To recruit skilled workers or to train one's...

To recruit skilled workers or to train one’s own?Vocational training in the face of uncertaintyas to the rate of retention of traineeson completion of training*

Lutz Bellmann and Florian Janik**

The article deals with the impact of uncertainty as regards the retention of traineescompleting vocational training programmes on firms’ training behaviour. As an alter-native to vocational training, the external recruitment of skilled workers is consideredvia an opportunity costs approach. The uncertainty is traced back to staff turnover,considerable product market competition and changing skill requirements. Indicatorsof these causes of uncertainty are examined as to their impact on vocational trainingand the recruitment of skilled workers with a fractional logit model, using a seeminglyunrelated cluster-adjusted sandwich estimator with data from the IAB EstablishmentPanel. There is no empirical evidence that the presumed correlations exist in manufac-turing. A negative correlation is identified in the service sector between high staffturnover coupled with rapid change in the skills required on the one hand, and commit-ment to vocational training on the other, while there is a positive link with the externalrecruitment of skilled workers. This is interpreted as an indication that uncertaintywith regard to retention after training can go some way towards explaining the growinggap between employment and vocational training in the tertiary sector.

* This paper was released for publication in July 2007.** We would like to thank Simon Jansen and two anonymous referees from the Zeitschrift für Arbeits-marktForschung for their helpful remarks.

Contents

1 Introduction

2 Vocational training or external recruitment?

2.1 Recruitment costs: opportunity income of vocational training

2.2 Uncertainty and vocational training

3 Empirical analysis

3.1 3.1 Data

3.2 Empirical strategy and econometric modelling

3.3 Variables and hypotheses

3.4 Results

4 Summary and conclusions

References

Annex

ZAF 2 und 3/2007, S. 205Ð220 205

To recruit skilled workers or to train one’s own? Lutz Bellmann and Florian Janik

1 Introduction

Year after year, many young people fail to obtain avocational training place and as a result frequentlyattend state-funded skill-building schemes. Manifoldreasons have been expressed in the public debatefor this dramatic situation, which has been continu-ing for years despite various promotional pro-grammes and the “Pakt für Ausbildung” (Pact forTraining). Too high training costs, inadequate or-ders, a lack of job prospects after training and insuf-ficiently qualified applicants are frequently referredto in this context. Tertiarisation is commonly identi-fied as a further reason for this development. Firstlybecause the rationalisation processes in manufactur-ing and the concomitant reduction of employmentalso reduce the number of training places, and sec-ondly because the ratio between training places of-fered and the number of employees is deterioratingin the service sector (Baethge et al. 2007).

The stiffer competition faced by firms at both na-tional and international level is indicated as a fur-ther reason for firms providing less vocational train-ing. The results of the IAB Establishment Panel alsosuggest that firms engage in less training becausethey are unsure as to whether they will be able tooffer subsequent employment to the trained work-ers if they enter into a training agreement, becauseof the considerable competitive pressure and the un-certainty with regard to their income situation whichthis entails (cf. Bellmann/Hartung 2005). Anotherreason that is frequently expressed for firms’ reti-cence when it comes to vocational training lies in theorganisation and structure of the German system ofvocational training as such. Where firms are not onlysubject to considerable competitive pressure, but arealso operating in innovative fields in which workerskill requirements are high and subject to rapidchange, one may presume that vocational training,with its firmly established training content over whichthe firms have little influence, is not the right sourceof skilled workers because it is too uncertain whetherthe training contents actually match the current re-quirements at the end of the training period. Forthese reasons, the external recruitment of skilledworkers may be preferred over vocational training.Both arguments entail the hypothesis that uncer-tainty influences firms’ training behaviour and thatincreasing uncertainty has a negative influence on it.

This uncertainty is modelled in this paper using theprobability of a trainee not remaining in the firm oncompletion of his/her training. Firms respond to alow expected retention rate and reduce their com-mitment to vocational training. At the same time,the recruitment of already trained workers to cover

206 ZAF 2 und 3/2007

the need for skills becomes more attractive becauseof falling opportunity costs.

Evidence of this circumstance is provided by Falk(2002), who studied firms employing or seeking ICTspecialists. The enterprises were surveyed in 2000, inother words during the IT boom. The ICT sector isa very good example of the modern, innovative sec-tors of the economy. If these firms need additionalskilled labour, then overtime, outsourcing and theuse of freelancers are preferred to creating trainingplaces. Particularly in technical and entrepreneurialservices, the firms state that they prefer ICT special-ists who have graduated from university over theirown in-house trainees (Falk 2002). This finding isnot without its detractors, however. Steedman et al.(2006) and Steedman/Wagner (2007) reach the con-clusion in a comparative analysis of the UnitedKingdom and Germany that, particularly in the ICTfield, dual training is more attractive for enterprisesin Germany than the recruitment of external skilledworkers, since the trainees in the dual system canbe deployed more flexibly if they are provided withappropriate further training.

Various studies have examined the impact of highresignation rates and retention rates on vocationaltraining. They all reach the conclusion that these re-duce the provision of training by firms (Niederalt2004, Beckmann 2002a, Neubäumer/Bellmann 1999).Smits/Zwick (2004) provide empirical evidence forthe presumption that the matching of the skills pro-vided in vocational training with the actual skill re-quirements poses a greater problem in the servicesector than it does in manufacturing. A number ofstudies (e.g. Wolter et al. 2006, Hartung/Leber 2004,Beckmann 2002a) take into account the impact ofthe share of skilled workers on training provisionand training intensity, but the impact on the externalrecruitment of skilled workers and commitment tovocational training has so far not been the subjectof studies using establishment data.

The text is structured as follows: the second chapterpresents the theoretical model, which is discussedparticularly with regard to the uncertainty of furtheremployment after completion of training. The thirdpart follows with the empirical examination, usingdata taken from the IAB Establishment Panel. Thefourth and final part summarises the results and alsointerprets them in relation to the refinement of thedual system of vocational training.

2 Vocational training or externalrecruitment?

In the model used here, vocational training is re-garded as a part of staff planning reaching beyond

Lutz Bellmann and Florian Janik To recruit skilled workers or to train one’s own?

the training period, with the aim of meeting firms’requirements in terms of skilled workers. As an al-ternative to vocational training, firms can also re-cruit skilled workers via the external labour market.These arguments are based on considerations madeby Franz/Soskice (1995) and taken up by Niederalt(2004), and are used below as a basis for modelling.

2.1 Recruitment costs: opportunityincome of vocational training

It is not worthwhile for firms to invest in generalhuman capital where there is complete competitionon labour markets, since they have no guaranteethat individuals who have received general trainingwill remain in the firm on completion of the skill-building measure. If the trainees leave, the firm canno longer benefit from the increase in productivityas a result of training and bears only its costs(Becker 1964). Investment in general human capitalis only worthwhile for firms if the trainees are paidwages after completion of training which are lowerthan their productivity, and thus a margin can berealised. This type of remuneration is conditional ona compressed wage structure in which, as skills in-crease, wages grow less quickly than productivity(Acemoglu/Pischke 1998, 1999a, 1999b). This meansthat vocational training is only profitable for enter-prises if those completing training can be paid lessthan skilled workers recruited externally. This cannot be described clearly in either theoretical or em-pirical terms (Niederalt 2004). If it is therefore as-sumed for the sake of simplicity that the wages andthe productivity of internally-trained workers arethe same as those of externally-recruited skilledworkers at least after a certain familiarisation phase,the vocational training decision can be modelled viathe concept of opportunity income (Niederalt 2004,Wolter/Schweri 2002, v. Bardeleben et al. 1995).

In this view, the focus is not on productivity beinghigher than remuneration as the source of earningsfrom vocational training, but on meeting the needfor skilled workers. The opportunity income derivedfrom vocational training can be deduced by compar-ing it with the recruitment of skilled workers on theexternal labour market, in other words cost advanta-ges gained by retaining trained workers who havecompleted in-house training (Niederalt 2004). Thisopportunity income consists of:

Ð staff acquisition costs being avoided (Franz/Soskice 1995), since it is no longer necessary tolook for and select skilled workers from the exter-nal labour market;

ZAF 2 und 3/2007 207

Ð familiarisation costs being avoided because thenew employees already know the firm (Kau 1997);

Ð savings being gained by stabilising the wage andsalary structure (Cramer/Müller 1994);

Ð costs being avoided as a result of lower staff turn-over since those completing training remain in thefirm longer (Cramer/Müller 1994) and because ofa reduced risk of recruiting the wrong employees(Franz/Soskice 1995).

Considerations made by Niederalt (2004) are usedin modelling this decision on the part of the firm.The modelling extends to the time when both alter-natives lead to equivalent results, which in this caseapplies after the familiarisation of the externally re-cruited skilled workers in the firm. If it is assumedthat an additional need for skilled workers existsprior to commencement of training for the periods + 1, the transaction costs of familiarisation lead tomore or less fixed costs being incurred for the re-cruitment of external skilled workers. If these trans-action costs, which conversely can also be regardedas opportunity income gained from internal training,are higher than the net costs of training, vocationaltraining is worthwhile for the firm. The followingdecision rule can thus be formulated:

�n

t�s�1

RKt ý �s

t�1

NKt (1)

where RK � recruitment costs for external skilledworkers, and NK � net costs of vocational training.

Therefore, if the recruitment costs, which converselycan also be interpreted as opportunity incomegained from vocational training, exceed the net costsof training, it becomes attractive for the firm to pro-vide apprenticeships. This decision rule is, however,based on the assumption that all people completingtraining remain in the firm. If some or all of thetrainees leave the firm on completion of training,the need for skilled workers must be at least partlymet via the external market. It is therefore only pos-sible to avoid recruitment costs for those traineeswho do remain in the firm. The decision rule musttherefore be supplemented by the retention rate qof those completing training

q �n

t�s�1

RKt ý �s

t�1

NKt (2)

with 0 � q � 1.

Therefore the lower the retention rate q is, the lowerthe opportunity income gained from internal train-


ing is, and hence the lower the costs must be to en-courage firms to invest in training. If it is assumedthat workers who have completed vocational train-ing may be substituted by workers who have otherqualifications (from vocational academies, technicalcolleges, universities etc.), it can be shown that thecosts of these substitutions are already accommo-dated in the possibilities of recruiting (Niederalt2004). Short-term productivity differences due tothe firm-specific human capital which is imparted inon-the-job vocational training, as well as temporaryadditional wage costs which may be caused by exter-nal recruitment, can also be modelled by the recruit-ment costs (Niederalt 2004). Vocational training isconsequently always profitable with a given need forskilled labour if the net training costs are lower thanthe costs of the external recruitment of skilled work-ers, taking into account the retention rate of thosecompleting internal vocational training.

2.2 Uncertainty and vocational training

This retention rate is, however, unknown to firms atthe time of the training decision. Whether voca-tional training is profitable or not is thus uncertainfor them at this time. The higher the retention rateis estimated to be, the more likely it is that firms’vocational training is also valued as worthwhile.1

Three different sources of this uncertainty are dis-cussed below with regard to the impact on the deci-sion-making situation described.

Acemoglu/Pischke (1998, 1999b) argue on the basisof the classic human capital theory that provisionof vocational training vitally depends on the firm’sresignation rate. The authors reach the conclusionthat there are two equilibrium states of participationin training, namely the “low quit Ð high training”and the “high quit Ð low training” equilibriums. Theargument is structured as follows: if a firm’s rate ofvoluntary departures is high, the continued employ-ment of trained workers is also uncertain; the reten-tion rate is therefore estimated to be relatively low.This leads to the employer having less pronouncedmonopoly power over the workers, which in turn re-duces the incentive to invest in general human capi-tal. Beckmann (2002a) shows that this model can beexpanded to accommodate the risk of poaching, inother words the probability that a trained worker isrecruited away by another firm. It can be said invery general terms that low stability of employment

1 If, however, the firm does not incur any costs at all due to train-ing (NK = 0), vocational training is even worthwhile with an ex-pected retention rate of q = 0.

208 ZAF 2 und 3/2007

relationships also reduces the expected retentionrate and hence the opportunity income, thereforealso reducing the provision of vocational training.

In their theoretical analysis Acemoglu/Pischke pre-sume that the retention rate q depends primarily onthe exogenously-stated resignation rate of workersin the firm, but state elsewhere that q denotes theprobability that the worker or the firm, or both, suf-fer an adverse shock and therefore separate (Ace-moglu/Pischke 1999b). Stevens (1994) also empha-sises that it is continued employment per se whichis vital to the vocational training calculation and notwhether employment is terminated by the employeror employee. The reasons for retention or separa-tion after training can thus be traced not only to thetrainees and to potential third-party employers, butalso to the training firm itself. These sources of un-certainty are discussed in detail below.

One of these sources of uncertainty is for instancethe argument put forward by Baethge et al. (2007)that firms which are subject to strong national orinternational competitive pressure react to this byapplying new methods of enterprise control whichreduce firms’ involvement in training. This argumentcan be integrated into the decision-making calcula-tion with the aid of the above uncertainty considera-tions. The more a firm is in competition, the moredifficult it is for the firm to foresee at the time ofthe training decision whether it will be possible tooffer jobs to the trainees on completion of training.A firm which finds itself in such a situation will thustend to estimate a low retention rate q, which in-creases the incentive to meet the need for skilledworkers via external recruitment. According to thisargument, high competitive pressure can be re-garded as a source of uncertainty. Hence firms whichare subject to strong national or international com-petitive pressure can be expected to train fewer ap-prentices. This does not mean, however, that firmswhich find themselves in such a competition situa-tion have no need of skilled workers. If competitionis not for the lowest price, but for the highest quality,one may indeed anticipate the opposite. On the ba-sis of this argument it is rather to be anticipated thatthese firms will tend to meet their needs via the ex-ternal labour market instead of by investing in theirown vocational training programmes.

The organisation of vocational training can also beregarded as an even more significant cause of theuncertainty of continued employment because of itsconsequences. If one presumes, as Acemoglu/Pischke(1999b) do, that the need for skilled workers alsoincreases with modern production techniques, it ap-pears initially also to be consistent that these firms


are very much committed to vocational training. Theopposite may apply, however, if modern productiontechniques are also accompanied by constant chan-ges in these techniques, if the firms are thereforeoperating in highly innovative fields. Bartel/Lichten-berg (1987) state that jobs held by the highly skilledare more short-term, since the half-life of knowl-edge relating to new products and procedures is par-ticularly short. Heidenreich (1998) argues that thevocational training system that is typical of an indus-trial society is therefore coming under pressure inthree areas in line with the transition towards theservice society, and the new forms of organisationand work which this entails:

� Firstly, today’s vocational training system spansfirms, in other words the standardised, uniformtraining content is negotiated between the cham-bers, the employers’ associations and the trade un-ions. Firm-specific, and hence variable, require-ments as to qualifications, can be integrated inprinciple in the context of the dual training sys-tem, but this is hardly practiced at present.

� Furthermore, it is presumed that the industrialform of organisation, compartmentalised as it isaccording to functions with specific qualifications,constitutes a stumbling block to firms’ innovative-ness. Work and organisation processes are thusbeing restructured in many enterprises in order totake account of these new requirements. Voca-tional training is, however, primarily orientated inline with the old functional units. Apprentices aretherefore trained mostly in a particular depart-ment, and are more rarely given access to differ-ent areas or processes. Training therefore fre-quently becomes less attractive for innovative or-ganisations in particular.

� Finally, the dual training system is also coming un-der pressure as a result of educational expansion,and by virtue of the expansion of skill-intensiveservices. Because of educational expansion, firmscan also place school-leavers or graduates in me-dium-ranking posts where more interdisciplinarythinking and acting is expected. This group of in-dividuals thus competes with vocational traininggraduates. People completing training in the dualsystem are in a relatively poor position, which de-creases the perspectives of those with vocationaltraining, and the same therefore applies to theirattractiveness for future employers.

There is therefore uncertainty for the firm as towhether the skills imparted in vocational trainingare indeed useable and whether they must separate

ZAF 2 und 3/2007 209

from the trainees after completion of training, whichcorresponds to a lower expected retention rate q.The firm would then have to meet the skill require-ments by recruiting external skilled workers, despitehaving invested in vocational training, which accord-ingly reduces the opportunity income obtained frominternal training, thus making it less attractive.

In the case of uncertainty, firms therefore tend toopt against vocational training for a variety of rea-sons, and are more likely to recruit a skilled worker.This is firstly because at the beginning of vocationaltraining it is uncertain whether the trained workerwill leave the training firm of his/her own volition.Secondly, it is also possible for changes to take placeon the sales and buying markets, in particular thosesubject to high competitive pressure, which canmake one or more jobs obsolete. Thirdly, what ismore, firms are uncertain as to whether the skill re-quirements are met by the content of the training.These may have altered for instance as a result oftechnical and organisational changes.

3 Empirical analysis

3.1 Data

The IAB Establishment Panel of 2005 is used forthe analysis below. This is a survey which has beenconducted annually since 1993 in western Germanyand since 1996 in eastern Germany, generally usingoral interviews. The population covers all establish-ments which have at least one employee subject tosocial security contributions.2 Because of the funda-mental differences existing between the labour andtraining markets in eastern and western Germany(cf. Beckmann 2002b), the following analyses areconfined to western Germany, only. One focus ofthe IAB Establishment Panel in 2005 was the staffrecruitment process, in particular that of skilledworkers (cf. Bellmann et al. 2006).

3.2 Empirical strategy and econometricmodelling

According to the theoretical considerations, boththe recruitment of skilled workers and the provisionof vocational training can serve to meet the need forskills. Hence, both are treated as dependent varia-bles in the analysis. The share of training posts of-

2 For further information on the IAB Establishment Panel cf.Bellmann (2002).


fered in the previous year, in other words both thosethat were occupied and those which were vacant, isused as a measure of the need for trainees. Themeasure of the need for skilled workers is the shareof jobs offered to skilled workers in the first half ofthe year under observation, in each case in relationto all workers in the firm.3 Workers who have un-dergone vocational training, who have comparablevocational experience or a university degree areclassed as skilled workers.

Both share values are between zero and one, andindeed assume these values. The use of a tobitmodel for estimation is not ideal since a share valuebelow zero is not possible, but the tobit model isorientated towards censored values because of un-observability. A logit model taking the shape ofE (log [y / (1 - y)] |x) = x� could be estimated, butdoes not provide a perfect description of the availa-ble data since it is not defined for y = 0, and y = 1,and E (log [y / (1 - y)] |x) cannot be transformed intoE (y |x) without difficulty (Papke/Wooldridge 1996).Both share values are thus estimated with the frac-tional logit proposed by Papke/Wooldridge (1996),since this can represent the spread of the share values.

The fractional logit model, which is used here forestimating the shares, takes on the general form:

E (yi |xi) = G(xi �), (3)

where 0 � yi � 1 applies and xi is the vector of theexplanatory variables, � is the coefficient vector,i = 1,2,. . .,N and N is the number of observations.G (xi �) assumes the form of a probability densityfunction of the logistic function G(z), and 0 �G(z) � 1 applies. The coefficients are obtained bymaximising the Bernoulli log likelihood function

li (b) = yi log [G(xib)] + (1 - yi) log [1 - G(xib)] (4)

(Papke/Wooldridge 1996).

In order to determine the influence of the depen-dent variables on both the independent variables“share of training places offered” and “share of jobsoffered for skilled workers”, these are estimated us-ing the Seemingly Unrelated Cluster-AdjustedSandwich Estimator (SURE) proposed by Weesie(1999): This method makes it possible to study

3 The variable actually needed, namely the expected demand forskilled workers at the time of completion of vocational training,in other words in two to three years, is not available in the IABEstablishment Panel. For this reason, the current demand forskilled workers is used as a proxy variable.

210 ZAF 2 und 3/2007

whether the same independent variables have a sim-ilar impact on two (or more) dependent variables:

y1 = α0 + x1i �� + u1 (5.1)

y2 = γ0 + x2i �� + u2 (5.2)

Using a test proposed by Breusch/Pagan (1980), it isexamined whether the precondition for the use of aSURE approach is met, namely the non-diagonalityof the covariance matrix of the disturbance values.Furthermore, as in a Hausman test, the coefficientsof the individual estimates are compared (Weesie1999):

H0 : ��-�� = 0 (5.3)

Other conceivable methods, for example a Heck-man correction or the use of instrumental variables,are impossible since the need for skills Ð expressedon the one hand by the demand for skilled workersand on the other hand by the demand for trainees Ðdepends on the same influencing values. Both proce-dures, however, would require variables which cor-relate with the demand for apprentices, but not withthe demand for skilled workers or vice versa. How-ever, such variables are not available in the IAB Es-tablishment Panel.

Along with Heidenreich (1998) and Baethge et al.(2007), one should presume that the matching of theskills imparted to those actually required tends tohave more of a reducing impact on training in theservice sector than in manufacturing. A decline inemployment in the secondary sector, underwaysince the mid-nineties, has also been accompaniedby a fall in the number of trainees. The number ofemployees in the tertiary sector has grown in thesame period, but the number of trainees cannotkeep pace with this growth (Bundesministerium fürBildung und Forschung 2005). In order to accommo-date these different presumptions and developmentsin the secondary and tertiary sectors, the estimatesare implemented separately for the two economicsectors.

It should also be pointed out that the study pre-sented here does not measure uncertainty as such.Rather, the impact of (soft) indicators on trainingbehaviour is analysed. Accordingly, it cannot beruled out that the results are distorted. Because ofthe theoretical considerations, it is presumed thatthese indicators have an impact, on the anticipatedretention rate via the unobservable uncertainty and,hence, that they influence training. This analysis alsodoes not make it possible to say whether increaseduncertainty leads to less training since only a cross-


section of the IAB Establishment Panel is used as adata basis. Fixed effects, such as firm sizes or thesector, may also be the cause of the correlations thatwere identified.

3.3 Variables and hypotheses

In order to model uncertainty on the basis of com-petitive pressure, as proposed by Schank (2003), adistinction is made between competitive pressure onthe international markets (globalisation) and that on

ZAF 2 und 3/2007 211

the domestic market. The influence of globalisationis measured by the firm’s export share, i.e. its ex-ports as a percentage of turnover in the last businessyear. In line with the Acemoglu-Pischke model, anegative correlation with the share of training placesoffered and a positive one with the share of skilledworkers is anticipated since firms with internationallinkages are under greater competitive pressure andthis reduces the probability that they can offer thetrained workers jobs on completion of training. Aswith Schank (2003), the export share is portrayed bythree dummies each taking on the value of “one” ifthe share is 1Ð10 percent, 11Ð25 percent or morethan 25 percent respectively. These and the follow-ing considerations are summarised in Table 1.

A lower training intensity and a larger share ofskilled workers recruited is also to be expected forfirms with higher competitive pressure at nationallevel. A yardstick for competition intensity is theconcentration of workers in the sector in which thefirm is active. The calculation of the index is basedon the procedure according to Herfindahl-Hirsch-mann. If n states the number of firms, the index va-ries between 1/n (all firms have the same number ofworkers, considerable competition) and 1 (a firmhas all workers, no competition).4 A high indexvalue should therefore coincide with a large demandfor trainees. This can be put down to lower uncer-tainty as to continued employment by virtue of theless fierce competition in the sector. Firms which aresubject to less competition need spend less timeworrying about orders. One may also assume thatthe workers have fewer external options in less com-petitive sectors.

Three variables are included in the model whichrepresents the stability of the employment relation-ships since, according to the theory, high staff turno-ver leads to lower training intensity in firms. In ac-cordance with the “low quit Ð high training” and“high quit Ð low training” equilibriums derived byAcemoglu/Pischke (1998), the share of worker resig-nations as a percentage of all workers is included inthe estimate equations. According to the theory, anegative correlation with the share of trainees is ex-pected since in a firm with a large number of resig-nations the margin obtained by investing in generalhuman capital is reduced by the departure of thetrained workers. Also, a large share of worker resig-nations may signify considerable instances of suc-cessful poaching by other enterprises, also leading

4 The calculation is implemented using the establishment file ofthe Federal Employment Agency at 3-digit level of the classifica-tion of sectors.


to the expectation that a firm will react by reducingits training efforts (Beckmann 2002a). A similar ar-gument can be applied to the variables “share oftemporary employment” and the churning rate,5

which indicates the share of staff turnover that doesnot contribute to changes in the staff headcount.Both are also used as a measure of the stability ofemployment relationships. Accordingly, a negativecorrelation is presumed to exist in each case withthe share of trainees, and a positive one with thejobs offered for skilled workers.

The third source of uncertainty, namely the match-ing of the skills taught in training to rapidly chang-ing markets, is represented by five variables in themodel. It appears plausible that firms with a consid-erable skill requirement which are operating in a dy-namic market also employ many skilled workers.Hence, their share of the workforce is included inthe model. Furthermore, the amount of investmentin information technology and additional invest-ment as a share of all investment expenditure is ac-commodated. A dummy takes on the value of oneif the enterprise’s technical plant is highly modern.Furthermore, the amount of investment per workeris incorporated into the model. On the basis of thearguments put forward by Heidenreich (1998), whoclaims that the dual system of training is less adapta-ble to dynamic markets, a negative correlation is ex-pected in each case. Conversely, it follows from thisargument that firms with a considerable demand forskilled workers will be more likely to attempt tocover this need by recruiting additional skilledworkers.

Furthermore, variables are included in the modelwhich, although they are not directly associated withuncertainty, may nonetheless be relevant to thetraining decision. Since wage structure compres-sions, which count against skilled workers, constitutean incentive to provide training (Acemoglu/Pischke1998, 1999a, 1999b), as with Beckmann (2002a) adummy is used for denoting whether firms arebound by collective agreements. It can be presumedthat such firms invest more in training. The sameapplies to remuneration above the collectivelyagreed rates, which Acemoglu/Pischke interpret asan indication of efficiency wages, which in turn con-

5 This “churning rate” (CR) describes the part of worker mobilitywhich cannot be explained by the change in the number of work-ers in a firm. It is calculated as follows:CR = (H + S - | H - S |) / L, where H stands for the number of re-cruitments, S for the number of departures and L for the averagenumber of workers in the period under observation. There areseveral measuring concepts for the “churning rate”. The “churn-ing rate” used here is calculated as in Alda/Allaart/Bellmann(2005).

212 ZAF 2 und 3/2007

tribute to wage compression and are assumed tohave a positive impact on the share of trainees. Afurther dummy therefore takes on the value of oneif wages are paid above the collectively agreed rates.On the basis of the data available, however, it is notpossible, in this way to portray which groups of indi-viduals (e.g. highly-skilled, low-skilled or the train-ees themselves) enjoy this increased remuneration.

The economic situation of the firm is represented bytwo variables. A dummy is included if the firms haveevaluated their profit situation in the past year as“three” or worse on an ordinal scale from “one”(very good) to “five” (poor). Secondly, employmentgrowth in the firm is incorporated into the model.On the basis of the costs incurred in vocationaltraining, it can be presumed that firms in which em-ployment is growing are more likely to be able toafford training and that firms with a poor incomesituation are accordingly less committed to training(Beckmann 2002a).

The company-internal decision-making structurescan also be significant for the training decision. Twodummies are therefore formed which state whetherthe firm is foreign-owned or is an independent com-pany. A positive correlation is presumed to exist inthe case of independent companies, and a negativecorrelation is expected in the case of foreign enter-prises. It has been shown in empirical terms that in-dependent companies are more committed to train-ing than branches (Niederalt 2004). Foreign-ownedfirms are said to provide less training because of alack of training traditions and higher informationcosts on the dual system (Niederalt 2004). A furtherdummy states whether a works council exists in thefirm. Although no well-founded theoretical reason-ing exists to date, it has been shown in several em-pirical studies that the existence of a works councilhas a negative impact on the provision of training(Niederalt 2004).

Both Neubäumer/Bellmann (1999) and Niederalt(2004) have shown that the provision of training in-creases with firm size, but that training intensityfalls. Since the analysis observes both firms whichengage in training and firms which do not, as withBeckmann (2002a) one may not expect an effect tobe exerted by the ten firm size dummies included.A further reason for including the dummies is thedata set used. Since the respondent firms in the IABEstablishment Panel are selected using a stratifiedrandom selection (Bellmann 2002), it would be nec-essary according to DuMouchel/Duncan (1983) andWinship/Radbill (1994) to carry out the control pro-cedure for the level variables. For this reason, dum-mies for the federal state and for sectoral affiliation


are included in addition to firm size. Furthermore,the sector dummies also partly control for differentnet costs of training in the various economic activi-ties (Niederalt 2004). The distribution of the varia-bles6 used can be found in Tables A and B in theAnnex.

3.4 Results

Manufacturing

Table 2 shows the estimation results for manufactur-ing. An unexpected picture is revealed regarding thepotential sources of uncertainty as to being able toretain trained workers after training. While it is pos-sible to recognise the expected negative influenceof international competition on training intensity,measured against the export share, a significant ex-planation of the share of the training places offeredis, however, only provided by the dummies for a1 % to 10 % and for a 11 % to 25 % export share.The national competition index also has an influ-ence, but not the positive one that was anticipated.

In contradistinction to the theoretical considera-tions, the variables on employment stability in thefirm have no independent influence on the share oftrainees. That said, the churning rate and a largeshare of temporary contracts both have a positiveinfluence on the share of jobs offered for skilledworkers. In contradistinction to Beckmann (2002a),the share of resignations is not significant. High staffturnover appears not to exert an influence on thetraining intensity in manufacturing, but it does exertan influence on the demand for already qualifiedworkers.

The indicators of the modernity of the firm showcontradictory results. Manufacturing firms whichmake large investments also train more. The shareof skilled workers exerts a positive influence on thedemand for external skilled workers, and has a nega-tive correlation, albeit not a significant one, withtraining intensity. Also, these variables thus impactat best on the demand for skilled workers, but noton that for trainees.

The estimation of the profit situation does not exertan influence on the training intensity, and not onthe demand for skilled workers in the firms, either.However, a link can be identified between the

6 The age of the firm and the regional unemployment rate wereinitially included in the estimate model. Like the separate esti-mate by firm size classes, however, this provided no additionalinformation.

ZAF 2 und 3/2007 213

growth of the firm and the demand for skilled work-ers, but there is no independent influence on train-ing intensity. As might be expected, foreign-ownedfirms also train less. The negative influence of theworks council on training is in line with the resultsobtained by Niederalt (2004). The positive influenceof collective bargaining on the provision of trainingalso corresponds with the results from other surveys(e.g. Niederalt 2004, Beckmann 2002a). Firms boundby collective agreements train more.

To sum it up, it can be stated that the considerationsmade regarding uncertainty and its impact on train-ing and the external recruitment of skilled workersin manufacturing are not confirmed. The theoreticalapproach is therefore not well suited to explain thetraining behaviour of firms in manufacturing. Theonly slightly significant Breusch-Pagan test also cor-responds with this picture. Hence, in manufacturingthere tends not to be a correlation between the de-mand for skilled workers and that for trainees. Itshould be pointed out once more here that the indi-cators used for the various causes of uncertaintytend to be soft, and therefore distortions due to mis-specifications cannot be ruled out.

Services

The fact that the demand for skilled workers andthe demand for apprentices in this sector of theeconomy are interrelated is also confirmed by theresult of the Breusch-Pagan test. The estimate re-sults for the service sector from Table 3 also confirmthe considerations as to uncertainty.

This does not apply to the variables which measurenational and international competitive pressure,however. None of these have any significant impacton training behaviour.

The negative link between the share of skilled work-ers and the provision of training, coupled with thepositive link with the demand for skilled workers,supports Heidenreich’s thesis (1998). Both of the co-efficients also differ highly significantly from one an-other. Service-providers with a relatively large num-ber of skilled workers are therefore more likely tomeet their need for skilled workers by recruitingnew workers rather than by engaging in vocationaltraining. Heidenreich (1998) puts this down to thelow level of adaptability of dual vocational trainingto the requirements of a knowledge-based society.It therefore also appears to be plausible that thiseffect is found more in the growing service sectorthan in the secondary sector. The other variables in-tended to portray the modernity or innovativenessof enterprises, however, have no significant effect on


214 ZAF 2 und 3/2007


ZAF 2 und 3/2007 215


the share of training places offered, which ratherputs the significance of these considerations intoperspective.

The findings on staff turnover and flexibility alsocorrespond with the theoretical considerations. Botha high churning rate and a large share of temporaryemployment relationships reduce firms’ provision oftraining. At the same time, both variables have apositive impact on the demand for skilled workersand differ significantly from one another. The resig-nation rate has no significant effect. In this sense,the thesis of Acemoglu/Pischke (1998) of the “lowquit Ð high training” and the “high quit Ð low train-ing” equilibriums appears initially to have beencountered. The idea on which this is based, namelythat high staff turnover is an indication of a lowerprobability of continued employment, to whichfirms react by reduced provision of training, is con-firmed, however.

As is the case in manufacturing, in the service sector,too, the profit situation and employment growth donot impact on the provision of training. However, apositive effect of growth on the demand for skilledworkers can be ascertained. The further control var-iables “works council”, “foreign owners” and “inde-pendent company” demonstrate a similar picture tothat in manufacturing. However, collective agree-ments do not impact on training behaviour, whichcan be explained by the wage structure in the servicesector. Collectively-agreed wages are very low insome of the sectors, in particular in personal servi-ces.

Summing up, the empirical results for the servicesector bear out the theses put forward in this paperregarding the link between uncertainty and voca-tional training and the external recruitment ofskilled workers.

4 Summary and conclusions

In addition to high staff turnover (Acemoglu/Pischke 1998, 1999a, 1999b), this paper analyses twofurther causes of a high separation probability, inother words of uncertainty as to the retention oftrainees in the firm after completion of training. Itis stated that high competitive pressure can reducethe provision of training (Baethge et al. 2007) sincefirms are in greater competition, and hence find itdifficult to estimate their requirements and cannotbe certain of being able to offer a job on completionof training. Furthermore, according to Heidenreich(1998), vocational training is also less attractive for

216 ZAF 2 und 3/2007

modern and innovative enterprises since Ð given itsrelatively inflexible organisation Ð firms have littlelatitude to influence the training content, and there-fore it cannot be predicted at the beginning of thetraining whether the acquired qualifications will beuseable once training has been completed. If for oneor more of the three mentioned reasons, firms areuncertain as to whether they will be able to take onworkers after training, one may presume that in thecase of a need for skilled workers the recruitmentof already qualified workers will be preferred, andthat commitment to training will be reduced accord-ingly.

This paper did not measure directly the indicatorsof uncertainty, but (soft) indicators from the IABEstablishment Panel for western Germany wereused and their impact on vocational training behav-iour studied. On the basis of the theoretical consid-erations, these indicators are presumed to exert animpact via the unobservable uncertainty regardingthe anticipated retention rate, and hence exert animpact on training. The fractional logit estimatorproposed by Papke/Wooldridge (1996) is deployedhere. In order to accommodate the alternative hiringof skilled workers, both shares of employment areestimated using the Seemingly Unrelated Cluster-Adjusted Sandwich Estimator proposed by Weesie(1999).

Little evidence is available in manufacturing toconfirm the theoretical considerations. The resultsobtained in the service sector correspond on thewhole to the considerations put forward in this pa-per as to the significance of uncertainty for voca-tional training behaviour. While competitive pres-sure does not take on any major significance, theexpected correlations are, however, confirmed forboth of the other uncertainty factors Ð i.e. high staffturnover and the matching of vocational training tothe actual skill requirements. Firms with considera-ble movement in the headcount and with a largeshare of skilled workers train less and at the sametime have a larger demand for skilled workers whoalready have the requisite skills. Vocational trainingis obviously not optimal for these firms with regardto their need for skilled workers. Uncertainty as toretention after training can help to explain the grow-ing gap between employment and training in the ter-tiary sector in this context. However, it should benoted here that the indicators used for the variouscauses of uncertainty tend to be weak, and thereforedistortions caused by mis-specifications cannot beruled out. Furthermore, because of the cross-sec-tional view, the results may be influenced by fixedeffects such as firm size. The results can thus onlybe evaluated as indications of a link between uncer-


tainty with regard to the retention rate and trainingcommitment.

If this relationship were to be confirmed in furtheranalyses, it would mean that the dual system of voc-ational training is faced not only by quantitativechallenges against the background that, firstly, fu-ture employment gains are to be expected more inthe tertiary sector and, secondly, employment rela-tionships are becoming increasingly flexible. In or-der to make the provision of training more attrac-tive in these sectors, and to give it a future, one ap-proach to be explored is that of greater interlinkingbetween practical work experience, higher educa-tion and dual training. The growing trend towardsmore dual courses of study (cf. Ausbildung-Plus.de2006) is an indication that the stakeholders have al-ready recognised this need. One must also considerhow current changes in the demand for skills can beintegrated more rapidly into the dual system. Fur-thermore, the increasing flexibilisation of the labourmarket, for which many are calling, is to be criticallyanalysed against the background of these results.

References

Acemoglu, D./Pischke, J.-S. (1998): Why do firms train?Theory and evidence. In: The Quarterly Journal of Eco-nomics 113, 79Ð119.

Acemoglu, D./Pischke, J.-S. (1999a): Beyond Becker: Train-ing in imperfect labour markets. In: Economic Journal109, pp. F112ÐF142.

Acemoglu, D./Pischke, J.-S. (1999b): The structure of wagesand investment in general training. In: Journal of Politi-cal Economy 107, 539Ð572.

Ausbildung-Plus.de (2006): Projekt Ausbildung-Plus. Infor-mationssystem über Zusatzqualifikationen in der Be-rufsbildung und duale Studiengänge. Institut der deut-schen Wirtschaft, gefördert vom Bundesministerium fürBildung und Forschung.

Alda, H./Allaart, P./Bellmann, L. (2005): Churning and in-stitutions. Dutch and German establishments comparedwith micro-level data. IAB Discussion Paper 12/2005.

Baethge, M./Solga, H./Wieck, M. (2007): Berufsbildung imUmbruch: Signale eines überfälligen Aufbruchs. Fried-rich Ebert Foundation: Berlin, Bonn. http://library.fes.de/pdf-files/stabsabteilung/04258/index.html,accessed on 12 March 2007.

Bardeleben, R. v./Beicht, U./Fehér, K. (1995): BetrieblicheKosten und Nutzen der Ausbildung. RepräsentativeErgebnisse aus Industrie, Handel und Handwerk. Be-richte zur beruflichen Bildung 187, Bielefeld: Bertels-mann.

ZAF 2 und 3/2007 217

Bartel, A. P./Lichtenberg, F. R. (1987): The ComparativeAdvantage of Educated Workers in Implementing NewTechnology. In: The Review of Economics and Statistics69(1), 1Ð11.

Becker, G. S. (1964): Human Capital. A Theoretical andEmpirical Analysis, with Special Reference to Educa-tion. New York, London Ð Columbia University Press.

Beckmann, M. (2002a): Lohnstrukturverzerrungen und be-triebliche Ausbildung. Empirische Analyse des Acemo-glu-Pischke-Modells mit Daten des IAB-Betriebspanels.In: Mitteilungen aus der Arbeitsmarkt- und Berufsfor-schung 35(2), 189Ð204.

Beckmann, M. (2002b): Wage compression and firm-spon-sored training in Germany: Empirical evidence for theAcemoglu-Pischke Model from a zero-inflated countdata model. In: Applied Economics Quarterly (Kon-junkturpolitik) 48, 368Ð387.

Bellmann, L. (2002): Das IAB-Betriebspanel. Konzeptionund Anwendungsbereiche. In: Allgemeines StatistischesArchiv 86(2), 177Ð188.

Bellmann, L./Bielenski, H./Bilger, F./Dahms, V./Fischer, G./Frey, M./Wahse, J. (2006): Personalbewegungen undFachkräfterekrutierung. Erste Ergebnisse des IAB-Betriebspanels 2005. IAB Forschungsbericht 11/2006.

Bellmann, L./Hartung, S. (2005): Zu wenig Stellen unddoch sind nicht alle besetzt. IAB Kurzbericht 27/2005.

Breusch, T. S./Pagan, A. R. (1980): The lagrange multipliertest and its applications to model specification in econo-metrics. In: The Review of Economic Studies 47(1),239Ð253.

Bundesministerium für Bildung und Forschung (2005): Be-rufsbildungsbericht 2005. Bonn, Berlin.

Cramer, G./Müller, K. (1994): Nutzen der betrieblichenBerufsausbildung. In: Beiträge zur Gesellschafts- undBildungspolitik 195, Institut der deutschen Wirtschaft.Köln.

DuMouchel, W. H./Duncan, G. J. (1983): Using sample sur-vey weights in multiple regression analysis of stratifiedsamples. In: Journal of the American Statistical Associa-tion 78, 535Ð543.

Falk, M. (2002): IKT-Fachkräftemangel und Anpassungs-reaktionen der Unternehmen. In: Bellmann, L./Velling,H. (eds.): Arbeitsmärkte für Hochqualifizierte. Beiträgezur Arbeitsmarkt- und Berufsforschung 256. Nürnberg,37Ð73.

Franz, W./Soskice, D. (1995): The German apprenticeshipsystem. In: Buttler, F./Franz, W./ Schettkat, R./Soskice,D. (eds.): Institutional frameworks and labor marketperformance. London, New York, 208Ð234.

Hartung, S./Leber, U. (2004): Betriebliche Ausbildung undwirtschaftliche Lage * empirische Ergebnisse des IAB-Betriebspanels. In: Krekel, E./Walden, G. (eds.): Zu-kunft der Berufsausbildung in Deutschland. Empirische


Untersuchungen und Schlussfolgerungen. Ergebnisseder BIBB-Fachtagung am 4./5. November 2003 in Bonn.Berichte zur beruflichen Bildung 273. Bielefeld. 111Ð129.

Heidenreich, M. (1998): Die duale Berufsausbildung zwi-schen industrieller Prägung und wissensgesellschaftli-chen Herausforderungen. In: Zeitschrift für Soziologie27(5), 321Ð340.

Kau, W. (1997): Ausbildung in Großbetrieben: Gewinnoder Belastung? In: Beicht, U./Berger, K./Herget, H./Krekel, E. (eds.): Berufsperspektive mit Lehre; Wertund Zukunft dualer Berufsausbildung. Berichte zur be-ruflichen Bildung 211. Bielefeld, 333Ð343.

Neubäumer, R./Bellmann, L. (1999): Ausbildungsintensitätund Ausbildungsbeteiligung von Betrieben: Theore-tische Erklärungen und empirische Ergebnisse auf derBasis des IAB-Betriebspanels 1997. In: Beer, D./Frick,B./Neubäumer, R./Sesselmeier, W. (eds.): Die wirtschaft-lichen Folgen von Aus- und Weiterbildung. Münchenand Mering: Hamp, 9Ð41.

Niederalt, M. (2004): Zur ökonomischen Analyse betrieb-licher Lehrstellenangebote in der BundesrepublikDeutschland. Europäische Hochschulschriften. Series 5,Volks- und Betriebswirtschaft * 3082. Frankfurt a. M.:Lang.

Papke, L. E./Wooldridge, J. M. (1996): Econometric meth-ods for fractional response variables with an applicationto 401(K) plan participation rates. In: Journal of AppliedEconometrics 11, 619Ð632.

Schank, T. (2003): Die Beschäftigung von Un- und Ange-lernten. Eine Analyse mit dem Linked Employer-Em-

218 ZAF 2 und 3/2007

ployee Datensatz des IAB. In: Mitteilungen aus der Ar-beitsmarkt- und Berufsforschung 36(3), 257Ð270.

Smits, W./Zwick, T. (2004): Why do business service firmsemploy fewer apprentices? A comparison between Ger-many and the Netherlands. In: International Journal ofManpower 25(1), 36Ð54.

Steedman, H./Wagner, K. (2007): The impact of rationalICT qualification systems on companies’ recruitmentpractices Ð an Anglo-German comparison. In: ZAF40(2 + 3).

Steedman, H./Wagner, K./Foreman, J. (2006): Das Angebotan IKT-Fachkräften im Vereinigten Königreich und inDeutschland. In: Europäische Zeitschrift für Berufsbil-dung 37, 9Ð23.

Stevens, M. (1994): A theoretical model of on-the-job train-ing with imperfect competition. In: Oxford EconomicPapers 46(4), 537Ð562.

Weesie, J. (1999): Seemingly unrelated estimation and thecluster-adjusted estimator. In: Stata Technical Bulletin52, 231Ð248.

Winship, C./Radbill, L. (1994): Sampling weights and re-gression analysis. In: Sociological Methods and Research23, 230Ð257.

Wolter, S. C./Schweri, J. (2002): The cost and benefit of ap-prenticeship training: The Swiss case. In: Applied Eco-nomics Quarterly (Konjunkturpolitik) 48, 347Ð367.

Wolter, S. C./Mühlemann S./Schweri, J. (2006): Why somefirms train apprentices and many others do not. In:German Economic Review 7(3), 249Ð264.


Annex

ZAF 2 und 3/2007 219


220 ZAF 2 und 3/2007

To recruit skilled workers or to train one's...

Documents

Transcript of To recruit skilled workers or to train one's...