepage: http://www.tandfonline.com/loi/raec20

Local employer competition and training of workers

Sylvi Rzepka & Marcus Tamm

To cite this article: Sylvi Rzepka & Marcus Tamm (2016): Local employer competition and

training of workers, Applied Economics, DOI: 10.1080/00036846.2015.1137550

To link to this article: http://dx.doi.org/10.1080/00036846.2015.1137550

Published online: 27 Jan 2016.

Submit your article to this journal

Article views: 5

View related articles

View Crossmark data

http://crossmark.crossref.org/dialog/?doi=10.1080/00036846.2015.1137550&domain=pdf&date_stamp=2016-01-27http://crossmark.crossref.org/dialog/?doi=10.1080/00036846.2015.1137550&domain=pdf&date_stamp=2016-01-27http://www.tandfonline.com/doi/mlt/10.1080/00036846.2015.1137550http://www.tandfonline.com/doi/mlt/10.1080/00036846.2015.1137550http://www.tandfonline.com/action/authorSubmission?journalCode=raec20&page=instructionshttp://www.tandfonline.com/action/authorSubmission?journalCode=raec20&page=instructionshttp://dx.doi.org/10.1080/00036846.2015.1137550http://www.tandfonline.com/action/showCitFormats?doi=10.1080/00036846.2015.1137550http://www.tandfonline.com/loi/raec20http://www.tandfonline.com/action/journalInformation?journalCode=raec20


2/16

Local employer competition and training of workers

Sylvi Rzepkaa and Marcus Tammb

a

RWI, and RUB, RWI, Essen, Germany; b

RWI, and IZA, RWI, Berlin, Germany

ABSTRACT

The new training literature suggests that in a monopsonistic market employers will not only payfor firm-specific training but also pay for general training if the risk of poaching is limited. Thisimplies that training should decrease with more competition for employees among firms. Usingworker-level data for Germany on training participation and on training duration, the authors findempirical support for this hypothesis. Specifically, the authors find that employees are signifi-cantly less likely to participate in training if the local density of firms in a sector is high and theyhave shorter training durations when the local sector concentration is low.

KEYWORDS

Training; local labourmarkets; monopsony

JEL CLASSIFICATION

I24; J24; J42

I. Introduction‘Why do firms train?’ Acemoglu and Pischke (1998)

asked. Still today this is a relevant question since

European governments seek to augment training

participation as they articulated in the Strategic

Framework for European Cooperation in Education

and Training (EU Council 2009). The rationale

behind this policy is clear: in countries such as

Germany where natural resources have become

increasingly scarce and where the service sector con-

tributes nearly 70% to the gross domestic product

(Statistisches Bundesamt 2013), flexible human capi-

tal that can adjust to new trends and technology is

crucial. Hence, to ensure the country ’s competitive-

ness encouraging lifelong learning is vital and on-

the-job training an important tool for keeping the

workforce qualified for the labour market. Though,

unlike with formal school, vocational, and university

education, the government only has limited means

to influence on-the-job training since the market is

managed primarily by private agents. In fact, in

2012, 67% of adult training investments in

Germany were funded by firms (Autorengruppe

Bildungsberichterstattung 2014), making firms the

prime training sponsor and research on their moti-

vation to provide training a timely issue.

Contributing to the recent literature that exploresdifferent factors that determine training incidence,

we ask: how does local employer competition influ-

ence individual training incidence?

Constituting the traditional view of human capital

theory on training, Becker (1993) argues that

employees reap the benefits of general training and

consequently should fund it themselves. In recent

years, the ‘New Training Literature’ has challenged

this traditional view (see, among others, Acemoglu

and Pischke [1998, 1999] and Bassanini et al.

[2007]). For example, Acemoglu and Pischke(1999) suggest that in monopsonistic labour markets

employers will not only pay for firm-specific training

but also pay for general training. They argue that in

imperfect labour markets there is a rent because

firms can pay wages that are lower than the marginal

product of labour.1 This circumstance enables firms

to benefit at least to some extent from general train-

ing since they may set the wage structure such that

they recover the costs of training investments and

hence they are inclined to finance at least a part of it.

There are different causes that bring about rents foremployers or monopsony power. While the classical

justification for monopsony power is based on the

presence of few employers, more recent literature

CONTACT Sylvi Rzepka [email protected] RWI, and RUB, RWI, Hohenzollernstr. 1-3, Essen 45128, Germany

This article uses data from the National Educational Panel Study (NEPS): Starting Cohort 6 – Adults (Adult Education and Lifelong Learning), doi:10.5157/NEPS:SC6:1.0.0. The NEPS data collection is part of the Framework Programme for the Promotion of Empirical Educational Research, funded by the GermanFederal Ministry of Education and Research and supported by the Federal States.1Analysing the returns to training, Konings and Vanormelingen (2015) provide evidence that productivity increases more than wages. This would be in line

with employers having monopsony power.

APPLIED ECONOMICS, 2016

http://dx.doi.org/10.1080/00036846.2015.1137550

© 2016 Taylor & Francis

http://www.tandfonline.com/


3/16

emphasizes the importance of market frictions.

Bhaskar, Manning and To (2002) and Manning

(2011) provide examples of such frictions: imperfect

information of workers about job opportunities, the

ability of firms to differentiate themselves from

others in the sector combined with preference het-

erogeneity of workers over jobs, searching costs and

mobility costs. Another reason for rents might becollusion between employers. Furthermore, if skills

are not perfectly transferable or if the mix of specific

human capital that a worker possesses makes it

difficult to find an employer who requires exactly

this mix (Lazear 2009), employers might also set

wages in a way that enables them to reap the benefits

of training.

Generally it is difficult to empirically determine

the degree of monopsonistic or oligopolistic power

directly or indirectly since measures for wage com-

pression, mobility constraints, or poaching risk are

unavailable. Therefore, the literature discusses sev-

eral proxies (e.g. Boal and Ransom 1997). For exam-

ple, the number of (same sector) firms in a region

and the Herfindahl index are measures that approx-

imate the classical monopsony power of a firm.

These measures relate to monopsony models of clas-

sic differentiation and of moving costs, since they

approximate the physical distance between firms (cf.

Boal and Ransom [1997] and Manning [2003]). The

rationale behind both monopsony measures is that

employers in areas with more competition, i.e. with

more firms in their sector or with less concentration,

have less monopsony power over their employees

since there are many job alternatives. In such a set-

ting, firms need to pay wages at or very close to the

marginal product to prevent employees from moving

to the firm next door. While the underlying rationale

behind the number of firms and the Herfindahl

index is the same, the two measure slightly different

aspects: the number of firms captures the density of a

sector and the Herfindahl index additionally takesthe firm size into account. Building upon the litera-

ture, we hypothesize that in regions with higher local

competition for labour, the monopsonistic influence

of each firm decreases and therefore, the incidence

and volume of firm-financed training will be lower.

This is our primary hypothesis.

Recent empirical studies on European countries

investigate this hypothesis from various angles.2

Given a regional industrial structure, Muehlemann

and Wolter (2007) examine how a firm’s decision to

train apprentices is affected by the likelihood of

trained workers quitting after the completion of

vocational training. They detect a negative impact

of the number of same-sector firms within a regionon the provision of training. Muehlemann and

Wolter (2011) explore the effect of regional firm

density on the extensive and intensive margin of

apprentice demand. They find that only firms that

bear net costs of training are negatively affected by

the number of same-sector firms in their region; the

others face a near to zero demand elasticity for

apprentices with respect to regional firm density.

Brunello and Gambarotto (2007) and Brunello

and de Paola (2008) study how and through what

channels regional economic density, measured as the

number of employees per square kilometre, affects

the training incidence of the individual. The authors

highlight that economic density has two opposite

effects on a firm’s motivation to finance training.

On the one hand, economically dense labour mar-

kets provide incentives to train since this makes it

easier to use positive knowledge spillovers present in

the region; on the other, high economic density also

discourages training due to the proximity of compe-

titors that can easily poach highly qualified staff.

Their empirical results suggest that the latter effect

dominates, since for both the UK and Italy they find

an overall negative sign for this correlation (Brunello

and Gambarotto 2007; Brunello and de Paola 2008).

In this article, we build on this literature and

investigate how the density and the concentration of

same-sector firms in a region affects training inci-

dence and training duration of employees to test the

hypothesis that individual training participation

decreases with the degree of local employer competi-

tion. In contrast to Muehlemann and Wolter (2007,2011), we focus on individual on-the-job training that

takes place after the initial education or vocational

training. This means we investigate on-the-job skill

formation from the individual perspective, rather

than preparation for the job from the firm’s perspec-

tive. Furthermore, while many articles in this strand

2A related strand of literature analyses the impact of product market competition on training (e.g. Bassanini and Brunello 2011; Görlitz and Stiebale 2011;Picchio and Van Ours 2011; Bilanakos et al. 2014).

2 S. RZEPKA AND M. TAMM


4/16

of literature use administrative geographic entities, we

rely on local labour markets defined by commuter

flows, which we find describes the local labour market

more realistically. The literature so far only considers

training participation; however, as Green et al. (2013)

point out, sheer participation may be a misleading

measure when one is ultimately interested in skill

formation. Hence, we include a measure of trainingduration into our analysis as a proxy for training

intensity. Motivated by the monopsony literature,

we use the Herfindahl index as a measure of local

labour market competition in addition to the number

of same-sector firms. Furthermore, most of literature

does not differentiate between sector-specific mea-

sures of local firm competition; we, however, rely on

variation of firm density and concentration across

regions and sectors, which can approximate local

employer competition more precisely. This also

enables us to control for sector and region fixed

effects in the empirical analysis. These fixed effects

pick up many aspects that might bias results if firms

or workers select into regions based on considerations

that correlate with competition or future skill forma-

tion. Moreover, we contribute to the literature by

studying heterogeneous effects by employee charac-

teristics that are of high policy relevance such as

gender, levels of education, migration background,

age and tenure, as well as by regional and sector

aspects.

The remainder of the article is structured as fol-

lows: Section II describes the data and research

methods. Section III presents our results and robust-

ness checks. Section IV examines heterogeneous

effects and Section V concludes.

II. Data and research methods

Data

We rely on two sources of data. First, we use the

2009/10 wave of the adult cohort of the NationalEducation Panel Survey (NEPS, Start Cohort 6).

NEPS is a panel study on educational, occupational

and family formation processes (Blossfeld, Roßbach,

and Von Maurice 2011). The adult cohort covers

detailed life course information from birth to adult

life for more than 11,000 individuals born between

1944 and 1986. For the analysis, we focus on indivi-

duals working as employees at the time of interview,

i.e. we exclude self-employed, unemployed and retir-

ees among others. Furthermore, we exclude from the

sample those employees that participate in publicprograms such as job creation schemes (e.g.

‘Arbeitsbeschaffungsmaßnahmen’ or ‘1-Euro-Jobs’),

or are in apprenticeship. The latter we exclude from

the estimation sample because the apprenticeship sys-

tem is a specialty of the German labour market that

provides upfront vocational training; hence an

apprenticeship is more of an educational than an

employment spell. In contrast, our research question

focuses on on-the-job training in regular jobs, which

is more similar to training in other countries.

Employees from the public sector and civil servants

are excluded as well since the public sector does not

necessarily follow a profit maximizing strategy when

deciding about investments. This leaves us with a

final, cross-sectional sample of almost 5,000 indivi-

duals employed in the private sector.3

We study two dependent variables. First, training

participation, a binary variable, which indicates

whether an individual has participated in training

with a professional interest at least once during the

previous 12 months. This variable captures any for-

mal and non-formal training like seminars or train-

ing courses that occurred while the individual was

employed and provides a measure of training inci-

dence. Around 33% of individuals in our sample

participated in at least one such training during the

last 12 months (see Table 1). For a subsample of

courses we know who financed the training and find

that 91% of individuals with training participated in

employer-financed courses,4 which is in line with

stylized facts presented in Bassanini et al. (2007).

This implies that our outcome variable is a goodapproximation of incidence of firm-financed train-

ing. Our second dependent variable is total training

duration in hours. It accounts for the duration of all

3Out of the entire sample (11,649 individuals), we exclude 2,518 individuals because they are doing an apprenticeship or are not working at the time of interview, 1,139 individuals who are self-employed, 33 individuals who are in a job creation scheme, 2,398 individuals who are civil servants or employedin the public sector, 488 individuals for whom information on region or sector are missing and 144 individuals for whom information on trainingparticipation is missing.

4For cost reasons, the NEPS interview covers details about training courses only for up to two courses per person. These are selected randomly out of thetotal number of courses. Thus, for individuals attending 3 or more training courses during the previous 12 months, information on financing is availableonly for a subsample of courses.

A PPLI ED ECONOM ICS 3


5/16

courses taken during the last 12 months. Training

duration, for those that participated in training, has

an average of 45 hours (Table 1).

Table 1 also provides summary statistics on back-

ground characteristics of the individuals in our sam-ple. Individuals are on average 44 years old and have

almost 10 years of tenure. Forty-four percent are

women, and 13% have children below the age of 6

living in the household. One out of nine individuals

has a migration background, i.e. was not born in

Germany. The majority of individuals are well edu-

cated: two-thirds completed vocational education

and another 16% has a university or college degree.

Almost 3 out of 10 individuals are working part-time

and more than 1 out of 6 has a fixed-term contract.Second, we use the Establishment-History-Panel

(BHP) from the Institute of Employment Research

(IAB) for information on the number of firms that

have at least one employee covered by the social

security system for each region and sector.5 With

Table 1. Descriptive statistics.

Mean SD Min Max

At least one course with professional interest 0.3276 0.4694 0 1Training duration in hours (conditional on training participation) 45.196 78.0537 1 999No. of firms in same sector and local labour market per km2 1.0871 1.1348 0 9Herfindahl index (concentration of employees) 0.0214 0.0480 0 1Female 0.4383 0.4962 0 1Born abroad 0.1170 0.3214 0 1Age 44.1367 10.3568 23 67No vocational education 0.1326 0.3392 0 1

Vocational education 0.6932 0.4612 0 1University degree 0.1635 0.3698 0 1Missing information on education 0.0107 0.1028 0 1Household has child(ren) below 6 years 0.1271 0.3331 0 1Job tenure in years 9.8584 9.4850 0 45Part-time worker 0.2818 0.4499 0 1Missings in part-time dummy 0.0005 0.0231 0 1Fixed-term contract 0.1736 0.3788 0 1Missings in fixed-term dummy 0.0201 0.1404 0 1Firm with between 1 and 10 employees 0.1975 0.3981 0 1Firm with between 10 and 50 employees 0.3206 0.4667 0 1Firm with between 50 and 250 employees 0.3709 0.4831 0 1Firm with more than 250 employees 0.0941 0.2920 0 1Firm size missing 0.0169 0.1290 0 1Working in local labour market different from region of birth 0.6912 0.4620 0 1Occupations

Managers 0.0424 0.2015 0 1

Professionals 0.1560 0.3629 0 1Technicians and associate professionals 0.1829 0.3866 0 1Clerical support workers 0.1504 0.3575 0 1Service and sales workers 0.1276 0.3336 0 1Agricultural, forestry and fishery workers 0.0114 0.1063 0 1Craft and related trades workers 0.1834 0.3870 0 1Plant and machine operators, and assemblers 0.0610 0.2393 0 1Elementary occupations 0.0811 0.2731 0 1

SectorsAgriculture (A) 0.0136 0.1158 0 1Electricity, water, waste, mining (B, D, E) 0.0169 0.1290 0 1Manufacturing (C) 0.3589 0.4797 0 1Construction (F) 0.0719 0.2583 0 1Sale and retail trade (G) 0.1316 0.3381 0 1Transporting and storage (H) 0.0423 0.2014 0 1Accommodation and food service activities (I) 0.0234 0.1513 0 1Information and communication (J) 0.0479 0.2136 0 1

Financial and insurance activities (K) 0.0444 0.2060 0 1Profess., technical, admin. and support service activities, real estate (L, M, N) 0.0912 0.2879 0 1Education (P) 0.0175 0.1312 0 1Human health and social work (Q) 0.0854 0.2796 0 1Arts, entertainment, recreation, act. of households, other services (R, S, T) 0.0550 0.2279 0 1

Observations 4929

NEPS Starting Cohort 6. Sample restricted to private sector employees. Training duration is only indicated for those that participated in training during thelast 12 months (N = 1518).

5This data set is based on the 100% sample of the BHP and was developed in a cooperation project between the RWI and the Research Data Center of theGerman Federal Employment Agency (BA) at the IAB. This specific aggregated data set is available for replication studies upon request by contacting theauthors and the Research Data Centre ([email protected]).



6/16

these data we construct our two versions of the

main explanatory variable. Sectors are categorized

by NACE6 and partly aggregated. Overall, we dis-

tinguish between 14 sectors (cf. Table 1). Regions

are defined by local labour markets as suggested by

Kosfeld and Werner (2012). The advantage of

using these local labour markets is that they do

not exclusively rely on administrative borders;rather these areas are characterized by close com-

muter links, which ensure high seclusion towards

other local labour markets. Overall, Kosfeld and

Werner (2012) distinguish between 141 local

labour markets. For the individuals in our sample,

these local labour markets are the region in which

most job changes take place; hence it is the rele-

vant region for employment competition.7 We

merge this information to the NEPS data based

on region of work place and sector of current

employment of individuals. Our first measure for

local employer competition is the number of firms

per sector and local labour market. Following

Muehlemann and Wolter (2011), Brunello and de

Poala (2008) and Brunello and Gambarotto (2007)

we standardize the number of firms per sector and

region by dividing by the size of the local labour

market in square kilometres.8 This provides an

intuitive density measure. As shown in Table 1,

the average number of same-sector firms in the

region is 1.1 per square kilometre in our NEPS

sample. The Herfindahl index, our second measure

for local employer competition, is calculated based

on the number of employees per firm f in local

labour market j and sector k:

Herfindahl Index jk ¼X14k¼1

employees fjk

employees jk

2

Since the Herfindahl index also accounts for firm

size, it describes whether the local labour market of a

sector is dominated by small or large firms, which isan important aspect of local employer competition

that the simple measure of firm density does not

capture. Taken together, the two measures depict

the density and the concentration of a sector in the

local labour market. The Herfindahl has a mean of

0.0214 across all sectors and local labour markets.

Combining the two sources of data, we find that

rates of participation in professional training vary

considerably between regions. As documented in

Figure 1, participation rates vary from 7% to 56%

across local labour markets. These differences are

only partly due to differences in sector affiliation.

For example, when restricting the sample to only

individuals in the manufacturing sector, we still

find that training participation rates between local

labour markets (with more than 10 interviewees)

0.0714 - 0.2609

0.2609 - 0.3158

0.3158 - 0.3630

0.3630 - 0.4368

0.4368 - 0.5556

Less than 10 observations

Figure 1. Training incidence by local labour market.

This graph displays the average individual training incidence foreach of the 141 local labour markets. Only for this graph we

have restricted the sample of local labour markets to those thathave more than 10 interviewees. Source: NEPS Starting Cohort 6.

6NACE is the industry standard classification system of the European Union. It is similar to the International Standard Industrial Classification (ISIC).7Combes and Duranton (2006) show that in France worker flows take place locally and mostly in the same sector. Examining the employment trajectories

covered in NEPS, we also find that approximately 72% of all job-to-job changes of workers take place between employers located in the same local labourmarket and only a small share of job-to-job changes take place across the borders of local labour markets. This suggests that workers are not very mobilebetween regions and that most competition for workers takes place within these regions.

8For this we use the data provided by BBSR (2011).



7/16


8/16

Table 2. Main estimation results.

Training participation Training duration

(1) (2) (3) (4)

No. of firms in same sector andlocal labour market per km2

−0.0234** 1.1550(0.0117) (2.2369)

Herfindahl index 0.1083 62.7097***(0.1549) (22.1566)

Female −0.0004 0.0002 1.3662 1.2218(0.0177) (0.0177) (3.4676) (3.4656)

Born abroad −0.0299 −0.0315 −3.8759 −3.8858(0.0261) (0.0262) (5.1185) (5.1193)

Age 0.0177*** 0.0177*** −1.0550 −1.0720(0.0061) (0.0061) (1.0500) (1.0490)

Age squared −0.0002*** −0.0002*** 0.0113 0.0115(0.0001) (0.0001) (0.0118) (0.0118)

Vocational education 0.0362 0.0364 1.8674 1.8871(0.0226) (0.0227) (4.3603) (4.3563)

University degree 0.0723** 0.0720** −4.8474 −5.2387(0.0290) (0.0290) (4.8340) (4.8503)

Missing information oneducation

−0.0228 −0.0247 −7.0481 −7.5291(0.0566) (0.0569) (21.8009) (21.9822)

Household has child(ren) below6 years

−0.0413* −0.0418* 9.7232* 9.6363*(0.0233) (0.0234) (5.1625) (5.1698)

Job tenure in years 0.0035* 0.0035* −0.2448 −0.2463(0.0021) (0.0021) (0.4307) (0.4294)

Tenure in years squared −0.0001* −0.0001* 0.0034 0.0036

(0.0001) (0.0001) (0.0126) (0.0125)Part-time worker −0.0650*** −0.0648*** −13.3996*** −13.3489***(0.0190) (0.0190) (3.8695) (3.8666)

Missings in part-time dummy −0.6131*** −0.6177*** – –(0.0736) (0.0735)

Fixed-term contract −0.0126 −0.0117 −0.0578 0.2843(0.0206) (0.0207) (4.3045) (4.2788)

Missings in fixed-term dummy −0.1202*** −0.1178*** 12.0186 11.7061(0.0383) (0.0381) (16.8652) (16.8344)

Firm with between 10 and 50employees

0.0384* 0.0377* 2.0780 1.9971(0.0211) (0.0211) (4.4789) (4.4708)

Firm with between 50 and 250employees

0.0771*** 0.0759*** 10.3412** 10.4085**(0.0238) (0.0239) (4.2906) (4.2730)

Firm with more than 250employees

0.1291*** 0.1287*** 20.1981*** 20.2739***(0.0305) (0.0305) (5.0492) (5.0263)

Firm size missing −0.0665 −0.0681 16.8526 17.5747(0.0458) (0.0456) (18.4008) (18.3608)

Managers 0.1065** 0.1060** 3.8479 4.2404(0.0433) (0.0434) (5.2829) (5.2861)Professionals 0.0495 0.0494 4.9282 5.3964

(0.0316) (0.0316) (3.5029) (3.5150)Clerical support workers −0.0827*** −0.0833*** −10.6777** −10.0855**

(0.0297) (0.0297) (5.0618) (5.0555)Service and sales workers −0.1130*** −0.1138*** 3.4126 4.1680

(0.0279) (0.0280) (6.1932) (6.1537)Agricultural, forestry and fishery

workers−0.1324* −0.1279* 2.2466 2.5503(0.0705) (0.0703) (12.3816) (12.4061)

Craft and related trades workers −0.1915*** −0.1913*** −6.3788 −6.2036(0.0316) (0.0317) (4.5566) (4.5645)

Plant and machine operators,and assemblers

−0.1997*** −0.1999*** −28.4452*** −28.2840***(0.0372) (0.0372) (7.0406) (7.0083)

Elementary occupations −0.2953*** −0.2961*** 11.2451 12.5487(0.0276) (0.0277) (14.0906) (14.1124)

Occupation missing 0.0438 0.0419 29.4602* 29.8104*

(0.1307) (0.1301) (16.8183) (16.5362)R2 0.1540 0.1535Number of clusters 1047 1047 607 607Observations 4929 4929 1518 1518

Table shows the regression results for four basic specifications including all control variables. The specifications additionally control for regional and sectorfixed effects. No vocational education is the reference group for the education dummies, technicians and associate professionals for occupations and firmsup to 10 employees for firm size. Columns 1 and 2 display the estimation results of the linear probability model for training participation. Columns 3 and 4show the marginal effects of a zero-truncated negative binomial model for training duration. Standard errors accounting for clustering at the sector-regionlevel are in parentheses; * p < 0.10, ** p < 0.05, *** p < 0.01.



9/16

by 2.3 percentage points. The Herfindahl Index

(Column 2) does not reach conventional levels of

statistical significance; however, the direction of the

coefficient is as expected.

Having shown the results for the participation

decision, we now turn to the results of the zero-

truncated negative binomial model for training

duration. They are displayed in Columns 3 and 4of Table 2. Many of the control variables have a

similar impact on training duration as on participa-

tion. For instance, training duration is much longer

in larger firms: those that participate in training

complete about 10 hours and 20 hours more of

training, in firms of 50–250 employees and firms

with more than 250 employees compared to firms

with up to 10 employees, holding everything else

constant. The effect of part-time work also has a

strong diminishing effect on training volume.

The coefficient for the Herfindahl index indicates

that as concentration increases, i.e. sector competi-

tion in the local labour market decreases, training

volume also rises (see Column 4 of Table 2). All else

equal, an increase of the Herfindahl index by one

standard deviation leads to a decrease of average

training duration by around 3 hours

(=0.048 × 62.7097). This is in line with the theore-

tical underpinnings of monopsony power influen-

cing training. In the specifications for training

duration, only the Herfindahl index reaches conven-

tional levels of significance; whereas firm density is

insignificant. As mentioned in Section II both indi-

cators capture slightly different aspects of local

employer competition; hence, it is plausible that

results differ somewhat. Yet, the overall picture is

similar and the analysis so far suggests that firm

density negatively affects training participation,

while lower levels of firm concentration decrease

training volume. Both results are in line with the

argument that local employer competition influences

training decisions.

Robustness checks

We conduct a number of robustness checks for all

four main specifications to explore whether con-

founders or potential sources of endogeneity hamper

our results. We also test several aspects of

estimation. First, there could be additional factors

at the sector-by-local labour market level that corre-

late with our employment competition measures and

the training variables, which might bias our results.

For instance, some sectors in a particular local

labour market may have many part-time workers

and therefore train less. Such circumstances could

bring about a similar effect as we present above butwithout being related to the sectoral density and

competition. This is why we explicitly control for a

set of potential confounding factors as a robustness

check. These factors are the skill composition among

the labour force, the female share of the labour force,

the migrant share and the part-time employment

share, each of them measured at the sector-by-local

labour market level (see Table 3).12 While the point

estimate of the number of firms per sector and local

labour market increases slightly in the participation

equation (Column 1), all of the sector-local labour

market characteristics are insignificant and hence do

not impact our estimation results. Similarly, the

relationship between training participation and the

Herfindahl index does not change fundamentally

when controlling for these additional characteristics

(Column 2). This is why our main specification only

includes sector and local labour market fixed effects

and no further observable sector-local labour market

factors since they appear unrelated to individual

training incidence. For training duration, some sec-

tor-local labour market factors do have an influence

(Columns 3 and 4 in Table 3). For example, a high

share of high-skilled employees in one sector

decreases overall training duration; whereas a high

share of migrants in a sector and local labour market

increases training duration. Both factors may be

interpreted using stylized facts: high-skilled employ-

ees may need fewer hours of training to keep their

skill set up-to-date, while sector-local labour market

combinations with many migrants require longer

training durations, e.g. for upgrading German lan-guage skills. Most importantly, however, controlling

for these additional sector-local labour market fac-

tors hardly changes the point estimates of the

Herfindahl index or the number of firms per sector

suggesting that these are not confounding factors.

Note that there might be other sector-by-local

labour market specific factors than those controlled

12These sector-by-local labour market characteristics are all calculated based on the Establishment-History-Panel from the IAB.



10/16

for in Table 3. For example, there might be policy

programs specifically designed for underdeveloped

sectors in specific regions that also influence train-

ing. If this is the case, our estimates are biased and

the direction of the bias depends on whether it is

sectors with above or below competition receiving

policy support. While we are not aware of such

programs, we cannot rule out such confounders,

which would not be captured by the sector and

region fixed effects included in the model.

Second, our measures of firm density and con-

centration may be endogenous if individuals select

into regions and sectors due to unobservables, for

instance, their willingness to participate in

training.13 While we think that most of such selec-

tive sorting is already accounted for by the indivi-

dual-specific control variables and the sector and

region fixed effects of the main specifications,

Table 4 shows estimates that directly check whether

additional factors drive the selection. Specifically,

we test for the impact of selection into local labour

markets by including a variable that indicates

whether individuals are currently working in a

local labour market different from the one they

were born in. This indicator variable could pick

up any additional variation related to both the

selection into a local labour market and the training

incidence of an individual since especially movers

might self-select into specific regions. However, the

coefficients in Columns 1 and 2 of Table 4 suggest

that this selection has no significant influence on

the training incidence and that selection into local

labour markets does not challenge our estimation.

For training volume, in contrast, selection appears

to have a (marginally significant) negative influence(Columns 3 and 4). Yet, this hardly changes the

point estimates of the competition measures.

Therefore, we conclude that selection into local

labour markets does not drive our main results.

Third, we explore whether the impact of our main

explanatory variables is non-linear by estimating

specifications that additionally include squared

Table 3. Robustness check – controlling for economic structure at sector-by-region level.

Training participation Training duration

(1) (2) (3) (4)

No. of firms in same sector and locallabour market per km2

−0.0256** 0.7267(0.0117) (2.2343)

Herfindahl index 0.0720 66.5411**(0.1792) (26.9071)

Share of low-skilled in sector andlocal labour market

0.0269 0.0425 −34.1941 −6.0905(0.2733) (0.2808) (38.6978) (39.9621)

Share of high-skilled in sector andlocal labour market

0.3594 0.3623 −121.4839*** −119.7780***(0.2615) (0.2630) (42.9412) (43.1090)

Share with missing educational infoin sector and local labour market

0.0669 0.0405 −34.7285 −1.7853(0.1850) (0.1935) (34.6492) (39.6307)

Share of female employees in sectorand local labour market

−0.0165 −0.0019 −61.3481 −49.5862(0.2439) (0.2442) (45.5991) (45.5068)

Share of migrants in sector and locallabour market

0.0522 −0.0019 144.8536* 119.7827(0.3591) (0.3611) (74.5627) (76.0016)

Share of part-time employees insector and local labour market

−0.0145 0.0266 −16.3458 −26.4928(0.2191) (0.2138) (36.9658) (37.6286)

Structural local labour marketvariable censored

0.0572 0.0442 −3.7205 −4.5366(0.0416) (0.0415) (9.0673) (8.9813)

R2 0.1547 0.1540Number of clusters 1047 1047 607 607Observations 4929 4929 1518 1518

See notes of Table 2. Specifications additionally control for several sector-local labour market characteristics.

Table 4. Robustness check – controlling for mobility acrossregions.

Trainingparticipation Training duration

(1) (2) (3) (4)

No. of firms in samesector and local labourmarket per km2

−0.0234** 1.3305(0.0117) (2.2516)

Herfindahl index 0.1090 59.5057***(0.1552) (22.7656)

Working in local labourmarket different fromregion of birth

0.0039 0 .0039 −5.5027* −5.1592*(0.0160) (0.0160) (3.1269) (3.1152)

R2 0.1541 0 .1535Number of clusters 1047 1047 607 607Observations 4929 4929 1518 1518

See notes of Table 2. Specifications additionally control for mobilityindicator.

13Brunello and de Paola (2008) also discuss this type of endogeneity.



11/16

terms (Table 5). For all our four main specifications,

results suggest no improvement over the linear spe-

cifications since the coefficients of the squared terms

are insignificant.Furthermore, we explore different methods of

clustering. All results presented so far control for

(one-way) clustering at the region-sector level;

results in Columns 2 and 3 of Table 6 use alternative

clusters that account for clustering at the regional

level and for clustering at the sector level.14 In all

specifications, significance levels hardly alter and do

not change our qualitative interpretations. We

generally prefer the sector-local labour market clus-

tering, since here we do not encounter a small-sam-

ple bias. In addition, these standard errors are often

the most conservative with which we avoid over-

rejecting the null hypothesis.

IV. Heterogeneity analysisAs outlined in the introduction, theory suggests that

firms consider the chances that their human capital

investment pays off when deciding on training

investments. The likelihood of an investment being

profitable depends on how wages are set and on the

poaching risk; both factors are interrelated and likely

to vary across socio-economic groups. For one thing,

the wage setting power that firms can exert differs by

socio-economic groups. For example, Hirsch and

Jahn (2012) point out that migrants have lower

labour supply elasticities and therefore firms havemore wage setting power over migrants. Therefore,

we would expect that firm density has a different

effect on migrant compared to native workers. For

another, poaching risks vary across different indivi-

duals mainly because of differences in job-to-job

mobility. Among others, the literature on job mobi-

lity highlights differences according to age, tenure,

gender and education (Booth, Francesconi, and

Garcia-Serrano 1999; Theodossiou and Zangelidis

2009): for instance, job mobility increases with edu-

cation and decreases with age. Hence, for our ana-

lysis, we expect that firm density has a stronger effect

on highly educated and/or younger employees. In

addition to these theoretical considerations, looking

into heterogeneities is policy relevant, since, for

instance, the recent Strategic Framework for

European Cooperation in Education and Training

calls for equity in training participation (EU

Council 2009). Therefore, we now examine whether

there are heterogeneities by age, tenure, educational

level, gender and migration background as well as by

local labour market characteristics. For sake of brev-

ity we only discuss results for the impact of number

of firms on training participation and for the impact

of the Herfindahl index on training duration, i.e.

those specifications where results in Section III

were significant. They are shown in Tables 7 and 8;

Table 5. Robustness check – testing a non-linear relationship.

Trainingparticipation Training duration

(1) (2) (3) (4)

No. of firms in same sectorand local labour marketper km2

−0.0012 4.3224(0.0325) (6.7273)

No. of firms in same sectorand local labour marketper km2 squared

−0.0032 −0.4592(0.0046) (0.9427)

Herfindahl index 0.2271 87.6708(0.3965) (83.4011)

Herfindahl index squared −0.2615 −50.0113(0.7449) (140.7619)

R2 0.1541 0.1535Number of clusters 1047 1047 607 607Observations 4929 4929 1518 1518

See notes of Table 2. Specifications additionally control for squared com-petition measure.

Table 6. Robustness check – testing alternative levels of

clustering.Cluster at sector-

local labour marketlevel (main

specification)Cluster at

sector level

Cluster atlocal labourmarket level

(1) (2) (3)

A) Training participationNo. of firms in same

sector and locallabour market perkm2

−0.0234** −0.0234*** −0.0234**(0.0117) (0.0074) (0.0110)

B) Training participationHerfindahl index 0.1083 0.1083 0.1083

(0.1549) (0.1717) (0.1040)

C) Training duration

No. of firms in samesector and locallabour market perkm2

1.1550 1.1550 1.1550(2.2369) (2.3324) (2.3154)

D) Training durationHerfindahl index 62.709 7*** 62.7097 *** 62.7097***

(22.156 6) (21.8 562) (22.2060)

See notes of Table 2. Specifications apply alternative levels of clustering.Column (1) replicates results from Table 2.

14For training participation, we also used two-way clustering by sector and region following Cameron, Gelbach and Miller (2011), which produced standarderrors in the same range as those presented, if not smaller.

10 S. R ZE PK A A ND M . TAM M


12/16

the tables also present the p-values of tests of equal-

ity of the interaction effects. For the other two main

specifications (i.e. those regressing training partici-

pation on the Herfindahl index and training dura-

tion on the number of firms), results often show a

similar pattern but generally insignificant results.15

While the job mobility literature finds quite het-

erogeneous effects by age (e.g. Theodossiou and

Zangelidis 2009), our specification including inter-

actions with age dummies does not reveal that the

local employment competition effect on training

incidence works differently for employees of differ-

ent ages (Panel A of Table 7). All the point estimates

for the interaction between firm density and age

groups are similar in size (Panel A). Likewise, the

results for the regressions of training duration (Panel

A in Table 8) display no clear pattern of

heterogeneity.

Similarly, the interactions of tenure with our main

explanatory variables do not reveal a clear pattern of

heterogeneity: the interactions of the number of

same-sector firms per local labour market with

Table 7. Heterogeneity analysis for training participation and firm density.

Heterogeneity by Interaction coeff icients p-Values of Wald’s tests for equality

of interactions R2Number of

clustersObser-vations

(A) Age Below 40 years × no. of firms −0.0193* Below 40 years = 40–50years interaction

0.9800 0.1532 1047 4929(0.0107)

Between 40 and 55 years × no. of firms

−0.0196* 40–50 years = older than55 years interaction

0.9797(0.0118)

Older than 55 × no. of firms −0.0192 Below 40 years = olderthan 55 years interaction

0.9943(0.0141)

All age interactions 0 .9996

(B) Tenure Tenure up to 2 years × no. of f irms −

0.0255* Up to 2 years = 3–10 yearstenure interaction

0.8446 0.1536 1047 4929(0.0154)

Tenure 3 to 10 years × no. of firms −0.0224 3–10 years = 10 + yearstenure interaction

0.9702(0.0137)

Tenure more than 10 years × no. of firms

−0.0220* Up to 2 years = 10 + yearstenure interaction

0.8064(0.0133)

All tenure interactions 0.9698(C) Education No vocational training × no. of firms −0.0178 No vocational

training = vocationaltraining interaction

0.8594 0.1542 1047 4929(0.0223)

Vocational training × no. of firms −0.0213* Vocationaltraining = universitydegree interaction

0.4282(0.0118)

University degree × no. of firms −0.0324** No vocationaltraining = universitydegree interaction

0.4374(0.0136)

Missings in education × no. of firms −0.0299 All education interactions 0.6189

(0.0329)(D) Gender Female × no. of f irms −0.0194 Male = female interaction 0.3501 0.1542 1047 4929

(0.0126)Male × no. of firms −0.0294**

(0.0126)(E) Region of birth Born abroad × no. of firms −0.0314** Foreign born = Germany

born interaction0.4164 0.1541 1047 4929

(0.0146)Born in Germany × no. of firms −0.0213*

(0.0122)(F) Labour market

sizeSmall labour market × no. of firms −0.0517*** Small = medium local

labour market interaction0.1639 0.1554 1047 4929

(0.0145)Medium labour market × no. of firms −0.0273* Small = large local labour

market interaction0.0014

(0.0151)Large labour market × no. of firms 0.0098 Medium = large local

labour market interaction0.0353

(0.0178)All local labour market sizeinteractions

0.0058

(G) Public service Public service × no. of firms −

0.0149 Public = private sectorinteraction

0.7407 0.1540 1288 7149(0.0192)

Private sector × no. of firms −0.0206*(0.0109)

Panel A shows interactions of the number of firms with age, Panel B with tenure, Panel C with the education level, Panel D with gender, Panel E with bornabroad, Panel F with labour market size and Panel G with the public and private sector. For this latter interaction, we rely on a larger sample that alsoincludes public servants. All covariates from the baseline specification and fixed effects are included as well but not displayed (cf. notes of Table 2).Standard errors in parentheses; * p < 0.10, ** p < .05, *** p < 0.01.

15These are available upon request.

A PPLI ED ECO NOM ICS 11


13/16

tenure dummies are all similar, suggesting that there

is no tenure-related heterogeneity for training parti-

cipation (see Panel B, Table 7). Likewise, for those

participating in training, differences in the impact of

the Herfindahl index on training duration across

tenure groups are not statistically significant (Panel

B in Table 8).

The interaction terms with education (Panel C in

Tables 7 and 8) suggest that monopsony power

affects individuals with higher educational levels

more. The estimates using the number of same-sec-

tor firms make this most apparent: in areas with

more same-sector firms, training participation is

reduced comparatively more for individuals with

higher educational levels. While we observe a nega-

tive but insignificant effect for employees without

vocational training, we estimate considerably larger

negative and significant effects (at the 5% and 10%

level) for those with vocational education and uni-

versity degree, respectively. This may be because

highly skilled individuals pursue human capital

intensive jobs, and poaching of this type of work-

force can be especially harmful to firms. Therefore,

firms refrain from investing more in high-skilled

employees. The coefficients of the regression of

training duration on the Herfindahl index also

Table 8. Heterogeneity analysis for training duration and Herfindahl index.

Heterogeneity by Interaction coefficients p-Values of Wald’s tests for

equality of interactions Number of clustersObser-vations

(A) Age Below 40 years × Herfindahl 58.4910 Below 40 years = 40–50years interaction

0.9949 607 1518(40.2288)

Between 40 and 55 years × Herfindahl 58.7869*** 40–50 years = older than55 years interaction

0.5932(22.4002)

Older than 55 years × Herfindahl −14.9071 Below 40 years = olderthan 55 years interaction

0.6070(138.0074)

All age interactions 0.8667

(B) Tenure Tenure up to 2 years × Herfindahl 28.2836 Up to 2 years = 3–10 yearstenure interaction

0.7130 607 1518(46.5356)

Tenure 3–10 years × Herfindahl 61.6307 3–10 years = 10 + yearstenure interaction

0.7781(57.5505)

Tenure more than 10 years × Herfindahl 80.2937*** Up to 2 years = 10 + yearstenure interaction

0.1595(26.8904)

All tenure interactions 0.1931(C) Education No vocational training × Herfindahl 2 .5545 No vocational

training = vocationaltraining interaction

0.7649 607 1518(199.3495)

Vocational training × Herfindahl 62.2509** Vocationaltraining = universitydegree interaction

0.9067(24.5298)

University degree × Herfindahl 66.0910** No vocationaltraining = universitydegree interaction

0.7613(31.8836)

Missings in education × Herfindahl 961.4072 All education interactions 0.9549

(801.3697)(D) Gender Female × Herfindahl 38.1292 Male = female interaction 0.5106 607 1518

(45.7976)Male × Herfindahl 67.0593***

(22.1353)(E) Region of birth Born abroad × Herfindahl 64.4781** Foreign born = Germany

born interaction0.9430 607 1518

(30.1040)Born in Germany × Herfindahl 62.4678***

(23.0029)(F) Labour market size Small labour market × Herfindahl 71.9821 Small = medium local

labour market interaction0.9087 607 1518

(115.3092)Medium labour market × Herfindahl 58.6541*** Small = large local labour

market interaction0.8131

(20.8546)Large labour market × Herfindahl 109.9916 Medium = large local

labour market interaction0.6778

(121.6000)All local labour market sizeinteractions

0.9149

(G) Public service Public service × Herfindahl 46.7905 Public = private sectorinteraction

0.6776 853 2547(47.2600)

Private sector × Herfindahl 70.2679**(29.9541)

Panel A shows interactions of the Herfindahl index with age, Panel B with tenure, Panel C with the education level, Panel D with gender, Panel E with bornabroad, Panel F with labour market size and Panel G with the public and private sector. For this latter interaction, we rely on a larger sample that alsoincludes public servants. All covariates from the baseline specification and fixed effects are included as well but not displayed (cf. notes of Table 2).Standard errors in parentheses; * p < 0.10, ** p < .05, *** p < 0.01.



14/16

support this trend (Table 8). Yet, formal tests of

differences between the interaction terms are gener-

ally insignificant (see p-values of the Wald’s tests in

Tables 7 and 8).

Examining gender differences, Panel D of Tables

7 and 8 provides evidence that the effect of firm

density and firm concentration is more important

for men than for women. While for women oneadditional same-sector firm per square kilometre

reduces training participation by about 1.9 percen-

tage points (not significant), it reduces training par-

ticipation for men by about 2.9 percentage points

(significant at 5% level). We observe the same pat-

tern in the specification regressing training duration

on the Herfindahl index: one standard deviation

increase in the Herfindahl index raises the training

volume of women by 0.8 hours and that of men by

1.4 hours. This might be explained by the fact that

women tend to be less mobile considering job-to-job

transitions (Theodossiou and Zangelidis 2009);

hence, poaching risks would generally be lower for

women. However, as for the differences by educa-

tion, the gender-specific results only suggest a ten-

dency since the differences between the interaction

terms are not statistically significant in any of the

specifications.

Furthermore, we analyse whether firm density

has a different effect on migrants’ training inci-

dence and duration (Panel E, Tables 7 and 8).

This analysis is motivated by the fact that migrants

have a special standing in the labour market (Borjas

1999). Our results also hint at a distinct effect for

migrants since the point estimates for migrants are

slightly larger than those for natives in the regres-

sion of training incidence on the number of same-

sector firms. For migrants, an increase in firm den-

sity is associated with a stronger decline (3.1 per-

centage points) than for natives (2.1 percentage

points). However, these differences are not statisti-

cally significant and they are almost identical in theregression of training volume on the Herfindahl

index.

We now analyse whether the impact of firm den-

sity and firm concentration on training participation

and volume differs by size of the area of the labour

market. To do so, we divide labour markets into

three size categories, each covering about a third of

the distribution of all local labour markets. Results

are shown in Panel F in Tables 7 and 8. In the

specification regressing training incidence on firm

density, we detect a clear pattern: the impact of

competition is strongest in small labour markets

and decreases with labour market size. Specifically,

the coefficients for the interaction with small- and

medium-sized local labour markets are negative and

significant, while that of large local labour markets isclose to zero, insignificant and has a positive sign. A

test on equality of the three interaction terms also

confirms that they are significantly different from

one another. This is an intuitive finding, since for

small local labour markets our measure of sector

density might be particularly precise gauging the

actual employment competition. Large labour mar-

kets are likely to have smaller but overlapping sub-

regions of labour market competition, which might

be more crudely captured by the density measure for

the entire labour market. Hence, all above-men-

tioned results may reflect lower bounds of the true

impact of local employment competition on train-

ing. But also note that we do not find any systematic

pattern when regressing training duration on the

Herfindahl index.

Finally, we examine whether similar mechanisms

apply to the public and the private sector. For this

we extend the sample to include public sector

employees and civil servants. Our estimates suggest

that firms in the private and the public sector act

differently (Panel G, Tables 7 and 8). While the

interaction between firm density and private sector

is negative and significant for training participation,

as would be expected from the results above, the

interaction with public sector is about a third lower

and remains insignificant. The same pattern can be

found for training duration using the Herfindahl

Index as main explanatory variable. This finding

that the public sector does not react to local employ-

ment competition as much as the private sector

could be explained by the fact that the public serviceusually does not act to maximize profit as firms in

the private sector do (for related findings see Booth

and Katic 2011).

Summing up, we reveal tendencies for differences

in the effect of firm density and concentration by

educational attainment, gender and migration back-

ground, between private and public sector firms and

by size of the labour market.



15/16

V. Conclusion

In this article we investigate how local competition of

firms for employees, measured by the number and

concentration of same-sector firms within a local

labour market, influences individual training incidence

and training volume. In accordance with what eco-

nomic theory postulates, our empirical findings provide evidence that the poaching risk is an influential

factor since training incidence decreases with same-

sector firm density and training volume increases with

employees being more concentrated at few firms

within a sector. Furthermore, we reveal tendencies

that the effect varies across different educational levels,

by gender and by migration background.

A possible policy conclusion that follows from our

results is that levels of training investments are higher

in regions where the industrial structure is less dense.

Hence, policies that aim at building up more specia-lized economies, while potentially having a positive

impact on innovation and economic growth (e.g.

Romer 1986; Porter 2003), may negatively influence

firms’ investment in training. One possible option to

tackle this potentially negative effect could be to

encourage the use of pay-back clauses for training

investments (Mueller 2014). These would diminish

the risk of poaching, since employees need to reim-

burse training costs in case they voluntarily leave the

firm before the investment is paid off.

The literature also mentions levy schemes such assectoral training funds (Bassanini et al. 2007;

CEDEFOP 2008; Ok and Tergeist 2003; Johanson

2009) that can make poaching less attractive since

all firms of one sector participate in funding train-

ing. However, so far there is no consensus on the

optimal design of such schemes since there are many

trade-offs attached to specific forms.16 Our results

underline the importance of taking regional aspects

such as the local labour market density into account

when deciding on a scheme.

Acknowledgements

We are indebted to Alexandra Schmucker for her coopera-

tion in analysing the BHP data and Paula Schneider for

excellent research assistance. Thanks to Giorgio Brunello,

Colin Cameron, Hanna Frings, Francis Green, Katja

Görlitz, Boris Hirsch, Samuel Mühlemann, Hessel

Oosterbeek and seminar participants at BeNA (Berlin), IAB

workshop on Perspectives on (Un-)Employment

(Nuremberg), NEPS (Bamberg), NIW (Hannover), SMYE

2014 (Vienna), SPP1646 Colloquium (Florence), CINCH

Academy 2014 (Essen), IWAEE 2014 (Catanzaro),

Jahrestagung VfS (Hamburg), BIEN Jahrestagung (Berlin)

and Workshop on Economics of Education (Barcelona) for

helpful comments and suggestions.

Disclosure statement

No potential conflict of interest was reported by the authors.

Funding

The authors acknowledge financial support from the

Deutsche Forschungsgemeinschaft [TA 829/2-1].

ReferencesAcemoglu, D., and J.-S. Pischke. 1998. “Why Do Firms

Train? Theory and Evidence.” The Quarterly Journal of

Economics 113 (1): 79–119. doi:10.1162/003355398555531.

Acemoglu, D., and J.-S. Pischke. 1999. “Beyond Becker:

Training in Imperfect Labour Markets.” The Economic

Journal 109 (453): 112–142. doi:10.1111/ecoj.1999.109.

issue-453.

Autorengruppe Bildungsberichterstattung. 2014. “Nationaler

Bildungsbericht”. Chapter G. Accessed 31 July 2014.

http://www.bildungsbericht.de

Bassanini, A., A. L. Booth, G. Brunello, M. de Paola, and E.

Leuven. 2007. “

Workplace Training in Europe.”

In Educationand Training in Europe, edited by Brunello, G., P. Garibaldi,

and E. Wasmer. Oxford: Oxford University Press.

Bassanini, A., and G. Brunello. 2011. “Barriers to Entry,

Deregulation and Workplace Training: A Theoretical Model

with Evidence from Europe.” European Economic Review 55

(8): 1152–1176. doi:10.1016/j.euroecorev.2011.05.004.

BBSR. 2011. “Raumabgrenzungen: Referenzdateien und

Karten – Kreise und kreisfreie Städte, Kreisregionen. ”

Accessed 7 June 2013. http://www.bbsr.bund.de/BBSR/

D E / R a u m b e o b a c h t u n g / R a u m a b g r e n z u n g e n /

K rei s e_ K rei s regi on en / Dow n l oad _ Ref 11_ x l s .x l s ?

__blob=publicationFile&v=2

Becker, G. S. 1993. Human Capital. A Theoretical and Empirical Analysis, with Special Reference to Education.

3rd ed. Chicago: The Univ. of Chicago Press.

Bhaskar, V., A. Manning, and T. To. 2002. “Oligopsony and

Monopsonistic Competition in Labor Markets.” Journal of

Economic Perspectives 16 (2): 155–174. doi:10.1257/

0895330027300.

Bilanakos, C., J. Heywood, J. Sessions, and N. Theodoropoulos.

2014. W orker Training and Competing on Product Quality .

16See Bassanini et al. (2007) for a discussion of the issues related to specific versions of sectoral training funds.


http://dx.doi.org/10.1162/003355398555531http://dx.doi.org/10.1111/ecoj.1999.109.issue-453http://dx.doi.org/10.1111/ecoj.1999.109.issue-453http://www.bildungsbericht.de/http://dx.doi.org/10.1016/j.euroecorev.2011.05.004http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://dx.doi.org/10.1257/0895330027300http://dx.doi.org/10.1257/0895330027300http://dx.doi.org/10.1257/0895330027300http://dx.doi.org/10.1257/0895330027300http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://www.bbsr.bund.de/BBSR/DE/Raumbeobachtung/%A0Raumabgrenzungen/Kreise_Kreisregionen/Download_Ref11_xls.xls?__blob=publicationFile%26v=2http://dx.doi.org/10.1016/j.euroecorev.2011.05.004http://www.bildungsbericht.de/http://dx.doi.org/10.1111/ecoj.1999.109.issue-453http://dx.doi.org/10.1111/ecoj.1999.109.issue-453http://dx.doi.org/10.1162/003355398555531


16/16

University of Cyprus Working Papers in Economics 08-2014.

Cyprus: Department of Economics, University of Cyprus.

Blossfeld, H.-P., H.-G. Roßbach, and J. Von Maurice, eds.

2011. Zeitschrift für Erziehungswissenschaft: Sonderheft.

Vol. 14, Education as a Lifelong Process. The German

National Educational Panel Study (NEPS). 1st ed.

Wiesbaden: VS Verlag für Sozialwissenschaften.

Boal, W., and M. Ransom. 1997. “Monopsony in the Labor

Market.” Journal of Economic Literature 35 (1): 86–112.Booth, A. L., M. Francesconi, and C. Garcia-Serrano. 1999.

“Job Tenure and Job Mobility in Britain.” ILR Review 53

(1): 43–70. doi:10.1177/001979399905300103.

Booth, A. L., and P. Katic. 2011. “Men at Work in a Land

Down-Under: Testing Some Predictions of Human

Capital Theory.” British Journal of Industrial Relations 49

(1): 1–24. doi:10.1111/bjir.2011.49.issue-1.

Borjas, G. J. 1999. “The Economic Analysis of Immigration.”

In Handbook of Labor Economics, edited by O. Ashenfelter,

R. Layard, and D. Card, 1697–1760. Vol. 3. Amsterdam:

Elsevier.

Brunello, G., and M. de Paola. 2008. “Training and Economic

Density: Some Evidence Form Italian Provinces.” Labour Economics 15 (1): 118–140. doi:10.1016/j.labeco.2006.10.001.

Brunello, G., and F. Gambarotto. 2007. “Do Spatial

Agglomeration and Local Labor Market Competition

Affect Employer-Provided Training? Evidence from the

UK.” Regional Science & Urban Economics 37 (1): 1–21.

doi:10.1016/j.regsciurbeco.2006.06.006.

Cameron, A. C., J. B. Gelbach, and D. L. Miller. 2011.

“Robust Inference with Multiway Clustering.” Journal of

Business & Economic Statistics 29 (2): 238–249.

doi:10.1198/jbes.2010.07136.

Cameron, C. A., and P. K. Trivedi. 2005. Microeconometrics:

Methods and Applications. New York: Cambridge University Press.

CEDEFOP. 2008. Sectoral Training Funds in Europe.

CEDEFOP Panorama Series 156. Luxembourg: Office for

Official Publications of the European Communities.

Combes, P.-P., and G. Duranton. 2006. “Labour Pooling,

Labour Poaching, and Spatial Clustering.” Regional

Science and Urban Economics 36 (1): 1–28. doi:10.1016/j.

regsciurbeco.2005.06.003.

EU Council. 2009. Council Conclusions of 12 May 2009 on a

Strategic Framework for European Cooperation in

Education and Training (ET 2020). Accessed 10 June

2014. http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?

uri=OJ:C:2009:119:0002:0010:en:PDFGörlitz, K., and J. Stiebale. 2011. “The Impact of Product

Market Competition on Employers’ Training Investments.

Evidence from German Establishment Panel Data.” De

Economist 159 (1): 1–23. doi:10.1007/s10645-010-9155-0.

Green, F., A. Felstead, D. Gallie, H. Inanc, and N. Jewson.

2013 “What Has Been Happening to the Training of

British Workers?” Published by the Centre for Learning

and Life Chances in Knowledge Economies and Societies.

Accessed 7 October 2014. http://www.llakes.org

Hirsch, B., and E. Jahn. 2012. Is There Monopsonistic

Discrimination against Immigrants? IZA Discussion

Paper #6472. Bonn: IZA.

Johanson, R. 2009. A Review of National Training Funds. SP

Discussion Paper No. 0922. Washington, DC: World Bank.

Konings, J., and S. Vanormelingen. 2015. “The Impact of

Training on Productivity and Wages: Firm Level

Evidence.” Review of Economics and Statistics 97 (2):

485–497. doi:10.1162/REST_a_00460.Kosfeld, R., and A. Werner. 2012. “Deutsche

Arbeitsmarktregionen - Neuabgrenzung Nach Den

Kreisgebietsreformen 2007 - 2011.” Raum forschung Und

Raumordnung 70 (1): 49–64. doi:10.1007/s13147-011-0137-8.

Lazear, E. 2009. “Firm-Specific Human Capital: A Skill-

Weights Approach.” Journal of Political Economy 117 (5):

914–940. doi:10.1086/598179.

Manning, A. 2003. “The Real Thin Theory: Monopsony in

Modern Labour Markets.” Labour Economics 10 (2): 105–

131. doi:10.1016/S0927-5371(03)00018-6.

Manning, A. 2011. “Imperfect Competition in the Labour

Market.” In Handbook of Labor Economics. Vol. 4, part

B, chapter 11, edited by O. Ashenfelter and D. Card, 973–

1041. Amsterdam: Elsevier.

Muehlemann, S., and S. C. Wolter. 2007. “Regional Effects on

Employer-Provided Training: Evidence from

Apprenticeship Training in Switzerland.” Zeitschrift Für

Arbeitsmarktforschung 2007 (2+3): 135–147.

Muehlemann, S., and S. C. Wolter. 2011. “Firm-Sponsored

Training and Poaching Externalities in Regional Labor

Markets.” Regional Science and Urban Economics 41 (6):

560–570. doi:10.1016/j.regsciurbeco.2011.04.003.

Mueller, N. 2014. “Does CVT of Firms in Germany Suffer from

Poaching?” Empirical Research in Vocational Education and

Training 6 (1): 1–

26. doi:10.1186/1877-6345-6-1.Ok, W., and P. Tergeist. 2003. Improving Workers’ Skills:

Analytical Evidence and the Role of the Social Partners.

OECD Social, employment and migration working papers:

10. Paris: OECD Publishing.

Picchio, M., and J. C. Van Ours. 2011. “Market Imperfections

and Firm-Sponsored Training.” Labour Economics 18:

712–722. doi:10.1016/j.labeco.2011.03.001.

Porter, M. 2003. “The Economic Performance of Regions.”

Regional Studies 37 (6–7): 545–546. doi:10.1080/

0034340032000108688.

Romer, P. M. 1986. “Increasing Returns and Long-Run

Growth.” Journal of Political Economy 94 (5): 1002–1037.

doi:10.1086/jpe.1986.94.issue-5.Statistisches Bundesamt. 2013. Bruttoinlandsprodukt 2012

Für Deutschland. Accessed 24 June 2013. https://www.

destatis.de/DE/PresseService/Presse/Pressekonferenzen/

2 0 1 3 / B I P 2 0 1 2 / P r e s s e b r o s c h u e r e _ B I P 2 0 1 2 . p d f ?

__blob=publicationFile

Theodossiou, I., and A. Zangelidis. 2009. “Should I Stay or

Should I Go? the Effect of Gender, Education and

Unemployment on Labour Market Transitions.” Labour

Economics 16 (5): 566–577. doi:10.1016/j.labeco.2009.01.006.

http://dx.doi.org/10.1177/001979399905300103http://dx.doi.org/10.1111/bjir.2011.49.issue-1http://dx.doi.org/10.1016/j.labeco.2006.10.001http://dx.doi.org/10.1016/j.regsciurbeco.2006.06.006http://dx.doi.org/10.1198/jbes.2010.07136http://dx.doi.org/10.1016/j.regsciurbeco.2005.06.003http://dx.doi.org/10.1016/j.regsciurbeco.2005.06.003http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:C:2009:119:0002:0010:en:PDFhttp://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:C:2009:119:0002:0010:en:PDFhttp://dx.doi.org/10.1007/s10645-010-9155-0http://www.llakes.org/http://dx.doi.org/10.1162/REST_a_00460http://dx.doi.org/10.1007/s13147-011-0137-8http://dx.doi.org/10.1086/598179http://dx.doi.org/10.1016/S0927-5371(03)00018-6http://dx.doi.org/10.1016/j.regsciurbeco.2011.04.003http://dx.doi.org/10.1186/1877-6345-6-1http://dx.doi.org/10.1016/j.labeco.2011.03.001http://dx.doi.org/10.1080/0034340032000108688http://dx.doi.org/10.1080/0034340032000108688http://dx.doi.org/10.1086/jpe.1986.94.issue-5https://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttp://dx.doi.org/10.1016/j.labeco.2009.01.006http://dx.doi.org/10.1016/j.labeco.2009.01.006https://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttps://www.destatis.de/DE/PresseService/Presse/Pressekonferenzen/2013/BIP2012/Pressebroschuere_BIP2012.pdf?__blob=publicationFilehttp://dx.doi.org/10.1086/jpe.1986.94.issue-5http://dx.doi.org/10.1080/0034340032000108688http://dx.doi.org/10.1080/0034340032000108688http://dx.doi.org/10.1016/j.labeco.2011.03.001http://dx.doi.org/10.1186/1877-6345-6-1http://dx.doi.org/10.1016/j.regsciurbeco.2011.04.003http://dx.doi.org/10.1016/S0927-5371(03)00018-6http://dx.doi.org/10.1086/598179http://dx.doi.org/10.1007/s13147-011-0137-8http://dx.doi.org/10.1162/REST_a_00460http://www.llakes.org/http://dx.doi.org/10.1007/s10645-010-9155-0http://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:C:2009:119:0002:0010:en:PDFhttp://eur-lex.europa.eu/LexUriServ/LexUriServ.do?uri=OJ:C:2009:119:0002:0010:en:PDFhttp://dx.doi.org/10.1016/j.regsciurbeco.2005.06.003http://dx.doi.org/10.1016/j.regsciurbeco.2005.06.003http://dx.doi.org/10.1198/jbes.2010.07136http://dx.doi.org/10.1016/j.regsciurbeco.2006.06.006http://dx.doi.org/10.1016/j.labeco.2006.10.001http://dx.doi.org/10.1111/bjir.2011.49.issue-1http://dx.doi.org/10.1177/001979399905300103

firm-specific training and competition

Documents