A comparative study of eight European countries: how life course events affect female migrant labour market integration under the perspective of welfare and production regimes

Abstract

This article analyses how life course events including education, citizenship, marriage status and having children would affect female migrant labour market integration. This is put in terms of employment status and job quality under the perspective of welfare and production regimes. To investigate different institutional effects, countries representative of these regimes’ typologies, such as the UK and Ireland for liberal market economies/liberal welfare states, Denmark, Finland, Norway and Sweden for coordinated market economies (CME)/social democratic welfare states, alongside Germany and France for CME/conservative welfare states, were employed. European Social Survey data are adopted within a linear probability model, alongside a nested model aggregating life course events. Through these, migrant penalty is analysed regarding labour market outcomes with respect to natives. Among the life course events, education and citizenship acquisition showed a particularly strong impact on migrant penalty, while marriage and having a child did not significantly affect migrant penalty.

Introduction

Female migrants’ labour market integration is a key issue as the secondary market expanded after post-industrialism, requiring a larger female labour force. However, barriers to integration persist. This article, therefore, analyses how life course events would affect female migrant penalty according to institutional backgrounds concerning production and welfare regimes in the United Kingdom, Ireland, Germany, France, Denmark, Finland, Norway and Sweden. The eight countries are classified differently under three types based on welfare and production regimes. The UK and Ireland represent liberal welfare states and liberal market economies (LME), while Denmark, Finland, Norway and Sweden belong to coordinated market economies (CME) and social democratic welfare states. Meanwhile, although Germany and France belong to CME in the production regime like Scandinavian countries, they represent conservative welfare states in welfare regime. Therefore, the characteristics of their labour market environment for female workforces would necessarily be differentiated according to these three types predicated upon the regimes.

Regarding migrant integration, the subject countries except Germany and France have been found to secure higher levels of integration in labour market outcomes. This is since labour market flexibility and less restricted accessibility to social rights endows migrants with favourable integration conditions in host societies. Specifically, labour market flexibility is the critical prerequisite in LMEs which rely on market demand and supply, so that the UK and Ireland secure this flexibility to a greater extent than other countries. Contrastingly, although the CME/social democratic welfare states are less flexible than LMEs, active labour market policies (ALMP) alongside universal welfare benefits produce flexicurity in the labour market. These institutional backgrounds could give migrants better labour market integration conditions (Ballarino & Panichella, 2013; Kogan, 2007). Nevertheless, after post-industrialism, practising a form of flexibility in CME/conservative welfare states continues to witness. Germany absorbed the impact of liberalisation alongside labour market dualization by institutionalising insiders and outsiders, whereas France produced “higher protection” which means repromoting employment protection while simultaneously expanding activation policies (Ferragina & Filetti, 2022).

Accordingly, beyond better conditions and outcomes regarding immigrant integration alongside the level of flexibility in the labour market, this article focuses on the difference between the regimes in the eight countries by incorporating life course events. This is since these diverse labour market conditions could differently affect female migrants’ labour market outcomes according to life course events. Hence, this article discusses how the association between individual life course events and the different institutions following the distinction of LME/liberal welfare states, CME/social democratic welfare states and CME/conservative welfare states, affects female migrant penalty regarding occupational integration. Previous studies mostly addressed male migrants’ labour market integration, including through immigrant policies which have been regarded as less significant than the welfare and production policy arenas (Czaika & Haas, 2013).

Moreover, female migrants have not been investigated regarding complementarity between both dominant regimes, and females have possible factors which affect their labour market outcomes based on maternity or care issues. This study therefore meaningfully reveals how welfare and production regimes are simultaneously coupled with life course events, as well as how this particularly affects migrant females. The institutional influence upon female migrant labour market integration was analysed through “migrant penalty,” indicating employment status and job quality differences between natives and migrants. Accordingly, the impact of each event on female migrant penalty with respect to female natives was uncovered through multivariate analysis. A linear probability model alongside a nested model is employed to aggregate individual life course events, namely education, citizenship acquisition, marriage and having a child.

The article is organised as follows. First, existing literatures relating to the core themes of interest will be presented by covering research into female labour market penalties, including that with explicit references to welfare and production regimes. The discussion continues with regards to female migrant penalties and the suggested importance of life course events to these penalties in labour market outcomes. Based on this, the hypotheses and methodology are considered, detailing data and model specifications. Through presenting model results, it is suggested that life course events, insofar as education and citizenship are concerned, exert a significant effect upon female migrant penalty according to the three typologies.

Previous literature

Female labour market penalty based on welfare and production regimes

Estevez-Abe (2005, 2006) revealed three major threats when women try to build professional skills and career paths: first, giving birth and any contingencies regarding family members; second, voluntary career breaks due to childcare and difficulties returning to the workplace; third, opportunity deprivation in terms of building particular skills and the degradation of the value of possessed skill due to career breaks. This discussion, therefore, implied how females are particularly constrained in the labour market due to family care work, and according to life course events. Furthermore, even though the demand for the female workforce has increased after post-industrialism, the gender gap regarding labour market outcomes in skill formation, wages and similar are significantly present in association with new social risks (Esping-Andersen, 2002; Lewis, Campbell, & Huerta, 2008; Timonen, 2004).

New social risk emerged from post-industrial society against old social risks which were typically treated by welfare states to support main actors or the male breadwinner model during the industrial era. Accordingly, the transition to post-industrial society massively demanded non-major workforces like females, migrants or obsolete skilled labourers who received little political attention in the industrial era, to fill expanded service industries. Therefore, alongside secondary labour markets, precarious work including low pay, short-term contracts increased significantly and, in turn, labour decommodification is hardly secured under these non-standard employment forms (Palier & Thelen, 2010; Rubery et al., 2018). In order to treat industrial demand and new social risks simultaneously, welfare states had to develop new policies, like work–family and activation policies, based on a lack of carers with females consequently taking on care responsibilities and remaining on career breaks (Bonoli, 2005; Taylor-Gooby, 2004).

In this regard, previous research employed work–family reconciliation policy based on the welfare regime and analysed the causality between the level of family policy (mostly expenditures), and labour market outcomes. These studies’ results commonly pointed towards higher effects of (public) day care services, but lesser impacts from family allowance, on female labour market participation rates (Ray, Gornick, & Schmitt, 2010; Stadelmann, 2008; Windebank, 2017). On the other hand, Kang (2020) examined the moderated effect of work–family reconciliation policy on the gender wage gap in the welfare and production regimes. This was in order to see the labour market environment in a broader perspective through both regimes. Thus, the author adopted a combined typology between the two regimes, as Soskice (2005) already introduced, to provide a more specific conceptualisation useful for analysing gender difference in the labour market.

This resulted in three categories combining pre-existing categories from production and welfare regimes previously defined by Esping-Andersen (1990) and Hall and Soskice (2001), respectively. Kang subsequently presented LMEs/liberal welfare states, CMEs/conservative welfare states and CME/social democratic welfare states categories. This new typology could certainly be regarded as overcoming limitations of the arguably confined explanations encountered that refer to only one regime to analyse labour market outcomes. This has been witnessed in existing scholarship alongside discussions of institutional complementarity between the regimes (Schroder, 2008).

Kang (2020) analysed the wage gap through the median earnings of men and women among full-time employees and the self-employed from 1981 to 2015 in 13 countries (OECD). Maternity and paternity leave and day care generosity, were included as independent variables to convey work–family reconciliation policy. The results showed that, in CME/social democratic welfare states, the highest positive effect lessening the wage gap stemmed from day-care but for other groups statistical significance was absent. Parental leave showed positive statistical significance only in CME/conservative welfare states. Nevertheless, in general, the gender wage gap was highest in CME/conservative welfare states, followed by LME/liberal welfare states, then CME/social democratic states (Kang, 2020).

However, paternity leave’s positive effect in CME/conservative welfare state can possibly be explained with the German case. Here, 60 per cent of formal salary is subsidised for paternity leave and the female workforce who gave birth is more likely to hold part-time positions in Germany (Asinta, 2018); the lesser gender wage gap could thus be attributed to this specific policy condition. Nevertheless, in accordance with the general highest gender wage gap in CME/conservative welfare states, the results well-reflected expanded labour market dualization after the Harz reform and the increase in mini-jobs mostly taken by posted workers and females in Germany (Rubery et al., 2018).

Furthermore, these results are in-keeping with how the male breadwinner model, concentrating male employment in the industrial era, placed the actual beneficiary from the welfare system based on the contribution of employment status. Females, by extension, could be indirectly decommodified through the dependence on male counterparts’ contribution in the household. This system is mostly highlighted in conservative welfare states (Esping-Andersen, 1999, 2002; Hobson, 1990; Hobson & Lindholm, 1997), and the welfare system is complementarily supported by the production system in which firm-specific skills have been developed (Hall & Soskice, 2001). Thus, the preference for accumulated skills in conservative welfare states is likely to make employers avoid female labourers, given potential career breaks due to marriage, giving birth or providing care services (Estevez-Abe, 2005, 2006; Soskice, 2005).

This is complemented by the formation of reconciliation policy regarding how family allowance has been developed more than care services in CME/conservative welfare states (Esping-Andersen, 2002; Ray, Gornick, & Schmitt, 2010). Consequently, females are more likely to remain in the household according to the complemented system of welfare and production. Hence, not only wage gaps but also female labour market participation rates have consistently been reported to be low compared to LME/liberal welfare states and CME/social democratic welfare states.

Moreover, according to OECD (2021) statistics from 2000 to 2019, this trend seems similarly present in female migrants’ case as well. Compared to the figures between this article’s eight subject countries, female immigrants in CME/conservative welfare states showed approximately 6–7 percentage points (pp) lower employment rate. To be specific, LME/liberal welfares states revealed 58 per cent female migrant employment status based on the 20 years’ average employment rates within the UK (58.5 per cent) and Ireland (57.4 per cent). Likewise, CME/social democratic welfare states’ average also stood at 59 per cent, when including Denmark (57.8 per cent), Finland (54.3 per cent), Norway (64.7 per cent) and Sweden (58.3 per cent). By contrast, CME/conservative welfare states’ average employment was uncovered at a lower rate of 52.5 per cent, with this captured by France (48.5 per cent) and Germany (55.6 per cent).

Female migrant penalty and life course events

Following previous studies regarding the gender gap in labour market outcomes and the brief statistics concerning female migrant employability above, this study questions how life course events would affect migrant penalty differently with respect to female natives according to the three typologies. Furthermore, whether LME/liberal welfare states and CME/social democratic states where relatively favourable institutional background were found for female labour market outcomes could entail lower migrant penalty than CME/conservative welfare states is asked as well. This is since migration is one process in which a family unit or individuals purposively make a strategy in order to migrate and, ultimately, integrate into destination countries. Thus, females could be more penalised as tied movers since the family migration strategy could involve voluntarily enduring a longer inactivity or low job quality status, even with greater possibilities to access employability or vocational training (Ballarino & Panichella, 2017).

Apart from this voluntary inactivity status, education, citizenship acquisition, marriage status and having a child could involuntarily exert varied effects upon female migrants’ labour market outcomes in relation to production and welfare regimes (Kogan, 2007; Sainsbury, 2006, 2012; Stier, Lewin-Epstein, & Braun, 2001). To elaborate, education and citizenship acquirement are closely related to employers’ preferences because visa support to migrants for obtaining work permits is clearly beneficial for high-skilled migrants. This can be found in the UK where employers’ sponsorship is required to obtain Tier 2 visas for skilled positions (UK Government, 2022). Similarly, the EU Council (2009) also specified favourable visa conditions to streamline residence and work permit processes for highly qualified third-country immigrants. When considering human capital issues across countries, selectively favourable border controls for highly qualified migrants meets the innovative area’s labour demand, so that stratified citizenship in association with education and high skills has expanded in receiving countries (Ellermann, 2020; Paul, 2018).

In LMEs, general skills are preferred to support main industries, centred on high and low service sectors, and migrants could resultantly be more affected by their educational level. Certificates of education in LMEs could be a central measure to prove their skills in the service industries due to general skills which do not require accumulation in practice over time like in CMEs (Hall & Soskice, 2001). Accordingly, employers prefer hiring highly educated migrants when reckoning with better return which can strengthen their business competitiveness in the skilled service industries. However, in low-skilled service sectors, there is no greater reason for employers to hire educated migrants unless they provide greater competitiveness, for instance, by accepting lower wages. In CMEs, meanwhile, main industries prefer accumulated skills indicating firm and industrial specific skills which can be obtained in apprenticeship and vocational training system, respectively (Hall & Soskice, 2001). Thus, migrants would not be critically affected by education levels compared to in LME since educational certificates do not secure adequate proof of their specific skill possession which would not necessarily be transferable from the education they may have.

On the other hand, the impact of life course events such as marriage and having a child on labour market outcomes could vary in association with the welfare regime. After marriage, household livelihood is considered to be accomplished between partners, commonly by using time and goods. Thus, the family strategizes how to produce commodities together by reckoning with (dis)advantages of labour market entrance. This could be determined in association with housework costs, including childcare. Consequently, female partners who typically take charge of housework could be discouraged from improving human capital and labour participation in the marriage relationship compared to male partners, or single women who can independently decide human capital investment for the sake of employability (Becker, 1965; Borck, 2014; Chen et al., 2014; Grossbard-Schechtman & Neuman, 1988; van Rensburg, Claassen, & Fourie, 2019). Hence, work–family reconciliation policies which help females choose participation in the labour market could be equally helpful for female migrants. However, policies have been developed differently according to the three welfare systems’ traits, so female migrant penalty levels are also expected to vary.

Liberal welfare states have less developed public care services along with means-tested accessibility to welfare services; while Scandinavian countries, centred on universalism, provide generous public care services as well as parental leave (Esping-Andersen, 2002). Therefore, even though flexible employment contracts which remove fire costs and social insurance contributions for employers have given preferable situations for migrants to be employed in liberal welfare states (Kogan, 2007), female migrants who confront the life course events could be penalised more than in social democratic welfare states.

In terms of CME/conservative welfare states, and contrastingly to the liberal welfare state, employers’ contributions of social insurances for mini-jobs are required in Germany and public care services also are less developed compared to Scandinavian countries through contribution-based benefit welfare systems (Esping-Andersen, 2002; Sainsbury, 2012). Thus, female migrants with a child could experience similar or more migrant penalty than in LME/liberal welfare states because migrants have less social capital and information than natives to mobilise private care services. However, migrants in Sweden have easy access to public care services at similar rates to natives (Sainsbury, 2019), and this could be coupled with highly developed ALMP within Scandinavian countries; this being connected with vocational training accessible to all irrespective of citizenship status. Subsequently, marriage and having a child would not seriously affect migrants’ employability and job quality in Scandinavian countries.

Hypotheses

The hypotheses concern: firstly, individual characteristics regarding life course events; secondly, structural factors in terms of institutional differences under the three typologies pertaining LME/liberal welfares states, CME/conservative welfare states and CME/social democratic welfare states. To elaborate, different life course conditions would affect female migrant penalty with respect to natives in employment status and job quality, and the level of penalty could vary according to specific institutional backgrounds. Therefore, hypothesis 1 refers to the direct and composition contribution of each event by incrementally adding the events into a nested model by country. Hypothesis 2 reflects the last full model, considering the eight countries’ migrant penalty simultaneously under the regime typologies. The hypotheses are as follows.

Hypothesis 1. Each life course event’s contribution to migrant penalty in employment status and job quality could vary according to each country’s institutional background.

Education, the contribution of education on migrant penalty would be lowest in Scandinavian countries since their labour market prefers industrial-specific skills which do not necessarily correspond to level of education and could be built through much accessible vocational training systems compared to LME/liberal welfare states and CME/conservative welfare states, respectively.

Citizenship, as citizenship acquisition is based on having a longer residence period in host countries, citizenship status would reduce migration penalty in CMEs regardless of conservative or social democratic welfare states. This is since, unlike general skills, more time is needed to obtain accumulated skills (industrial or firm-specific) to improve employability and job quality. Thus, migrants who received citizenship would be less penalised than newcomers who would not yet have acquired citizenship under both CME types.

Marriage, marriage status would not significantly affect migrant penalty. This is because third-country migrants are expected to mostly take service positions in which workforce demands and flexibility in the labour market have expanded since post-industrialism. Therefore, there would be no expected penalisation upon migrants from marriage status in the three systems.

Child, migrants having a child in CME/social democratic welfare states would not be penalised, particularly in employability, since care service accessibility is similar to natives. However, migrants from the other two systems would be penalised due to lower care service accessibility and less social capital.

Hypothesis 2. When controlling all life course events, migrant penalty with respect to natives could be varied according to the three typologies, driven by welfare and production regimes.

Based on the welfare regime, social democratic welfare states would show higher migrant penalty with respect to natives in employment status than in liberal and conservative welfare states. This is because the lenient universal welfare benefits afford more room for migrants to remain in an unemployed status. This corresponds to this system’s characteristic of securing high individual autonomy since reliance on family and market is limited alongside high decommodification levels (Ferragina & Seeleib-Kaiser, 2011). However, migrants in liberal and conservative welfare states have less possibility to remain unemployed because they need to rely on the market more given means-tested and contributory welfare systems, respectively.

Based on the production regime, LME’ migrants could be less penalised with respect to natives than those in the other two systems regarding job quality. This is since, with general skills, it can be relatively easier to apply skills migrants already hold in contrast to specific skills. Furthermore, although migrants have higher educational backgrounds or certain skills, they would not transfer easily into CME labour markets where mainstream industries require accumulated skills obtained mostly in destination countries.

Methodology

Data and variables

To investigate the hypotheses, ESS cumulated data was employed. This includes the eight rounds held at 2 years’ interval from 2002 to 2016 (European Social Survey, 2018). Migrants were defined as those who answered that they were not born in the surveyed country (UK, Ireland, Denmark, Finland, Norway, Sweden, Germany and France), and second-generation migrants as those whose father or mother was not born in the case country. This is since their integration particularly from third country in the labour market was investigated to still be disadvantageous or little difference compared to first generations (see Ballarino & Panichella, 2013; Cheung, 2014; Heath & Cheung, 2007; Silberman, Alba, & Fournier, 2007).

Accordingly, immigrants from developed countries were excluded in the analysis since their labour market status is hardly penalised with respect to natives (see Ballarino & Panichella, 2013, 2017; Khattab & Johnston, 2013, 2015; Kogan, 2007; Modood & Khattab, 2016). The developed countries included the EU-15 and EFTA (Norway, Iceland, Switzerland, Lichtenstein), alongside North America and Oceania (US, Canada, Australia and New Zealand). With this definition, the independent variable for migrant status coded migrants as 1, whereas respondents who were born in the country, both themselves and their parents, were coded 0. Age range extended from 25 to 60 and the total sample size across the eight countries was 34,615 individuals which included 2,738 migrants.

Regarding employment status, the question concerning “what respondents’ main activity during the last 7 days was,” was used to define employed and unemployed status. Considering the questionnaire’s specification is confined to the last 7 days’ main activity, “looking after children or others” is defined as an unemployed status rather than inactivity since respondents may have the potential to find jobs or get back to work sooner or later. This reflects precarious labour market contracts offered in service sectors where low-skilled female workforces are likely to be hired. Moreover, in this way, turnover period can be better considered since there is the possibility that skilled females currently looking after children could also choose this answer, yet they could return to work in a place similar to where they used to be after the 7 days. Therefore, this way can less be biased while securing more statistical power by not discarding a valid sample under the expanded fixed-term contracts regardless of skill level in the labour market.

Accordingly, the employed who recently participated in paid work were coded as 1, while the unemployed were coded as 0. In line with this, an unconditional model was adopted as a better measurement of job quality. This was so that both the job quality of the employed and the unemployed were investigated. Since both categories may have experienced a certain level of skilled job position previously.

For the unconditional model, job quality was divided into high- and low-skilled jobs through ISCO88 and ISCO08 in ESS rounds 1–5 and 6–8, respectively. Specifically, ISCO88 skilled jobs include managers, professionals and technicians, coded from 1000 to 5168, but excluding domestic and manual service workers (5000–5160). Unskilled jobs include manufacturing labourers and elementary work, from 5169 to 9330. ISCO08 alternatively codes skilled jobs from 1000 to 4419, comprising managers, professionals and high technicians, including some public officers (5411–5413). Furthermore, precision-instrument makers and repairers, musical instrument, jewellery or precious metal workers (from 7311 to 7313) were classified as skilled workers. Meanwhile, manual service sales workers and elementary labourers, which are coded from 5000 to 5410 and between 5419 and 9700, respectively, were defined as low-skilled. Thus, by unifying the differently coded job qualities from ESS survey rounds, skilled jobs were coded as 1 and low-skilled jobs as 0.

Control variables represented the four life course events. The first condition is education, based on the ISCED frame. Less than lower secondary education (ISCED 0–1) and lower secondary education completed (ISCED 2) were classified as “lower secondary or less” and coded 1. Upper secondary education completed (ISCED 3) and post-secondary non-tertiary education completed (ISCED 4) were labelled “upper secondary education” and coded 2. The last level is tertiary education completed (ISCED 5–6), coded 3. For citizenship, demographic information like “are you a citizen of [country]?” is collected in all rounds (non-citizenship coded 1, citizenship coded 0). Marriage status was divided as married (1) and single (0), the latter including divorcees and widows. Lastly, having children was measured regarding whether the respondent lives with children in the household (living with children was coded 1, not as 0) (see Supplementary Material S1 for descriptive statistics for the variables).

Nested linear probability models

This article adopted a nested linear probability model with robust standard errors to compare different life course events’ impacts upon female migrant penalty. The models present the probability of female migrants being employed and entering skilled jobs with respect to female natives in each country. Reduced models accumulate one additional variable which represents another life course event, extending sequentially from the basic model until the full, which included all life course events of interest.

Therefore, a set of nested regressions are presented following these five models which sequentially add each event variable and, in turn, it is conceivable to see how the terms affect the outcome variables (Breen, Kalson, & Holm, 2018). Consequently, the first basic model shows the probability of migrant penalty with respect to natives regarding employability and job quality without consideration of the events. Thus, the direct contribution of each event on migrant penalty conveyed via Model 2 can be revealed through comparison with the basic model, so that Model 2 is conducted separately four times according to each event.

Afterwards, the reduced models are aggregated under the full model following the order of education, citizenship, marriage and having a child. The third reduced model adds citizenship status alongside education, uncovering the impact of citizenship while controlling education. The fourth reduced model introduces marriage status while controlling the effects of education and citizenship. The full model reveals the final migrant penalty which is controlled by all life course event terms. There is an additional term used to keep the influence of specific times under control. This is because the ESS rounds reflect eight different survey years. Subsequently, $ {\beta}_i\mathrm{YR} $ is included in every model, but this term’s results are not included in the regression tables below.

Nested model specification

The variables and five models are specified below.

Dependent variables: employment status and job quality. Independent variable: migrant status ( $ \mathrm{FMIG} $ ). Control variables: four life course events including education ( $ \mathrm{EDU} $ ), citizenship ( $ \mathrm{CTZ} $ ), marriage ( $ \mathrm{MRRG} $ ) and child ( $ \mathrm{CHD} $ ) plus the year term ( $ \mathrm{YR} $ ).

• • Model 1. Basic model of female migrants.

$$ E\left(X|Y\right)=\alpha +{\beta}_1\mathrm{FMIG}+{\beta}_2\mathrm{YR}+\varepsilon . $$

• • Model 2. Reduced model including each event separately.

$$ E\left(X|Y\right)=\alpha +{\beta}_1\mathrm{FMIG}+{\beta}_2\mathrm{EDU}\;\left({\beta}_2\mathrm{CTZ}/{\beta}_2\mathrm{MRRG}/{\beta}_2\mathrm{CHD}\right)+{\beta}_3\mathrm{YR}+\varepsilon . $$

• • Model 3. Reduced model including education and citizenship.

$$ E\left(X|Y\right)=\alpha +{\beta}_1\mathrm{FMIG}+{\beta}_2\mathrm{EDU}+{\beta}_3\mathrm{CTZ}+{\beta}_4\mathrm{YR}+\varepsilon . $$

• • Model 4. Reduced model including education, citizenship and marriage.

$$ E\left(X|Y\right)=\alpha +{\beta}_1\mathrm{FMIG}+{\beta}_2\mathrm{EDU}+{\beta}_3\mathrm{CTZ}+{\beta}_4\mathrm{MRRG}+{\beta}_5\mathrm{YR}+\varepsilon . $$

• • Model 5. Full model including education, citizenship, marriage and having a child.

$$ E\left(X|Y\right)=\alpha +{\beta}_1\mathrm{FMIG}+{\beta}_2\mathrm{EDU}+{\beta}_3\mathrm{CTZ}+{\beta}_4\mathrm{MRRG}+{\beta}_5\mathrm{CHD}+{\beta}_6\mathrm{YR}+\varepsilon . $$

Results

Migrant penalty regarding employment status

The results of the basic model are presented alongside the second model (Table 1). Firstly, the basic model indicated the migrant penalty regarding employment status with respect to natives without controlling any life course event, and all countries were statistically significant except Ireland. This figure for each country was later used to analyse the level of direct contribution from events by checking differences between the basic and second models, visualised in Figure 1.

Table 1. Basic model and direct contribution of life course event in employment status.

Note. CI in parentheses.

^*** p < 0.01;

^** p < 0.05.

Figure 1. Direct impact of each life course event on migrant penalty in employment.

The basic model results ranged from −1 to −21 pp, indicating the smallest migrant penalty in Ireland and the highest in Denmark and Finland. The second highest migrant penalty with respect to natives was seen in France, at −19 pp, followed by Germany (−17 pp), Sweden (−15 pp), the UK (−14 pp) and Norway (−11 pp). When adding education into Model 2, the migrant penalty increased in the UK and Ireland by aggregating a migrant penalty of around −2 and −5 pp, while there was no difference in Norway and Finland. However, a reduced penalty between the basic model and Model 2 was found in Germany, France (3 pp) and Denmark, Sweden (1 pp), respectively (Figure 1).

Concerning citizenship, Figure 1 and Table 1 demonstrate a positive contribution reducing migrant penalty in employment in CME/conservative welfare states and CME/social democratic welfare states. The two strongest positive contributions from citizenship to reduce migrant penalty were found in Finland and France, the difference from basic model standing at 17 and 11 pp (Figure 1), although both were non-statistically significant. However, Germany and Denmark showed a respective 9 and 6 pp decreased migrant penalty, while Sweden and Norway also reduced the penalty by 2 and 1 pp, respectively. Contrastingly, there was no effect and negative contribution of citizenship on migrant penalty in Ireland (non-statistically significant) and the UK (−1 pp). Hence, these results showed the higher positive direction of citizenship contribution in reducing migrant penalty around employment with respect to natives in CMEs, regardless of welfare regimes.

For marriage and having children, none of the countries presented any great difference from the basic model when adding these two life course events. The figures returned were similar to those of the basic model in each country (Table 1). This means that these life course events would not have great effect on migrant penalty in employment, although there were subtle differences of a 1 pp negative (UK, DK, NO and SE) and positive (IE and DE) contribution on migrant penalty. This trend somewhat coincided with the effect of having a child, indicating a positive direction reducing migrant penalty in UK, IE and DE, while a negative 1 pp contribution increasing migrant penalty was seen in DK and SE.

However, in terms of the total female workforce including natives and migrants, there was a significant negative effect of having a child on employment in the UK, Ireland and Germany at −8, −15 and −10 pp, respectively (see Supplementary Material S2 for Model 5’s results), which is opposite migrant penalty results. This is a very interesting result since the having a child condition affects natives and migrants differently.

The negative impact of the child condition on employment for female workforces reflect less developed public care services in the three countries which induce female natives to remain carers or temporarily unemployed (Esping-Andersen, 2002). Therefore, there could be more opportunity for migrants to occupy these redundant positions, or the less chance of having the right of decommodification in these countries could lead migrants who are in the more urgent situation alongside a child to work to support the household. In this regard, the expansion of work benefit in Germany cannot be ignored since, unless working in low-income jobs, there is no eligibility for social assistance benefit. Thus, around 1.3 million were reported to receive household means-tested benefits as low-paid workers in 2015 (Knuth, 2015; Rubery et al., 2018).

The results in Table 2 indicated how the composition contribution between life course events on migrant penalty significantly varied according to the three typologies. The UK and Ireland showed no greater change after education in Models 3–5. However, for Scandinavian countries, there was no contribution from Model 2 (education) while greater effect from citizenship was found in Model 3 and continued until the full model. With Germany and France, incremental positive effects from education to citizenship continued to last into full model, although France could not secure statistical significance after introducing citizenship. Therefore, through this reduced model regarding employment, it can be confirmed that the strong determinants among life course events on migrant penalty are education in LME/liberal countries and citizenship in CME/social democratic welfare states. Meanwhile, the significant positive composition effect from education and citizenship was seen in CME/conservative welfare states, meaningfully in Germany with statistical significance.

Table 2. Composition contribution from basic (model 1) to full model (model 5) in employment status.

Note. CI in parentheses.

^*** p < 0.01;

^** p < 0.05.

Migrant penalty regarding job quality

For job quality, a significant negative impact of education on migrant penalty was found in LME/liberal welfare states. The difference between the basic model and Model 2 was −8 and −9 pp in the UK and Ireland, respectively, while the education contribution was less significant in CME countries in general. Nevertheless, there were positive educational effects in Germany, France, Denmark (4 pp) and Sweden (3 pp), reducing migrant penalty. No education effect was observed in Finland and Norway (Figure 2). When compared to employment results, no outstanding division between the three systems existed regarding the impact of citizenship status on migrant penalty concerning job quality. This is since Denmark and France only showed significant positive impacts from citizenship decreasing migrant penalty by up to 9 and 5 pp from the basic model, respectively. Ireland and Germany also showed decreased penalty through a 2 and 1 pp difference, while the UK (non-statistical significance) and Norway indicated a penalty increase of 1 and 3 pp. Meanwhile, no change was found in Sweden regarding the direct contribution of citizenship and Finland could not secure statistical significance (Table 3 and Figure 2).

Figure 2. Direct impact of each life course event on migrant penalty in job quality.

Table 3. Basic model and direct contribution of life course event in job quality.

Note. CI in parentheses.

^*** p < 0.01;

^** p < 0.05.

In this regard, Model 3’s results which indicated the composition contribution between education and citizenship needs to be carefully interpreted alongside the direct contribution of citizenship (Table 4). Specifically, a clear positive composition effect between education and citizenship which reduced migrant penalty regardless of the different three systems existed, apart from in Norway. In Germany, France, Denmark and Sweden, the positive direct effect of education on migrant penalty was intensified by including citizenship. Accordingly, the penalty gradually decreased by adding education and citizenship from the basic model although France and Denmark could not secure statistical significance in Model 3.

Table 4. Composition contribution from basic (model 1) to full model (model 5) in job quality.

Note. CI in parentheses.

^*** p < 0.01;

^** p < 0.05.

Between Finland and Norway, the direction of composition effect was opposite; a 2 pp negative effect increasing migrant penalty was found in Norway, and 2 pp positive effect decreasing penalty was revealed in Finland. Therefore, regarding job quality, citizenship was not a strong measure impacting upon migrant penalty in both CME/conservative welfare states and CME/social democratic welfare states. However, the composition of education and citizenship significantly affected migrant penalty in relation to accumulated skills which can be acquired via education and longer residence in the destination countries.

Although the UK and Ireland also showed decreased migrant penalty in the composition impact between education and citizenship, it was clearly an opposite pattern compared to the other two systems. Here, a substantially increased migrant penalty from education in the direct contribution was seen at −8 pp (UK) and −27 pp (IE). The penalty was offset thanks to the composition contribution along with citizenship, decreasing the migrant penalty figure to −7 and −22 pp, respectively (see Models 2 and 3 in Table 4). Therefore, it can be highlighted how in LME/liberal welfare countries, there is also a positive composition impact regarding education and citizenship on migrant penalty for job quality, but in a way which reduced the negative direct impact of education.

There was still no greater change in the direct contribution on migrant penalty around job quality like that of employment for marriage and having a child. However, a 1 pp increased penalty from marriage could be found in Scandinavian countries except Denmark, while having a child decreased the migrant penalty by 1 pp from the basic model in the UK and Ireland (Figure 2). Resultantly, there could be a positive selection of migrants in terms of job quality in liberal states when considering having a child’s coefficient in the full model (see Supplementary Material S3 concerning Model 5 for the UK and IE). This is since the figures showed a significant penalty on female workforces, standing at −7 and −3 pp in the UK and Ireland, respectively. Accordingly, although an increased penalty upon female job quality attributable to having a child in LME/liberal welfare states was proved, the impact of having a child on migrant penalty was opposite; reducing the penalty by 1 pp. Therefore, positive selection among the migrant sample could be assumed to indicate that migrants who have a child could have higher skills.

Composition contributions of life course events in the nested model

The nested models’ results indicating the composite contribution of the four life course events are visualised in Figure 3, which is based on Table 2 (employment) and Table 4 (job quality). The significant change of migrant penalty from the basic to the full model was found in the three institutional backgrounds, but with different life course events and varied strengths and directions. According to Figure 3, greater migrant penalty occurred after the basic model in the UK and Ireland, indicating the direct impact of education in a negative direction and its sequential effects up to full model. Contrastingly, a lesser degree of positive education impact reducing the basic model’s migrant penalty (3–4 pp in both countries) can be found in Germany and France, but this effect did not last to full model once adding citizenship. In Scandinavian countries, meanwhile, positive citizenship effects are the strongest predictors of migrant penalty in employment and job quality as seen in Figure 3 where migrant penalty is stable until the full model from Model 3 where citizenship was introduced. Therefore, the hypotheses regarding the lower impact of education in CME/social democratic welfare states and positive impact from citizenship on migrant penalty in CME/social democratic welfare state and CME/conservative welfare states, was confirmed based on these results.

Figure 3. The result of the nested model regarding migrant penalty. Model 1, basic model without any life course events; model 2, education; model 3, education and citizenship; model 4, education, citizenship and marriage; model 5, full model including education, citizenship, marriage and child. Ireland, France and Finland could not secure statistical significance from model 3 to model 5 in employment. In terms of job quality, France and Denmark could not secure statistical significance from model 3 to model 5.

Interesting additional findings regarding education in relation to regime typologies include how the direction of education is different between LME/liberal welfare states and CME/conservative welfare states. This means higher educational levels positively reduces migrant penalty in CME/conservative welfare states, but not in LME/liberal welfare states. This is somewhat unexpected considering the strong restrictiveness of professional (business) service regulations in France and Germany which are at least twice higher than in the UK and Ireland for those crossing from one country to another (European Commission, 2017).

Accordingly, this result could be interpreted in three ways: first, racialisation in the UK and Ireland’s labour markets could be severer than Germany and France’s considering that the migrant subjects of this study are third-country migrants. This could be supported by recent migrant labour market experiments examining call-backs for job interviews indicated that third-country migrants applying to job markets across European countries are less discriminated in Germany than the UK (Di Stasio & Lancee, 2020). This is also in line with previous correspondence tests (Zschirnt & Ruedin, 2016).

Second, in relation with migrant policy, citizenship restrictions like family unification were investigated as more rigorous in the UK than France and Germany (Sainsbury, 2012). This could lead female migrants having high education backgrounds in the UK to be more limited in labour market participation in tied mover cases. Third, in association with recent trends of expanding innovative industries, although France and Germany have typically preferred firm-specific skills which impose narrow opportunities for migrants to access than natives, new innovative industries such as ICT workers have expanded notably since 2000s with simultaneously increased demand for skilled migrants in these countries. Therefore, higher education positively affects (decreases) migrant penalty in the countries, which coincides with the migrant laws enacted in 2005 (Germany) and 2006 (France) to promote high-skilled migrants (Murphy, 2006; OECD, 2013).

Meanwhile, the hypothesis regarding how marriage status would not significantly impact upon migrant penalty can be confirmed. This is because there is no great change in both measures; employment and job quality. The last life course event was having a child, which could be explained through the perspective of work–family reconciliation policies under the welfare regime. In this regard, it was expected that there might be more penalty upon female migrants from having a child in LME/liberal welfare states and CME/conservative welfare states. However, surprisingly, there was a 1 pp decreased penalty in both measures for the UK and Ireland, and in employment for Germany (Figure 2). This is interesting since having a child is a critical condition for female labour market status, especially for negative effects. Consequently, this study expected more penalty in liberal and conservative welfare countries given less developed public care services. Therefore, this hypothesis was not accepted.

Accordingly, migrant issues need to be considered beyond institutional expectations. Two possibilities can be considered regarding the results. First, as explained earlier, highly educated migrants could be positively selected. Second, urgent situations regarding livelihood maintenance as a migrant mother may influence them to find a job quicker in unskilled positions. In this regard, this article carefully emphasised the former in terms of the UK case since job quality showed the least penalty (−7 pp) among the eight countries, but increased employment (+1 pp) after adding having a child. In contrast, the latter explanation could reflect Ireland and Germany’s case since the two largest migrant penalties regarding job quality in the full model were found in Ireland (−23 pp) and Germany (−18 pp), alongside low migrant penalty in employment in Ireland (−6 pp) and Germany (−5 pp) (Figure 3).

Migrant penalty based on different regime typologies

The welfare and production regimes’ impact under the three typologies on migrant penalty levels are conveyed via the second hypothesis. Thus, the full model results regarding employment and job quality are referred to since the institutional effects can be analysed comprehensively when controlling for the four life course events (see Tables 2 and 4 and Figure 3, regarding Model 5 results).

Regarding the welfare regime, the hypothesis assumed that Scandinavian countries’ female migrant penalty in employment with respect to natives would be higher due to generous welfare benefits based on universalism, compared to liberal and conservative welfare states. According to the full model results in Table 2, however, this hypothesis is partially accepted since migrants in the UK were the most penalised among the eight countries (−15 pp), alongside Denmark (−15 pp). Nevertheless, except Finland which was not statistically significant, Model 5’s results for CME/social democratic welfare states ranged from 10 to 15 pp; lower than the other systems.

Regarding the production regime related to skill preference (see model 5 in Table 4), it was expected that migrant penalty in job quality would be lowest in LME/liberal welfare states since general skills are easily transferable unlike accumulated firm- or industry-specific skills. This hypothesis is partially accepted since Ireland, belonging among LMEs, had the highest penalty of around −23 pp; same figure for Norway. However, as expected the UK showed a significantly lower penalty, at −7 pp, while Germany, Finland and Sweden indicated a migrant penalty of −18, −15 and −18 pp with respect to natives (France and Denmark were not statistically significant).

An interesting point regarding education and skill preference in association with labour market outcomes should be emphasised here by referring to Model 5’s tertiary education average coefficients according to the three regime typologies (see the full model in Supplementary Materials S2 and S3). The third level of education posed a significantly beneficial probability for being in a skilled position with respect to lower secondary education in CME/social democratic welfare (65 pp) and CME/conservative welfare states (62.5 pp), but was less significant in LME/liberal welfare states (54 pp); indicating around a 10 pp average difference. Nevertheless, the employment average for tertiary education with respect to lower secondary was higher in LME/liberal welfare states (29 pp) than in CME/conservative welfare (24.5 pp) and CME/social democratic welfare states (21 pp).

Consequently, the highest and the least positive impact of tertiary education upon employment status and job quality can respectively be found in LME/liberal welfare states throughout the entire female workforce. This means skilled jobs are less affected by tertiary education in LME/liberal welfare states compared to the other systems, so employing natives over migrants despite their higher levels of education could occur since general skills are relatively easier to acquire. Thus, education’s negative direct contributions to migrant penalty in the UK and Ireland found in this study can be explained in these statistical results.

In contrasts, the strongest positive impact of tertiary education in gaining a skilled position was observed in CME/social democratic welfare states, followed by CME/conservative welfare states. Therefore, the composition contribution between citizenship and education to reduce migrant penalty regarding job quality can be vindicated, since skilled positions in the two systems are related to accumulated skills which are obtainable within the country. Additionally, migrant penalty in employment, tellingly attributed to citizenship for CME/social democratic welfare states, also coincides with the result by indicating the least positive impact of higher education upon employment status in Scandinavian countries (Supplementary Materials S2 and S3).

Discussion and conclusion

Among the four life course events, education and citizenship significantly affected migrant penalty according to the different institutional backgrounds in the regard to hypothesis 1. This result proved how labour market and industrial characteristics related to skill preference could be associated with the individual conditions of migrants according to different regime typologies. Therefore, migrant penalty attributed to educational background increased in LME/liberal welfare states associated with a general skill preference in employment and job quality, while a positive contribution of education reducing migrant penalty was uncovered in CME/conservative welfare states. Citizenship status was also a significant positive determinant reducing migrant penalty regarding employment; associated closely here with CME/social democratic welfare states’ industrial-specific skill preferences. Regarding job quality, citizenship contributed positively to reduce migrant penalty with respect to natives alongside education as the composition contribution in CME/social democratic welfare states and CME/conservative welfare states. This relates to how migrants from developing countries would need time to acquire skills and this is associated with obtaining higher education and citizenship, especially in these two systems.

Each regime effects on migrant penalty also revealed through the full model by controlling every event based on hypothesis 2. As expected, as for welfare regime, migrant penalty in employment with respect to natives was higher in CME/social democratic welfare states than the other two systems due to the lenient welfare scheme to give migrants a room for not to rely on market immediately. Regarding production regime, migrant penalty in job quality was assumed to be less in LME/liberal welfare states based on general skills easily acquirable. Although the UK showed the lowest migrant penalty among the countries, the highest penalty also was found in Ireland, so it was half confirmed. Through these two hypotheses, life course events and dominant regimes’ effects on migrant penalty were revealed under the three typologies. However, there are two questions to be considered based on the results found in this study.

First, why did the trade-off phenomenon in the UK turn out opposite to that of Ireland by showing higher job quality and lower employment status? This article expected a lower penalty in LME/liberal welfare states based on a strong, flexible labour market and general skill preference which would not pose significant obstacles to labour market integration with both measures. However, the most penalty among the eight countries was for employment in the UK and job quality in Ireland, while the reverse was found for least penalty; the UK for job quality, and Ireland for employment. This could be relevant to migrant demographic characteristic and positive selection, or skilled job proportions in major industries between the two countries. Furthermore, different institutional arrangements between these countries could be the fundamental reason for the results. Although they are classified as the same LME/liberal welfare states, this is since Ireland was investigated as less close to the “pure” liberal welfare states compared to the UK (Ferragina & Seeleib-Kaiser, 2011).

Second, why was there more migrant penalty in the direct contribution of having a child in CME/social democratic welfare states, albeit with greater accessibility to public care services? Is this condition more preferential for natives, or is the penalty instead attributable to the family strategy behind migration? In this regard, the latter could be the answer since the high accessibility to social benefits in Scandinavian countries could impose better decommodification for female migrants having a child, rather than commodifying themselves like in the UK, IE and DE.

As Czaika and Haas (2013) highlighted that the significant determinant of immigrant integration can be found in dominant policy arenas rather than in politics or direct migrant policy, this article has proved how migrant penalty occurs differently according to the dominant regimes alongside the four life course events. Nevertheless, there are some limitations to this study. Accordingly, suggestions for future studies to conduct a more detailed investigation are made based on female migrant characteristics and relevant institutions. The specific characteristics associated with female migrants may include their age category, various partnership forms including cohabitation, the spouse’s employment and job quality status. Furthermore, detailed employment and job quality status could be analysed beyond binary categories in future studies. Although ESS data provided better-detailed conditions such as citizenship and child status which are critical for female migrant life course studies, it has relatively less sample size than the EU-LFS so that certain conditions such as ethnicity difference and detailed labour market outcomes could not be adopted with this data in order to secure statistical power.

Additionally, the importance of changed welfare states based on an institutional interplay between the labour market and work–family reconciliation policies should be investigated regarding whether there are any detailed differences in these policies among the countries and how they lead to female migrant penalty. This is because the institutional similarity and relevant outcomes have kept changing across countries beyond the regimes’ typologies (see Ferragina & Filetti, 2022). Meanwhile, path dependency regarding institutions and political constituency is still acknowledged as an analytical view supporting the typologies (Arts & Gelissen, 2002; Pierson, 1993; Taylor-Gooby, 2004). However, the application of non-static welfare and production regimes would be necessary to give a precise answer to unexplained results (ie the opposite effect of citizenship in Norway compared to other Scandinavian countries), as well as future studies based on dominant regimes.