Is labour migration disrupting dual vocational education and training systems? Empirical evidence from the Danish and Norwegian construction sectors

Abstract

This article studies how labour migration affects dual vocational education and training (VET) systems. We argue that because dual VET systems rely on employers engaging in training, an alternative source of labour – such as labour migrants – may make employers less likely to train, especially when pressure on industrial relations institutions makes it possible for employers to use migrant labour as low-wage labour. Drawing on linked employer-employee register data from Denmark and Norway, we us logistic and Poisson regression to analyse whether changes in the level of labour migration in regional subsectors of the construction sector in these two countries affects firms’ hiring of apprentices. We find that despite dissimilar developments in the labour migration level, these levels nonetheless correlate with the intake of apprentices in both countries. The results suggest that labour migration into countries with dual VET systems may have long-term effects on their skill formation systems.

Keywords

Labour migration vocational education and training EU enlargement construction sector skill formation

Introduction

How does labour migration affect dual vocational education and training systems? A key question raised by the transnationalisation of labour markets is how the growing flow of labour across borders affects national skill formation systems. Research has shown that countries with a longer history of immigration are typically liberal market economies (LMEs) (Devitt, 2011). In contrast, countries with more coordinated market economies (CMEs) and a stronger reliance on vocational educational systems typically experience less immigration. However, since the EU enlargement of 2004, CMEs in Western Europe have experienced an unprecedented inflow of migrant labour, which allows us to study the effect of migrant labour on the vocational education and training (VET) systems of these countries.

This article concerns dual VET systems (Fürstenau et al., 2014) in which part of the education is based on on-the-job training while the other part is in-school education. These systems have been very successful in striking a balance between securing school-to-work transition and long-term employment prospects (Jørgensen, 2018). The combination of regulated firm-based training and in-school education ensures that students get work-relevant skills that can be used broadly within their occupation and not just the training firm. The strong involvement of employers is generally considered a strength of these VET systems, but it can also be a weakness because the system depends on the willingness of employers to partake in the training efforts. Because the dual VET systems requires that employers provide training beyond their specific skill needs, some employers may defect from training efforts to avoid the cost of training. This may be especially true if labour migrants offer employers with an alternative source of labour, or if increasing cost competition from firms using low-wage migrant workers makes it difficult for firms to sustain the training cost. The research question raised in this article is therefore whether labour migration will reduce the hiring of new apprentices in dual VET systems.

To study this question, we present a statistical analysis of the relationship between labour migration and trends in apprenticeship uptake by firms in the construction sectors of Norway and Denmark. Both countries have a dual VET system and both countries have experienced a pronounced growth in labour migration since the EU enlargements of 2004. We argue that when labour migration makes the enforcement of industrial relations such as minimum wages and collective agreements difficult, firms can increasingly substitute skilled, high-wage workers with low-wage migrant workers. This may affect firms’ incentive and willingness to train apprentices. We then draw on linked employee-employer register data from the two countries to show that increases in labour migration are associated with a decline in the number of new apprentices hired relative to the economic situation. Consequently, labour migration is incrementally destabilizing dual VET systems.

Labour migration, institutional strain, and decline in apprenticeships

Since the 2004 EU enlargement, the free movement has spurred unprecedented westward flows of labour from Central and Eastern European countries. This flow has been driven by the combined effect of relatively low wages and high unemployment in the new member states (push) and relatively strong demand for labour in older member states (pull) (Windzio et al., 2021). Substantial research on this migratory flow has shown that labour migration puts pressure on wages and working conditions, while also challenging national industrial relations institutions, such as minimum wages or collective bargaining (Dølvik and Eldring, 2015; Felbo-Kolding et al., 2019; Meardi, 2012). Specifically for the construction sector, studies have shown that migrant labour is used as a hyper-flexible buffer to carry the burden of market uncertainty (Frangi et al., 2021; Meardi et al., 2012), while at the same time shifting the balance of power between employers and workers (Arnholtz et al., 2018). This is especially true for so-called posted workers. While regular labour migrants are in principle covered by all labour regulation of the country they work in, this is only partially so for posted workers. Therefore, the presence of these posted workers increases the pressure on labour market institutions (Arnholtz, 2021; Bosch and Weinkopf, 2013; Lillie et al., 2014). However, the pressure is also observed for ordinary migrant workers. All in all, these studies show how labour migration puts industrial relations of especially the construction sector under strain, and this is particularly true for the enforcement of these institutions.

By contrast, few studies have focused on how labour migration affects skill formation and especially dual VET systems in CMEs (although see Røed and Schøne, 2016; Arnholtz and Wright, 2023). To understand the potential impact of labour migration on dual VET systems, we must first understand how the dual VET systems work. Here, it is important to distinguish between countries with purely school-based VET systems (like Sweden and France) and countries with dual VET systems (like Germany, Denmark, and Norway) (Busemeyer and Trampusch, 2012, 2019). The dual VET systems provide the apprentices with a mix of school-based education and workplace training. Comparative studies of VET systems show that dual VET systems tend to achieve a smooth transition from education to working life. On the one hand, the gradual socialization into work life and firms’ use of apprenticeship as a channel of recruitment ensures good immediate employment prospects (Jørgensen, 2013; Nyen et al., 2015). On the other hand, regulation of the workplace training ensures that apprentices acquire portable occupational skills (rather than only firm-specific skills), and trade unions and employer associations’ involvement in the governance of the VET system guarantees a high degree of legitimacy and recognition of these occupational skills (Jørgensen and Tønder, 2018; Streeck, 1989a).

However, the dual VET systems rely on employers’ willingness to train apprentices and can therefore face challenges regarding insufficient apprenticeship positions (Streeck, 1989a). In this context, Busemeyer and Iversen (2011) highlight that employers may defect from VET training if they can recruit the labour and skills they need through the market rather than training. This is where labour migration comes in. Instead of poaching skilled workers from domestic competitors via higher wage costs, firms may access a new source of labour by recruiting labour from abroad. In some cases, firms may need to restructure their production to utilize migrant workers’ skills, but research suggests that firms have done so in situations with a high inflow of labour migrants (Haakestad and Friberg, 2017). This often involves opting for a low-wage, low-skill strategy (Arnholtz and Ibsen, 2021).

However, different types of institutional embeddedness restrain firms’ opportunity to use labour migrants for such low-wage, low-skill strategies. On the one hand, industrial relations institutions such as minimum wages or collective agreements may hamper employers’ use of migrant labour as cheap labour (Edo and Rapoport 2019). Additionally, relatively high minimum wages may incentivize firms to train apprentices to increase workers’ productivity under conditions of down-inelastic wages (Busemeyer and Iversen, 2011). On the other hand, different types of occupational licencing and systems for recognition of qualifications may limit the use of labour migrants in specific sectors and jobs while also incentivizing employers to train workers that can comply with licencing requirements (Aerne and Trampusch, 2023; Sweetman et al., 2015; Trampusch, 2019). Thus, the institutional embeddedness of firms in CMEs may limit their use of labour migrants (Kerr, 1977; Streeck, 1989b) and thus explain why CMEs are found to experience less labour migration than LMEs (Afonso and Devitt, 2016).

However, if industrial relations institutions and their enforcement are put under pressure from labour migration as suggested by the literature, employer may have more leeway to use low-wage migrant labour as a replacement for both apprentices and skilled workers. While Nordic construction sectors have typically followed a high-road of high-skilled workers and high wages (Arnholtz and Ibsen, 2021), studies have found that firms gradually restructure their production processes to accommodate the skills and wage demands of migrant workers (Haakestad and Friberg, 2017). We suggest that a link between increasing labour migration and declining apprenticeship can occur through two mechanisms. First, there is a direct link where a firm fills an open position with a labour migrant rather than an apprentice. More realistically, the firm will scale down its involvement in training apprentices while scaling up its use of labour migrants because it no longer views training efforts as important now that migrant labour is available. The key thing in this direct link is that the replacement occurs within the same firm. Additionally, there is an indirect link between increasing labour migration and diminishing numbers of apprentices. When some firms start to use migrant labour as a low-wage source of labour, this may increase cost competition in the market. This may give other firms not using low-wage migrants less leeway to accept the risk and cost of training. Thus, even firms that do not recruit labour migrants themselves may become less willing or able to take on apprentices because of increasing market pressure. Especially smaller firms that act as subcontractors or suppliers for larger firms may find themselves under increasing competitive pressure entailing that they can no longer risk taking on apprentices. Rather than focussing only on the direct replacement of apprentices with labour migrants within each firm, we therefore need analyse whether the overall level for labour migration within a labour market affects firms’ propensity to take on apprentices. Yet, we hypothesize that labour migration may reduce the hiring of new apprentices in dual VET systems.

Obviously, there are many other factors than labour migration that firms’ propensity to hire apprentices. Most important among them are firm size, developments in the business cycle and demographic effects that will influence the number of students seeking apprenticeship positions (Wolter et al., 2006; Brunello 2009; Muehlemann et al., 2009). First, the size of a firm matters for its ability to take on apprentices both for reasons of cost and capacity to train. We therefore need to control for firm size to ensure that developments in apprenticeship uptake is not due to changes in the size composition of firms. Second, the economic cycles affect both the recruitment of migrants and the intake of apprentices. Firms hire more labour migrants and are more willing to take in apprentices during economic booms than during periods of economic slump. To identify the effect of labour migration we thus need to control for business cycle developments. Third, the number of students seeking apprenticeship positions will influence how many positions are offered. With some reservations, the number of young people entering the first part of the VET education can be viewed as students’ demand for VET education, while the number of students who get an apprenticeship during the second part of the education can be viewed as firms’ supply of apprenticeship positions¹. We want to study whether the share of labour migration correlates with changes in the supply of apprenticeship positions when controlling for the demand for them.

Case selection and background on case countries

To study how labour migration may affect dual VET systems, we focus on the construction sectors of Denmark and Norway. Both countries are CMEs with dual VET systems, which is the topic of interest in this article. We focus on the construction sector since it has a long tradition for VET in most CMEs, and it is a sector where substantial labour migration occurs in most countries. While this makes the sector ideal for observing the effect on labour migration on dual VET apprenticeship employment, a long-term adjustment of apprenticeship levels to labour migration may make the effect unobservable. Incidentally, the Nordic construction sectors have experience relatively modest labour migration up until the 2004 EU enlargement. Thereafter, the construction sectors of both Denmark and Norway were exposed to what can be regarded as a natural experiment in the form of a massive increase in the share of migrant workers (see Tables 1 and 2). Therefore, they are ideal cases for observing an effect that may be otherwise hidden due to gradual adjustment. That said, Norway has received far more labour migrants than Denmark, while Denmark has a longer history of dual VET than Norway. Our initial hypothesis (based on interviews in both countries) is therefore that we will only find an effect in Norway, while stronger institutional traditions and lower inflow will result in no effect in Denmark.

Table 1.

Training and Migration in the Norwegian Construction Sector, 2003–2015.

Year	Training firms		New apprentices		Labour migration
Year	Firms	Percent	Persons	Percent	Persons	Percent
2003	2.301	16.2	3.672	2.9	4.547	3.5
2004	2.517	17.0	4.092	3.1	4.951	3.8
2005	2.685	17.4	4.330	3.2	6.102	4.5
2006	2.961	18.2	5.194	3.6	8.581	5.9
2007	3.170	18.1	5.850	3.7	12.791	8.1
2008	3.092	16.2	5.391	3.2	16.026	9.6
2009	2.557	13.4	4.142	2.5	15.921	9.7
2010	2.797	14.4	4.576	2.7	18.542	11.1
2011	2.767	13.7	4.768	2.7	23.844	13.6
2012	2.760	13.1	4.855	2.7	27.510	15.2
2013	2.732	12.3	4.766	2.5	30.817	16.4
2014	2.822	12.3	4.800	2.5	33.092	17.1
2015	2.879	13.2	4.875	2.6	32.382	17.2

Table 2.

Training and Migration in the Danish Construction Sector, 2003–2015.

Year	Training firms		New apprentices		Labour migration
Year	Firms	Percent	Persons	Percent	Persons	Percent
2003	3.081	11.8	4.148	2.5	4.780	2.9
2004	3.294	12.3	4.438	2.6	5.094	3.0
2005	3.669	13.0	4.917	2.7	6.365	3.5
2006	4.213	13.9	6.093	3.1	9.815	5.0
2007	4.467	14.1	6.562	3.3	13.904	6.9
2008	3.865	11.2	5.539	2.9	16.033	8.4
2009	2.776	8.5	3.664	2.2	12.079	7.2
2010	2.818	9.0	3.710	2.3	10.107	6.4
2011	2.918	9.7	3.987	2.5	10.917	6.7
2012	2.781	9.4	3.723	2.3	11.895	7.4
2013	2.549	8.6	3.472	2.2	13.323	8.4
2014	2.672	8.9	3.791	2.3	16.529	10.1
2015	2.766	9.1	4.080	2.4	18.154	10.7

Overall, the two countries are quite similar. In both Denmark and Norway, the VET programmes consist of two different distinct periods. The first period is primarily school-based and gives students basic skills within a broad educational field (for instance, construction). The second period is more specialized and consists primarily of workplace-based training as apprentices (with some periods of school teaching in Denmark) (Jørgensen, 2018). Therefore, this second part of the education requires students to obtain an apprenticeship position in a relevant firm. It is this transition from the first to the second part of the education we use to control for the demand for apprenticeship positions. There are also some differences between the two countries. The Danish VET system has long historical roots and trade unions and employer associations are strongly involved in its governance. In comparison, the current Norwegian VET system has a shorter history and is more state-governed. The Danish VET education starts with 1 year of school followed by 3 years of firm-based apprenticeship, while in Norway it starts with 2 years of school followed by 2 years of firm-based apprenticeship. This difference implies that Danish firms can be expected to take in new apprentices less often than Norwegian firms if they stay continuously involved in VET training (every 3 years as opposed to every other year). This may imply that it is more difficult to observe the association between labour migration and intake of apprentices in Denmark because there is a greater time lag between the rise in labour migration share and firms’ potential reaction to it. However, in both countries, firms take on apprentices every year, implying that some firms will be able to react every year.

Regarding the economic cycles, both the Danish and Norwegian construction sectors were booming before the 2008 financial crisis, creating a labour demand that stimulated labour migration (Figure 1). However, the crisis affected the two countries and their construction sectors quite differently. The Norwegian construction sector experience only a 6% decline in employment from 2008 to 2010 and had surpassed its 2008 employment level by 2012. By contrast, the Danish construction sector saw an 18% drop in employment from 2008 to 2010 and recovered slowly. In other words, the demand for apprentices and labour migrants should be much stronger in Norwegian than in the Danish construction sector.

Figure 1.

Employed persons in the Danish and Norwegian building and construction sectors. Source: OECD data.

Methods and data

The literature on changes to VET systems is often based on case studies of policies and institutional developments (Busemeyer and Trampusch, 2012, 2019; Carstensen and Ibsen, 2019; Ibsen and Thelen, 2017). However, such case studies of institutional changes should be supplemented by ‘econometric micro-level studies which investigate “from below” how the hiring and recruitment behaviour of firms (or workers’ training behaviour) is affected by immigration, as firms replace the training by importing skills from abroad’ (Trampusch, 2019: 168). While policies and institutions clearly matter, in a system where firms play such a decisive role in the skill formation system, changes in their aggregate behaviour are also of great importance. Therefore, we focus on how a firm’s propensity to hire apprentices changes when levels of labour migration change.

More specifically, we have two dependent variables: on the one hand, whether firms become ‘training firms’ by taking on new apprentices each year, and on the other hand, firms’ ‘training rate’ understood as the share of newly hired workers who are hired into an apprentice position. While both concepts focus on firms’ recruitment of apprentices, the former is better at grasping changes among smaller firms in which hiring an apprentice is an either or choice, while the latter is better at grasping changes among larger firms, where hiring apprentices is more a matter of how many. Our focal variable is the share of labour migrants out of total employment in the specific sector and region in which the firm is situated. There are three considerations behind this measure. First, as explained, there may be an indirect link between labour migration and apprenticeship development, which requires us to look at labour migration in the labour market rather than the individual firm. Second, to construct these, labour markets exploit geographical differences in the distribution of labour migrants. Third, we exploit that occupational licencing and other forms of occupational regulation makes developments in labour migration less pronounced in licensed occupations (such as electricians and plumbers) than in unlicensed occupations (painters, carpenters). We thus create variation in the share of labour migration between these labour markets. We control for the supply of apprenticeship candidates, firm size, business cycle effects, and subsector and region invariant effects.

Empirically, we use linked employer-employee data drawn from different administrative registers in Norway and Denmark to construct panel data with annual observations for the period 2003–2015. This is the period where a significant increase in the inflow of labour migrants took place in both countries. Furthermore, this is a period of relative stability in the institutional setup of the two countries. In 2015, the Danish VET system underwent a noteworthy reform that might have changed both the supply of apprenticeship candidates and apprentices’ intake. Therefore, we ended our analysis before this reform took effect. The data do not include so-called posted workers since they are formally employed abroad while working in Denmark or Norway. However, while much discussed in the literature posted, workers make up a relatively small share of migrant labour in the two countries construction sectors. Due to confidentiality concerns about the administrative data, there are restrictions to the use of the data that imply that we can analyse the data from the two countries only separately. This excludes the possibility of integrating country differences into the models. Still, having data for both countries is important for strengthening the validity of our argument. Developments in both apprentice intake and labour migration levels can be affected by idiosyncratic national factors, but performing an identical analysis in two countries strengthens our belief in their results if these are similar.

To construct the panel, we started by identifying all active establishments in building and construction (BaC) industries each year from 2003 to 2015. In both countries, we identified around 30,000 such firms each year, most of which were observed repeatedly over the years (>5 years pf observation per establishment on average). For the selected establishments, we collected detailed information on sector (3-digit NACE codes), region (NUTS level 3), the number of employees, and the number of newly hired employees (employees registered for the first time in the establishment) each year 2003–2015. For each person with a valid employment relationship with one of the selected establishments, we used data on ongoing education from education registers to identify apprentices within the fields of building and civil engineering (ISCED-F code 0732) or electricity and energy (ISCED-F code 0713).

We use standard and fixed-effects logistic and Poisson regression to estimate the effect of labour migration on the odds of being a training firm and the effect of labour migration on firms’ training rate, respectively. The structural model is the same in all specifications:

Y_{i t} = {L M R}_{i t} + {S T U D E N T S}_{i t} + {F I R M S I Z E}_{i t} + {SEC}_{i} + {R E G}_{i} + {Y E A R}_{i t} + u_{i} + ε_{i t}

Here, Y_it represents our two dependent variables of interest (either being a training firm or a firm’s training rate). We classify firms as training firms if they had at least one employee who started their apprenticeship in a given year. This implies that some firms with older apprentices are not classified as training firms because they have no new apprentices. The reasoning for this is to sensitize variation in the dependent variables to changes in the independent variables. A firm’s training rate was calculated as the number of new apprentices (persons identified as apprentices for the first time in the establishment) divided by the number of newly hired employees in that firm.

The labour migration rate (LMR_it) is our focal variable. For each year, we calculate the labour migration rate as the number of migrant workers within the establishment’s detailed subsector and region divided by the total number of workers within the establishment’s detailed subsector and region. Rather than focussing on development in the entire national construction sector, we calculate the labour migration rate in this way to sensitize our analysis to both geographical and subsectorial differences in the labour migration rate. In addition, we focus on the labour migration rate at the subsector-region level, rather than the firm level, to allow not just for direct substitution of apprentices by migrant labour within the firm but also the indirect effect that labour migration can have through increasing wage cost competition on firms’ willingness to invest in training (as discussed in the theory section).

As our main control variable, we use the supply of apprenticeship candidates (STUDENTS_it) calculated as the logarithm of the regional number of students in the first school-based part of the VET education within the relevant fields of study. We also control for firm size (the logarithm of the number of employees, FIRMSIZE_it) because size substantially affects a firm’s propensity to hire an apprentice in a given year. Additionally, we control for trend and period effects that are common to all BaC firms (e.g. business cycle effects) via a set of (time-varying) calendar-year indicators (YEAR_it). Two sets of indicator variables (SEC_i and REG_i) capture time-invariant characteristics of subsectors and regions. Additionally, u_i is unobserved time-invariant characteristics of firms, and e_it is the remaining idiosyncratic errors.

We used logistic regression to model the effect of labour migration on a firm’s odds of being a training firm by having a new apprentice. We use standard logistic regression (with robust standard errors) on the pooled dataset to obtain estimated effects referring to all BaC firms. We use conditional (fixed-effects) logistic regression to estimate the effect of changes in the labour migration level on changes in the odds of having a new apprentice. Fixed-effects regression controls for all (measured and unmeasured) time-invariant characteristics of firms, so IND_i, REG_i, and u_i drop out of the equation above.

We used Poisson regression to model the effect of labour migration on a firm’s training rate. The number of new apprentices was used as a dependent variable, and the number of newly hired employees as an ‘exposure variable’ (the logarithm of the number of newly hired employees was included as an independent variable with the coefficient constrained to be one). This implies that only firms that hired employees in year t are included in the estimation sample used for the Poisson regressions (the training rate is the fraction of all newly hired employees that are apprentices). The conditional (fixed-effects) Poisson regression estimate of the labour migration effect is the estimated effect of changes in the labour migration level on changes in a firm’s training rate, controlled for all (measured and unmeasured) time-invariant characteristics of firms.

We follow convention and report the exponentiated, rather than the raw logit and Poisson, regression coefficients. The exponentiated coefficients should be interpreted as rates (odds ratios or incidence rates), with negative effects between one and zero and positive effects above one.

Results – descriptive data

Table 1 shows the training and labour migration development in the Norwegian BaC industries from 2003 to 2015. There was an absolute increase in the annual number of training firms (from 2301 in 2003 to 2879 in 2015) and in the annual number of new apprentices (from 3672 in 2003 to 4875 in 2015). However, these absolute numbers must be viewed in relation to the expanding economy of the sector indicating a relative decline in both the share of training firms and the training rate. While the share of training firms increased at the start of the period (from 16.2% in 2003 to 18.2% in 2006), it started to decline from 2008 onwards. Despite the substantial increase in the total number of new apprentices, the apprentices’ share of total employment has remained relatively stable with a slight decline.

There is a visible reduction in training levels around the years of the Great Recession, both in terms of the number of training firms (from 3092 in 2008 to 2557 in 2009) and in terms of the number of employees who started training (from 5850 in 2007 to 4142 in 2009). The fact that these reductions also pertain to the percentage of firms that are training firms (from 18.1% in 2007 to 13.4% in 2009) suggests that employers reduced their uptake of apprentices as a response to the downturn. However, when the economy recovered and the number of new apprentices also started to recover, the training rates remained low. At the same time, the number of labour migrants in the sector increased sevenfold, while the labour migration rate relative to the total number of employees increased from 3.5% in 2003 to 17.2% in 2015). While there was a slight absolute decline in the number of labour migrants in 2009, the share of labour migrants has continued to grow throughout the entire period under analysis.

A somewhat similar pattern is visible in Denmark (Table 2), but here the absolute numbers of training firms and apprentices observed in our data have declined during the period. The Great Recession caused an abrupt decline in the number and share of training firms. The number of training firms declined from 4467 in 2007 to 2776 in 2009, causing the share of training firms to decline from 14.1% of all firms in 2007 to only 8.5% in 2009. The share never recovered to pre-crisis levels. In a similar manner, the new apprentices’ relative share of employment dropping from 3.3% in 2007 to 2.2% in 2009 and the training rate has been slow to pick up again.

While these developments could be seen as being part of the general employment decline of the crisis, it is important to note that both the number of labour migrants and the labour migration rate has increased in Denmark during the post-crisis period. While there was also a clear break in the pre-recession growth in labour migration levels in Denmark during the downturn, labour migration was faster to pick up again and surpassed 2008 levels in both absolute and relative terms by 2014. While the crisis had a huge impact, firms were far quicker to start hiring labour migrants than apprentices as the economy improved.

Thus, in both countries, we observe that the share of training firms and the training rate has declined at the national level while there has been a significant increase in labour migration.

Results – estimated models

While the crisis has affected the two countries differently and levels of labour migration have differed, the regression estimates (Tables 3 and 4) suggest – to our surprise – that labour migration has a significant impact on training levels in both countries.

Table 3.

Pooled and Fixed-effects Logit Regressions of Training firms in Norway and Denmark.

	Norway		Denmark
	Pooled logit	FE logit	Pooled logit	FE logit
Labour migration rate (LMR)	0.963***	0.951***	0.981***	0.987***
Labour migration rate (LMR)	[0.96–0.97]	[0.95–0.96]	[0.97–0.99]	[0.98–0.99]
Students (cohort size)	1.117*	1118	1.120**	1.082*
Students (cohort size)	[1.02–1.23]	[0.98–1.27]	[1.03–1.22]	[1.004–1.17]
Firm size	3.250***	6.024***	4.522***	7.258***
Firm size	[3.18–3.33]	[5.71–6.35]	[4.43–4.62]	[6.94–7.59]
Year dummies	Yes	Yes	Yes	Yes
Region dummies	Yes	No	Yes	No
Sector dummies	Yes	No	Yes	No
Observations	244221	87319	397740	123208
Firms	46092	9642	72629	13129
Log likelihood	−71867.8	−31875.1	−86531.9	−42166.7
BIC	144219.4	63920.8	173476.4	84509.3
Pseudo R-squared	0.297	0.098	0.354	0.134

Note: Exponentiated coefficients (odds ratios) from (pooled or fixed-effects) logit regressions. Confidence intervals and statistical tests in the pooled logit regression are based on cluster-robust standard errors.

*p < .05, **p < .01, and ***p < .001.

Table 4.

Pooled and Fixed-effects Poisson Regressions of Training Rate in Norway and Denmark.

	Norway		Denmark
	Pooled Poisson	FE Poison	Pooled Poisson	FE poison
Labour migration rate (LMR)	0.959***	0.976***	0.977***	0.996*
Labour migration rate (LMR)	[0.95–0.96]	[0.97–0.98]	[0.97–0.98]	[0.99–0.99]
Students (cohort size)	1.132**	1.108**	1.055	1.003
Students (cohort size)	[1.04–1.24]	[1.03–1.20]	[0.98–1.13]	[0.99–1.08]
Firm size	1.113***	0.988	1.017	0.957**
Firm size	[1.09–1.13]	[0.95–1.02]	[0.999–1.03]	[0.93–0.98]
Year dummies	Yes	Yes	Yes	Yes
Region dummies	Yes	No	Yes	No
Sector dummies	Yes	No	Yes	No
Observations	142543	61824	184961	91216
Firms	43874	9225	63840	13035
Log likelihood	−101258.7	−48643.2	−117549.8	−63336.4
BIC	202980.2	97451.9	235.487.7	126844.1
Pseudo R-Squared	0.150		0.065

Note: Exponentiated coefficients (incidence risk rates and RRR) from (pooled or FE) Poisson regression. All models also include the logarithm of the within-firm number of newly hired employees as an offset variable (with the coefficient constrained to be one). Confidence intervals and statistical tests in the pooled Poisson regression are based on cluster-robust standard errors.

Significance level: *p < .05, **p < .01, and ***p < .001.

Table 3 presents pooled and fixed-effect regressions of how labour migration affects the odds of a BaC firm being a training firm. The pooled regressions show that, when we control for the regional supply of potential apprentices (the logarithm of the number of BaC students still in school), firm size, business cycle effects (calendar-year dummies) and time-invariant characteristics of industries and regions, the odds of being a training firm are lower when labour migration levels are higher in both Norway and Denmark. In Norway, the estimated odds of a BaC firm having a new apprentice is around 3.5% lower for each percentage point higher the labour migration rate is in the detailed sector-region (OR = 0.963 and p < .001). In Denmark, the corresponding estimate is around 2% lower odds for each percentage point higher the labour migration rate (OR = 0.981 and p < .001).

Fixed-effects regression controlling for unobserved heterogeneity broadly confirms these findings. In Norway, the odds of a BaC firm having a new apprentice are reduced by 5% for each percentage point increase in the labour migration rate (OR = 0.951 and p < .001). In Denmark, the corresponding estimated effect of a one percentage point increase in labour migration is around 1% (OR = 0.987 and p < .001). That is substantially lower but still statistically significant.

The cumulative effect also differs between the two countries. Thus, with labour migration levels having increased by almost 14 percentage points in Norway between 2003 and 2015 (Table 1), our estimates suggest that labour migration alone is responsible for a decline of around 50% in the odds of a Norwegian BaC firm having apprentices in training. Obviously, the actual decline has been more moderate due to countervailing forces, with the economic boom being among the most important. In Denmark, where labour migration levels were around eight percentage points higher in 2015 than in 2003 (Table 2), our estimates suggest that labour migration would have been responsible for only a 10% decline in the odds of a Danish BaC firm having apprentices. While the effect of labour migration on BaC firm’s willingness to take on apprentices is much larger in Norway than Denmark, due to both the greater level of labour migration and the stronger impact of that migration on BaC firm’s reactions, we nonetheless find effects in both countries.

A similar pattern is evident in the estimated effects of labour migration on the training rate (Table 4). In Norway, the pooled regression shows that a one percentage point increase in labour migration was associated with a 4% reduction in the training rate (IR = 0.959 and p < .001). In Denmark, the pooled regression suggests a 2% reduction in the training rate for each percentage point increase in the labour migration level.

For the fixed effect regression, the estimated effect of a one percentage point increase in the labour migration level is a 2% reduction in the training rate in Norway (IR = 0.976 and p < .001). With an increase in the labour migration level of 14 percentage points over the period 2003–2015, this implies that immigration alone was responsible for a 28% decline in the training rate of Norwegian BaC firms during this period. In Denmark, the corresponding effect is only 0.4% per one per cent increase in labour migration level, which amounts to an overall effect of only a 3% decline in the training rate due to the eight-percentage point increase in the labour migration level.

In sum, the data suggest that, in both countries, there is a negative association between the share of labour migrants employed in a specific regional subsector and firms’ uptake of new apprentices, even when we control for the supply of apprenticeship candidates, firm size, and period effects. We discuss the implications of these results and the reasons for variations between the two countries below.

Discussion

While the negative effect is statistically significant in both countries, there are substantial differences in the effect between the two countries. Without the ability to investigate these differences by including country as an independent variable in our analysis, we nonetheless have three potential explanations. First, the inflow of labour migrants has been much stronger in Norway than in Denmark – especially after the financial crisis. Labour migrants have become a much larger and integrated part of the Norwegian construction sector, with Norwegian firms having restructured their labour process to accommodate the skills of labour migrants at the expense of skilled Norwegian workers (Haakestad and Friberg, 2017). By contrast, larger Danish construction firms have retained part of their Danish workforce and their apprentices while relying on subcontractors to deliver cheap migrant labour (Arnholtz, 2021). This also implies that it is mainly the smaller Danish firms that have diminished their recruitment of apprentices due to uncertainties created by both the economic cycle and the increased competition from subcontractors using migrant labour. Second, the Danish VET system has a longer tradition, and there is more support among employers than in Norway. Interviews suggest that many Danish employers feel a strong commitment to passing on the educational skills of their vocation. In addition, there is a stronger reliance on social partner governance of the VET system in Denmark than in Norway, and that may imply a stronger sense of ownership towards the VET system. Third, a more technical reason may be the difference in length of the firm-based apprenticeship between the two countries. Since Danish firms take on apprentices for 3 years (as opposed to 2 years for Norwegian firms), they are less able to respond to changes in the level of labour migration. This may limit the ability of our model to capture employers’ responses in the Danish data.

That said, having both countries in our analysis strengthens the validity of the conclusion drawn from it. The explosive and continuous growth in labour migration into the Norwegian construction sector shows the full potential for the undermining effect this may have on the dual system. However, a single country study of Norway might have been disregarded as an outlier since Norway has experienced an exceptional increase in labour migration since the 2004 EU enlargement. In fact, our initial expectation was that this would be the case, and that we would observe no effect in Denmark, but while the effect was much more moderate this case shows that labour migration may affect dual systems even when levels of labour migration are more moderate.

Conclusion

This article has asked how increasing labour migration affects dual VET systems. We have used the increasing East-West mobility of labour that occurred after the 2004 EU enlargement as a quasi-experimental test of what happens when dual VET systems are exposed to low-skilled labour migration. This mobility has given rise to much debate about potential wage undercutting and erosion of industrial relations institutions in receiving countries, but little research has been devoted to the effect of this labour mobility on skill formation systems, including VET systems. We argue that once industrial relations institutions and their enforcement is put under pressure from labour migration, as documented by the literature, this also opens for companies to use labour migrants as a low-wage replacement for skilled domestic workers. This in turn makes firms less likely to invest in skill formation by taking on apprentices.

Empirically, our results supports this argument by showing that an increase in the level of labour migration within specific regional subsectors of both the Norwegian and Danish construction sector correlates with both a decrease in the odds of a firm becoming a training firm by recruiting at least one new apprentice (most important for small firms) and a decrease in the training rate understood as the number of new apprentices relative to the number of new employees overall (most important for larger firms). These results suggest that dual VET systems in coordinated market economies can be adversely affected by increasing labour migration. Increasing levels of labour migration may disrupt the delicate balance that underpins the dual VET systems without any formal institutional changes having taken place.

These results have implications for our understanding of the relationship between the institutional setup of CMEs and labour migration. While most research has focused on the short-term effect on wage and working conditions as well as medium-term industrial relations institutions, our research suggests that labour migration may also have a long-term on skill formation. This may eventually spill back into the area of industrial relations if a lack of adequate skill formation undermines the viability of the ‘high-skill, high-wage’ oriented wage-setting system. If domestic workers in, for example, the construction sector are no longer adequately trained, maintaining high-wage standards becomes untenable in the face of international competition.

This raises the question to what extent the findings of this article can be generalized both beyond the two countries under study and the construction sector. Further research should investigate this, but our considerations are as follows: The fact that we find a negative impact of labour migration on apprenticeships in both countries under study suggests that we are not dealing with a purely idiosyncratic effect. At the same time, the marked differences in impact in also suggest that institutional differences and differences in levels of migration matter. While other countries with dual VET systems, such as Germany, Austria, and Switzerland may also experience pressure on their industrial relations institutions, they generally have stricter rules on occupational licencing, which may hamper firms’ ability to employ migrant workers as substitutes for skilled domestic workers. Furthermore, as these countries have had higher levels of labour migration over a longer period of time, there may have occurred a gradual adaption that would make the impact impossible to observe. Still, the existence of such countervailing forces does not exclude that we have identified a general phenomenon.

Regarding the selection of the construction, this sector may have particularities that make it more exposed to the impact of labour migration than other sectors. In contract to manufacturing, where production can be relocated to other countries, labour migration is the main source of internationalization in construction. This also implies that industrial relations institutions have been put more under pressure from labour migration in construction than in manufacturing. At the same time, the construction sector has a longer tradition for VET in most CMEs than the service sector, thus making an impact on this type of skill formation more likely. Therefore, we would expect mainly to find the impact of labour migration on dual VET the construction sector. However, to the extent that policy makers are trying to reap the benefits of dual VET in service sector jobs, they should still take account of the impact labour migration may have on skill formation.

It should be noted that there are some limits to the causal interpretation of this papers findings, since there are risks of endogeneity because migrants may move to regions and subsectors where there is an insufficient supply of apprentices. While controlling for the supply of students should counteract some of this problem, a shift share strategy could potentially have solved this problem more satisfactorily (Card, 2001). However, when we tried to implement such a strategy previous settlement patterns failed to predict settlement patterns of new migrants with significant strength.

Despite these limitation and in light of the suggestive qualities of this paper, future research should focus on the interaction between national skill formation systems, international labour mobility, and wage-setting systems (Arnholtz and Wright, 2023; Trampusch, 2019; Wright et al., 2019). Given the increasing levels of labour migration in CMEs, we should study how labour migration affects their VET system. Additionally, research should focus on how labour migration affects the interaction between skill formation and wage-setting systems in CMEs.

Footnotes

Acknowledgements

The authors would like to thank Anna Hagen Tønder, Christian Helms Jørgensen, Torgeir Nyen, Christian Ibsen, and Jon Horgen Friberg for comments on a previous draft of this paper. This research was conducted within the project ‘Moving Trades: Skill formation and the Role of National Vocational Training in Transnational European Labour Markets’.

Declaration of conflicting interests

The author(s) declared no potential conflicts of interest with respect to the research, authorship, and/or publication of this article.

Funding

The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the project is funded by the Norwegian Research Council, FINNUT programme, research grant # 255001.

ORCID iD

Jens Arnholtz

Note

Author biographies

Jens Arnholtz is an Associate Professor at the Employment Relations Research Centre (FAOS), Department of Sociology, University of Copenhagen. His research interests are in the field of employment relations, with a special focus on cross-border labour mobility, posting of workers, Europeanization of national labour markets and power resource theory. His research has been published in journals like ILR Review, Regulation & Governance, British Journal of Industrial Relations, Work, Employment and Society, European Journal of Industrial Relations and Economic and Industrial Democracy. He is the co-editor of a forthcoming book entitled Workers, Power and Society: Power Resource Theory in Contemporary Capitalism (Routledge).

Ståle Østhus is a researcher at the National Institute of Occupational Health (STAMI) in Norway. His research interests are quantitative sociology, the organization of work and labor market dynamics, and social determinants of health. He has published research in journals such as Work, Employment and Society, Social Science and Medicine and Addiction. When the research for this paper was conducted, he worked as a researcher at the Fafo Institute of Labour and Social Research.

References

Aerne

Trampusch

(2023) Including migrant skills in a knowledge economy: the politics of recognition of foreign qualifications, work experience and industry courses. Social Policy and Administration 57(1): 16–33.

Afonso

Devitt

(2016) Comparative political economy and international migration. Socio-Economic Review 14: 591–613.

Arnholtz

(2021) Posted work, enforcement capacity and firm variation: evidence from the Danish construction sector. Economic and Industrial Democracy 42: 1149–1164.

Arnholtz

Ibsen

(2021) Sustaining ‘high road’ employment relations in the Swedish and Danish construction industries. In: Belman

Druker

White

(eds) Work and Labor Relations in the Construction Industry - an International Perspective. New York: Routledge, 183–204.

Arnholtz

Wright

(2023) Labor migration as a source of institutional change: Danish and Australian construction sectors compared. ILR Review 76(3): 532–555.

Arnholtz

Meardi

Oldervoll

(2018) Collective wage bargaining under strain in northern European construction: resisting institutional drift? European Journal of Industrial Relations 24(4): 341–356.

Bosch

Weinkopf

(2013) Transnational labour markets and national wage setting systems in the EU. Industrial Relations Journal 44: 2–19.

Brunello

(2009) The Effect of Economic Downturns on Apprenticeships and Initial Workplace Training: A Review of the Evidence. Paris: OECD.

Busemeyer

Iversen

(2011) Collective skill systems, wage bargaining, and labor market stratification. In: Busemeyer

Trampusch

(eds). The Political Economy of Collective Skill Formation. Oxford: Oxford University Press.

10.

Busemeyer

Trampusch

(2012) The comparative political economy of collective Skill Formation. In: Busemeyer

Trampusch

(eds) The Political Economy of Collective Skill Formation. Oxford: Oxford University Press, 3–31.

11.

Busemeyer

Trampusch

(2019) The politics of vocational training. In: Guile

Unwin

(eds) The Wiley Handbook of Vocational Education and Training. New York: Wiley Blackwell, 137–164.

12.

Card

(2001) Immigrant inflows, native outflows, and the local labor market impacts of higher immigration. Journal of Labor Economics 19: 22–64.

13.

Carstensen

Ibsen

(2019) Three Dimensions of Institutional Contention: Efficiency, Equality and Governance in Danish Vocational Education and Training Reform.

14.

Devitt

(2011) Varieties of capitalism, variation in labour immigration. Journal of Ethnic and Migration Studies 37: 579–596.

15.

Dølvik

Eldring

(2015) Labour mobility in the enlarged single European market. Comparative Social Research 32.

16.

Edo

Rapoport

(2019) Minimum wages and the labor market effects of immigration. Labour Economics 61: 101753.

17.

Felbo-Kolding

Leschke

Spreckelsen T

(2019) A division of labour? Labour market segmentation by region of origin: the case of intra-EU migrants in the UK, Germany and Denmark. Journal of Ethnic and Migration Studies 45: 2820–2843.

18.

Frangi

Zhang

Banerjee

(2021) Constructing inequalities: tenure trajectories of immigrant workers and union strategies in the milan construction sector. British Journal of Industrial Relations 59(2): 474–502.

19.

Fürstenau

Pilz

Gonon

(2014) The dual system of vocational education and training in Germany – what can Be learnt about education for (other) professions. In: Billett

Harteis

Gruber

(eds) International Handbook of Research in Professional and Practice-Based Learning. Dordrecht: Springer Netherlands, 427–460.

20.

Haakestad

Friberg

(2017) Deskilling revisited: labour migration, neo-Taylorism and the degradation of craft work in the Norwegian construction industry. Economic and Industrial Democracy 41: 630–651.

21.

Ibsen

Thelen

(2017) Diverging solidarity: labor strategies in the new knowledge economy. World Politics 69: 409–447.

22.

Jørgensen

(2013) The role and meaning of vocations in the transition from education to work. International Journal of Training Research 11: 166–183.

23.

Jørgensen

(2018) Vocational education and training in the Nordic countries: different systems and common challenges. In: Jørgensen

Olsen

Thunqvist

(eds) Vocational Education in the Nordic Countries - Learning from Diversity. New York: Routledge, 1–28.

24.

Jørgensen

Tønder

(2018) Transitions from vocational education to employment in the Nordic countries. In: Jørgensen

Olsen

Thunqvist

(eds) Vocational Education in the Nordic Countries: Learning from Diversity. New York: Routledge, 29–50.

25.

Kerr

(1977) The balkanization of labor markets. In: Kerr

(ed). Labor Markets and Wage Determination. The Balkanizaiton of Labor Markets and Other Essays. Berkeley: University of California.

26.

Lillie

Wagner

Berntsen

(2014) Posted migration, spaces of exception, and the politics of labour relatioons in the European construction industry. In: Hauptmeier

Vidal

(eds) Comparative Political Economy of Work. New York: Palgrave Macmillan, 312–331.

27.

Meardi

(2012) Social Failures of EU Enlargement : A Case of Workers Voting with Their Feet. New York: Routledge.

28.

Meardi

Martín

Riera

(2012) Constructing uncertainty: unions and migrant labour in construction in Spain and the UK. Journal of Industrial Relations 54(1): 5–21.

29.

Muehlemann

Wolter

Wüest

(2009) Apprenticeship training and the business cycle. Empirical Res Voc Ed Train 1: 173–186.

30.

Nyen

Skålholt

Tønder

(2015) Vocational education and school to work transitions in Norway. In: Bohlinger

Haake

Jørgensen

, et al. (eds) Working and Learning in Times of Uncertainty. Rotterdam: Sense Publishers, 173–182.

31.

Røed

Schøne

(2016) Impact of immigration on inhabitants' educational investments. The Scandinavian Journal of Economics 118: 433–462.

32.

Streeck

(1989a) Skills and the Limits of Neo-Liberalism: The Enterprise of the Future as a Place of Learning. Work, Employment & Society 3: 89–104.

33.

Streeck

(1989b) The territorial organization of interests and the logics of association action: the case of Handwerk organization in West Germany. In: Coleman

Jacek

(eds) Regionalism, Business Interests and Public Policy. London: Sage.

34.

Sweetman

MDJT

Hawthorne

(2015) Occupational regulation and foreign qualification recognition: an overview. Canadian Public Policy/Analisis Politico 41: S1–S13.

35.

Trampusch

(2019) Social partners’ policy reactions to migration in occupational labour markets: the case of the Swiss construction industry. European Journal of Industrial Relations 26: 157–172.

36.

Windzio

Teney

Lenkewitz

(2021) A network analysis of intra-EU migration flows: how regulatory policies, economic inequalities and the network-topology shape the intra-EU migration space. Journal of Ethnic and Migration Studies 47: 951–969.

37.

Wolter

Mühlemann

Schweri

(2006) Why some firms train apprentices and many others do not. German Economic Review 7: 249–264.

38.

Wright

Knox

Constantin

(2019) Using or abusing? Scrutinising employer demand for temporary sponsored skilled migrants in the Australian hospitality industry. Economic and Industrial Democracy 42: 937–959.