In-work poverty and labour market trajectories: Poverty risks among the working population in 22 European countries

Abstract

Is in-work poverty a low-wage or an unemployment problem, and is it the same problem all across Europe? Because of the definitional ambiguity, we really do not know. In this article, we use longitudinal European Union-Statistics on Income and Living Conditions (EU-SILC) data from 22 countries and derive a set of distinct clusters of labour market trajectories (LMTs) from information about monthly labour market position from a 36-month observation window and estimate in-work poverty risk for each LMT. The results show that in-work poverty is a problem that affects the self-employed and people in a marginal labour market position, that is, those who for different reasons move in and out of employment. Hence, in-work poverty is mainly an unemployment problem, not a low-wage problem. Besides the fact that the size of LMTs varies between countries, we also expected to find systematic country differences in the effect of LMTs. The analysis did not support that assumption.

Keywords

In-work poverty labour market trajectories low-wage unemployment welfare regimes

Introduction

Although the notion of the working poor has received a great deal of attention, it is still not clear what kind of phenomenon it represents, that is, whether it is mainly a low-wage problem or an unemployment problem. The roots of this lacuna lie in the ‘definitional chaos’ (Crettaz, 2011: 189) that surrounds the notion of in-work poverty. We will use longitudinal data from the European Union-Statistics on Income and Living Conditions (EU-SILC) to derive a set of typical labour market trajectories (LMTs), defined as clusters of individuals who share similar labour market experiences. The LMTs will be related to in-work poverty following the standard definition used in analyses of in-work poverty. The fact that we use comparative data from 22 European countries also means that we will be able to see whether in-work poverty means the same thing in different European countries and to what degree possible differences can be linked to labour market and welfare state typologies. From a policy perspective, the issue is far from trivial. In order to formulate policies to combat in-work poverty, politicians need to know what kind of phenomenon we are actually talking about. Without this knowledge, the risk is that the implemented measures will have little or no effect – or possibly even make things worse. There is ample evidence that lack of work causes poverty and, consequently, paid work is generally seen as the main antidote to poverty. Fighting unemployment is therefore closely interlinked with the fight against poverty. The fact that the EU, in its ‘Europe 2020 targets’, not only looks at unemployment as a cause of poverty, but in fact also uses unemployment (or to be more precise households with very low work intensity) as one dimension in its definition of poverty indicates the perceived close link between the two phenomena (Copeland and Daly, 2012; De Graaf-Zijl and Nolan, 2011). But, what if employment does not protect people from poverty? The presence of in-work poverty constitutes a central challenge, as its very existence entails that the labour market creates jobs with wages that are too low to lift people out of poverty, which in turn implies that we need to reconsider our traditional view of the relationship between employment and poverty. We also need to change our focus from promoting jobs to promoting higher minimum wages, improving income support systems and developing better family policies. However, because of the way in-work poverty is typically defined, we do not know whether the working poor are poor because they have too low hourly wages, experience recurrent unemployment spells, work too few hours or are experiencing a mix of all of these conditions (Crettaz, 2011; Halleröd and Larsson, 2008a; Larsson and Halleröd, 2011). Thus, we do not know whether we should promote higher minimum wages or job creation, or both.

Background

In-work poverty is often perceived as a distinct phenomenon, something different from the ordinary poverty that affects the unemployed, the retired and groups at the margins of society. There are also connotations that hark back to the old distinction between the deserving and undeserving poor, the working poor being deserving as they fulfil their moral duty to do the best they can to support themselves (Halleröd, 2004; Polanyi, 1968). The most clear-cut example of ‘working poor’ is a person who lives in a single-adult household without children and who works full time and full year, but at a wage that is too low to lift him or her above the poverty line. In this case, wage-setting in combination with taxation is the cause of the problem, and the solution is found in increased minimum wages and/or lowered taxation. One may suspect, of course, that higher wages will turn some working poor into non-working poor, as their work will be priced out of the market, but that higher wages will nevertheless solve the working-poor problem. However, the correlation between low wage and in-work poverty is surprisingly weak. Low wage is in most cases a necessary condition for in-work poverty, but most low-wage workers are not poor (Andreß and Lohmann, 2008; Corluy and Vandenbroucke, 2014; Fraser et al., 2011; Halleröd and Larsson, 2008a; Maitre et al., 2012; Marx and Nolan, 2014).

As indicated above, defining in-work poverty is problematic. Poverty typically refers to a person who lives in a household that has an annual income below the poverty line. Determining whether a person is working (employed or self-employed) is usually done in reference to another time frame and the individual, that is, another unit. According to the most common definitions of in-work poverty, a poor person is considered to be working poor if he or she has worked at least 6 or 7 months (Bardone and Guio, 2005; Eurostat, 2010; Klein and Rones, 1989; Larsson and Halleröd, 2011; Lohmann, 2009; Marx and Nolan, 2014; Mosisa, 2003; Peña-Casas and Latta, 2004) during the past 12 months. There are also examples where the working criterion is set to only 1 month (Marx et al., 2012; Peña-Casas and Latta, 2004), and Nightingale and Fix (2004) define a person as working poor if he or she lives in a poor household containing one working person. At the other end of the spectrum we find Maitre et al. (2012), who set the working criterion to full-time full-year employment. The different ways in which the working poor are distinguished from the non-working poor reflect a substantial definitional problem – that is, what kind of phenomenon are we actually investigating? The working poor problem could, after all, be an unemployment problem – which is something that has fundamental policy implications, changing the focus from wage-setting to job creation.

Because labour markets are gendered, measures of in-work poverty tend to produce systematic, and sometimes at first glance, counterintuitive, gender differences. As it seems, the family offers better protection against in-work poverty for women than for men in similar situations. Women with weak labour market positions and low-wage incomes often share a household with a full-time working husband and are therefore less likely to be defined as working poor. Full-time working men, on the other hand, are more likely to share a household with a wife who has a low-wage or no income at all, which increases the in-work poverty risk (Crettaz, 2011; Maitre et al., 2012). We will also observe between-country in-work poverty differences that are generated by differences in the organization of families. Both Lohmann (2009) and Crettaz (2011) show, for example, that the in-work poverty rate among full-time employed Spanish males is relatively high not because they have very low wages, but because they are the sole breadwinners in large families. At the same time, concerns are raised that an increasing rate of non-standard employment, given the gendered labour market, will affect women more negatively than men (Van Lancker, 2012). It is also the case that young people in the Nordic countries are more exposed to in-work poverty than are young people in most other European countries. One reason for this is that young people in the Nordic countries have the ability or, put in another way, are rich enough to leave the parental home at a relatively early age before they are fully established on the labour market, which means that they are not protected from poverty by the parental household during the transition from education into the labour market (Halleröd and Ekbrand, 2014; Larsson and Halleröd, 2011; Maitre et al., 2012).

The ambiguities in the definition of in-work poverty call for more detailed analyses of the working poor’s labour market positions (Van Lancker, 2012). That is, we need to know to what degree they are working, temporarily employed or in some other kind of labour market position, and to what degree in-work poverty occurs in transitions between different labour market positions and, in that case, in what types of transitions. We also need to know to what degree differences between countries are caused by disparities related to LMTs, that is, whether in-work poverty in Country A is more common than in Country B because a larger share of the workforce in Country A has a peripheral labour market position and moves back and forth between employment and unemployment. We also need to know whether the causes of in-work poverty differ between countries. It could be, given the standard definition of in-work poverty, that Countries A and B have the same level of in-work poverty, but that the composition of the group working poor is totally different in the two countries.

Country differences

It is well known that the organization of welfare states is related to country differences in both general poverty (Korpi and Palme, 1998, 2004) and in-work poverty (Bardone and Guio, 2005; Fraser et al., 2011; Lohmann, 2009; Lohmann and Marx, 2008; Peña-Casas and Latta, 2004). The degree of decommodification – that is, to what extent redistribution via social benefits, income protection schemes and pensions decreases individuals’ and households’ market dependency and market distribution of incomes (Esping-Andersen, 1991; Scruggs and Allan, 2006) – can, at least partly, explain differences in between-country poverty rates (Korpi and Palme, 1998). Depending on commonalities in institutional labour market arrangements, countries have been clustered, as an analogue to welfare regimes, according to their labour market regimes (e.g. De la Porte and Jacobsson, 2012; Gallie, 2007; Visser et al., 2009). As shown by Lohmann (2009), labour market regime features, such as centralized wage bargaining, have an important impact on pre-transfer in-work poverty, while welfare state redistributive features are vital to understanding the final post-transfer in-work poverty. Hence, it is the combination of labour market and welfare state features that explains country differences in in-work poverty. Even though welfare regimes and labour market regimes are related to different institutional characteristics, the actual clustering of countries often looks very similar, which basically reflects the interdependency between the organization of the labour market and the welfare state.

One way of understanding how labour market and welfare regimes affect in-work poverty is to relate regime features to theories of labour market segmentation, according to which labour markets can be divided into a core and a periphery (Atkinson, 1984; Barron and Norris, 1976; Ellingsaeter, 1998; Gallie et al., 1998). Work in the periphery is typically characterized by low skill levels, short-term insecure contracts, part-time work and, consequently, low wages (Kalleberg, 2003). Looking at the current development in the EU, there is a trend towards growth in employment both among the highest skilled professional and managerial positions and the low-skilled peripheral occupations primarily within the service sector (Marx and Nolan, 2014), generating what Standing (2011) refers to as the precariat. The way in which labour market regimes are organized will, in the first instance, affect the relative size of the core and peripheral labour force. Labour market regimes will also have an impact on how deep the dividing line between the core and periphery is and what type of peripheral labour force exists. The welfare state will affect to what extent a peripheral labour market position causes in-work poverty. The distinction between core and peripheral labour market positions is also central in relation to the definitional ambiguities surrounding in-work poverty. A high degree of in-work poverty among the core labour force indicates that in-work poverty is, after all, a low-wage problem, while a high prevalence of in-work poverty among the peripheral labour force would indicate that in-work poverty is an unemployment problem.

We will use a typology of labour market/welfare regimes with five different categories: Nordic, Anglo-Saxon, Continental European, Southern European and Eastern European. The classification of countries is shown in Table 1. As can be seen, the analysis does not cover all EU-SILC countries. The reason for this is that the data needed for the longitudinal analysis (see below) are not available for all EU-SILC countries. This is of course unfortunate, and especially Continental Europe is poorly represented, with both Germany and France missing. This typology largely follows the three worlds of welfare capitalism originally suggested by Esping-Andersen (1991), with the addition of a Southern European cluster (Gal, 2010) and a block of Eastern European former communist countries. Whether all countries actually fit the regime type they have been assigned to is another explorative empirical issue. The countries included in the analysis are pragmatically selected. They are the EU-SILC countries that have provided the longitudinal data necessary for extracting the LMTs and for conducting the required multivariate analyses (see below).

Table 1.

Countries in the study and regime type classification.

Country	Regime
Denmark	Nordic
Finland
Norway
Sweden
Ireland	Anglo-Saxon
United Kingdom
Austria	Continental European
Belgium
The Netherlands
Cyprus	South European
Spain
Greece
Italy
Portugal
Czech Republic	East European
Estonia
Hungary
Lithuania
Latvia
Poland
Slovenia
Slovakia

The Nordic countries have an inclusive employment regime, where polarizing tendencies in the labour market are counteracted by employment policies, a strong safety net (Gallie, 2007) and well-developed active labour market policies, family policies and educational systems (Nikolai, 2012). Industrial relations are organized with trade unions and employer associations that are highly self-regulating (Visser, 1996). Policies are aimed at reducing differentials between different types of employees (Gallie, 2007; Visser et al., 2009). Accordingly, we expect to find low in-work poverty rates among the full-time employed core labour force. The outcome among individuals with more peripheral labour market situations is less clear. Most income maintenance programmes are tied to work-related eligibility criteria (Bengtsson and Berglund, 2012; Jørgensen, 2009; Junestav, 2011). Thus, we expect to find the working poor in the Nordic countries first and foremost among people with a peripheral labour market position. Because of early nest leaving, we also expect to find that young people are at high risk for in-work poverty (Halleröd and Ekbrand, 2014; Larsson and Halleröd, 2011). The Nordic countries represent almost the archetype of a dual-earner system. Hence, we expect to find small gender differences in in-work poverty.

The Anglo-Saxon countries are examples of market employment regimes in which the relationship between capital and labour is characterized by non-intervention or absenteeism on the part of the state (Visser et al., 2009: 49–51). Given that these are more self-regulated market regimes, we expect to find that differentials between employment statuses primarily reflect skill differences. However, the existence of minimal employment regulation could result in a ‘relatively high risk of polarization’ among more vulnerable groups at the margins of the labour market (Gallie, 2007: 20). In relation to in-work poverty, we expect to find relatively high poverty rates both among sections of the low-skilled core labour force and among working people with a peripheral labour market position. Although the focus is on getting people into jobs, the difference between the male and female employment rate is considerable, reflecting the lack of a comprehensive dual-earner policy. As a consequence, we expect to find that in-work poverty among the male core labour force is a consequence of low wages and high support burden.

The Continental European countries are generally characterized by high spending on both investment-related and compensatory social policies (unemployment benefits and old age insurance) (Nikolai, 2012). In contrast to the Nordic countries, these ‘Bismarckian welfare systems’ are strongly oriented towards the male-breadwinner model and categorized in a dualist employment regime, with less focus on common employment rights and larger differences between insiders and outsiders (Gallie, 2007; Palier and Martin, 2007). Just as in the Nordic countries, due to work-related performance and merit, we expect to find that in-work poverty is mainly a problem for people in a peripheral labour market position, while the full-time employed with permanent contracts are well-protected from poverty (Van Lancker, 2012).

The Southern European countries could also be classified as having a dualist employment regime (Visser et al., 2009). Compared to the northern parts of Europe, active labour market policies and employment policies (e.g. to increase female labour force participation) are underdeveloped. There is a stronger element of compensatory welfare policy with higher levels of expenditure on unemployment benefits and on old age insurance (Nikolai, 2012). As the institutionalization of capital and labour is weak, unions often use the strike weapon to achieve their ends (Visser, 1996), which most often protects the core labour force and leads to a sharp insider–outsider divide. Thus, we again expect to find that in-work poverty is a problem for people with a more insecure labour market situation (Van Lancker, 2012). However, low female labour force participation, high youth unemployment and late nest leaving also mean that we expect to find in-work poverty among males within the core labour force (Crettaz, 2011; Halleröd and Ekbrand, 2014).

The East European countries’ main characteristic is national agendas dominated by ‘business-friendly’ free market regulations, underdeveloped industrial relations and very low levels of social expenditure (De la Porte and Jacobsson, 2012; Kohl and Platzer, 2007), which could be described as a market employment regime with low compensatory social policies. The statutory minimum wage is significantly lower in comparison with EU-15. These countries also spend less on social protection, and expenditures on active labour market policy and family policy are very limited (Aidukaite, 2011; Nikolai, 2012). Given this, we expect to find that in-work poverty is large not only among those with a peripheral labour market situation, but also relatively common among the core labour force.

LMTs

At any point in time, people can be assigned a specific relationship to the labour market: full-time employed, self-employed, inactive, student, unemployed and so on. These positions, which from a cross-sectional perspective appear to be fixed, are always more or less temporary over time. Rather than dividing people into specific categories depending on their current labour market position, we should try to understand different employment positions as existing along a continuum from total inclusion in the core labour force, different forms of peripheral labour market positions, to total exclusion from the labour market. Also, what from a static position look like identical labour market positions can in a longitudinal perspective in fact be very different transitions. It is one thing to be temporarily unemployed during the transition from higher education into employment, and a completely different thing to start out as a full-time employee and then be pushed into unemployment, only to end up in an early retirement scheme (Halleröd et al., 2013; Halleröd and Westberg, 2006). In our analysis, we will use EU-SILC data in a, to the best of our knowledge, novel way and, based on longitudinal data, distinguish a set of LMTs that will reveal what types of dominating labour market positions exist within the EU. Distinguishing between different LMTs among the working poor will provide vital knowledge about in-work poverty, thus telling us to what degree the working poor are part of the core labour force or to what degree they are found in different peripheral positions.

Data, operationalization and methods

We use an EU-SILC longitudinal dataset from 2007 and 2008. The data make up a representative sample of the total population, but in this case we have restricted our sample to people aged 16–65: the working population. The EU-SILC panel data follow a 4-year cycle, which means that every individual is followed for 4 years and that every year, a fourth of the sample is replaced by a new panel section. An effect of this procedure is that the fewer years we study, the larger sample we get. In order to have a sufficient sample, we have decided to use a 3-year panel. In a first step, we have selected those who participated during the period 2006–2008. In order to further boost the sample size, we have also added the sample section that participated during 2005–2007. When calculating LMTs (see below), we used the full sample (N = 167,454), but in further analyses, we analyse those who were employed or self-employed at least 6 months during the past 12 months of the observation period (N = 93,178). That is, our final sample includes the population that, in the in-work poverty literature, is most commonly defined as working.

Dependent variable

We use the standard EU definition of poverty, that is, a person is defined as poor if he or she lives in a household with an annual equivalent disposable household income that falls below 60 percent of the median equivalent disposable household income in his or her country. We are aware of the intense debate about this definition and, for that matter, other ways of defining and operationalizing poverty (compare Bradshaw and Mayhew, 2011; Guio et al., 2012; Halleröd, 1995; Halleröd and Larsson, 2008b), but this is not the place to delve into that discussion. We use the standard definition because this is the way poverty has typically been measured in previous studies of in-work poverty (e.g. Andreß and Lohmann, 2008; Crettaz, 2011; Fraser et al., 2011; Peña-Casas and Latta, 2004). We measure poverty during the third observation year. The country-specific in-work poverty rates are shown in Appendix 1.

LMTs

At each annual interview, respondents are asked to give a 12-month retrospective description of their main activity. For every month, nine alternatives are offered:

Employee (full time);

Employee (part time);

Self-employed (full time);

Self-employed (part time);

Unemployed;

Retired;

Student;

Other inactive;

Compulsory military service.

We use this information to derive clusters of specific LMTs. As a first step, before moving on with the cluster analysis, we reduce the number of main activity positions. The ninth category is very small and exists only in some countries, and therefore this group is excluded from the analysis. We also collapse full-time and part-time self-employed. As a consequence, seven clusters, those who do not change position during the observation period, are given from the beginning. Among the non-static, transitions are observed on a monthly basis during a 3-year period, which gives 7³⁵ theoretically possible transitions. The cluster analysis handles each observed month as a single variable and looks for a solution that minimizes the number of unique combinations during the 36-month period.¹ The cluster analysis reduces the complexity of the data into 34 clusters that describe different LMTs. Once we have derived the clusters, we start a manual process of reducing the number clusters. This process is necessary both from a theoretical and practical perspective, as we need clusters that are both theoretically interpretable and statistically manageable. Some clusters are deleted automatically because none in these groups fulfils the working criteria set for the third observation year, for example, full-time employed who move into retirement, those who move from employment into unemployment or/and inactivity that lasts more than 6 months, and those who are more or less permanently excluded from the labour market during all 3 years. Thereafter, we reduce the data by merging clusters with similar main activities and, finally, we end up with six different clusters, described in Table 2.

Table 2.

Description of LMTs.

Core labour force	Is mainly made up of individuals who, without interruptions, have been employed full time during all 3 years. Also included in this category are those who have been employed part time all 36 months. This is a relatively small group (7%) with generally low poverty rates, which in some countries in fact is zero.
Peripheral labour force	Mainly full-time employed, but who have experienced periods of unemployment and inactivity. Are also mixing part-time and full-time employment.
Into core labour force	Individuals who were unemployed or inactive in the beginning of the period but who are moving into employment by the end of the period. Also included are students who are making the transition from education to full-time work.
Core self-employed	Self-employed during all 36 months.
Peripheral self-employed	Mainly self-employed but who have experienced periods of unemployment and inactivity.
Marginalized peripheral labour force	Are, during the 36-month period, mixing episodes of employment with significant periods of unemployment and/or inactivity.

The first group we define as the core labour force. Here, we find individuals who have been full-time or part-time employed every month during the 3-year observation window. The vast majority of this group, 94 percent, is full-time employed and part-timers are included for practical reasons. The next cluster contains individuals who are moving from a peripheral position into full-time employment, that is, presumably into the core labour force. Included, and of equal size, are also students who are leaving education and are about to establish themselves on the labour market. In the peripheral labour force, we find people who are mixing different employment positions, the most common being full-time employment with episodes of unemployment and/or inactivity. In the next two clusters, we find the self-employed. They are divided into a core group, that is, those self-employed for all 36 months and a peripheral cluster consisting of people who mix self-employment with other activities. Finally, we have the marginalized peripheral labour force. Here, we find people who have mainly been unemployed or inactive during the past 36 months, but who happen to fulfil the working criteria during the last 12 months of this period. The size of the clusters in each country is shown in Appendix 1.

Our approach means that we move away from the static snapshot approach used in most analyses of in-work poverty. It also means that our definition of that core labour force is stricter than Maitre, Nolan and Whelan’s (2012) definition of a full-time, full-year worker, which only takes into account 1 year’s experience. Another way of differentiating among the different labour market positions is to distinguish between temporary (time limited) and permanent (without time limit) employment contracts. This is problematic, as the actual permanence invoked by a permanent contract varies both within and, especially, between countries (Van Lancker, 2012). Our approach avoids this problem because we use information about main activity, not type of contract. The measure also relates to measures of work intensity, because it is based on the individuals’ monthly labour market engagement, but instead of giving a 1-year snapshot, it provides information about a more long-term LMT. However, because it is based on the individual LMT, it only gives partial information about household work intensity. Creating LMTs for all adult household members is possible in theory, but apart from making the analysis additionally complicated, it would also entail a need to restrict the analysis to those countries that provide information for all household members. Among the 22 countries in this study, data on all household members are not available for the Nordic countries and Slovenia.

Control variables

To estimate the impact of LMTs, we need to control for a number of conditions. During a lifetime, most people make age-related transitions, first into the labour force and later out of the labour force. Both these transitions are to varying degrees associated with unemployment and other forms of employment interruptions (Halleröd and Ekbrand, 2014; Halleröd et al., 2013; Halleröd and Westberg, 2006). We therefore control for age and an age-squared function. The latter is intended to capture the curve linear impact of age. As discussed above, in-work poverty is a phenomenon that occurs when an individual is, at least partly, employed or self-employed and lives in a household that falls under the poverty line. Hence, in-work poverty risk is not only dependent on the individual’s LMT and related income, but also on the number of dependent household members, other household members’ employment and incomes, and of course welfare state arrangements. The latter aspect is one of the reasons as to why we compare countries. To control for household situation, we include variables for household type (single-adult household with or without children, married or cohabitant with or without children), the number of children and ratio between number of children and number of adults in the household. As an indicator of household work intensity, we estimate the proportion of adults in the household that are employed or self-employed. For two reasons, we have chosen not to use the household work intensity measure provided by the EU-SILC. First, work intensity is only available for the cross-sectional dataset, not for the longitudinal dataset. Second, the main component of the work intensity measure used in the cross-sectional EU-SILC originated from the information on monthly main activity, that is, the same EU-SILC question we use to identify individual LMTs. Finally, we also control for education as an indicator of the individual’s human capital and, because of the observed differences between men and women, for gender.

Method

We will use a mixed model type of analysis taking into account both within- and between-country differences (Hox, 2010). We will carry out the analysis in the following way:

Estimate the null-model, that is, fixed and random intercepts;

Estimate fixed LMT effects on the individual in-work poverty risk, including the interaction between sex and LMTs;

Estimate the fixed effect model including control variables;

To check whether the impact of LMTs varies between countries, we include random slope terms for the LMTs and for the interaction between LMT and sex.

Because we expect to find systematic gender differences, an interaction term between sex and LMTs is included in all models (except, of course, the null-model). By doing so, and not estimating separate models for men and women, we get more straightforward statistical tests of gender differences.

Results

Descriptive statistics

Table 3 shows the distribution of LMTs. About 57 percent of the in-work population belongs to the core labour force. In this group, only 2.8 percent were poor during the third year of the observation window. Because of the size of the core labour force, they nevertheless make up almost 25 percent of all working poor. The peripheral labour force makes up 16 percent of the population. Here, the poverty rate is 8.2 percent, and close to 22 percent of all working poor is found in this cluster. A slightly higher poverty rate, 9.3 percent, is found in the core labour force category. Among all other LMTs, the poverty rate is above 14 percent. Among the marginalized peripheral labour force, 17 percent is working poor. Although the group is small, less than 3 percent, they still make up 7 percent of the working poor. Worth noticing is that among the core self-employed, the in-work poverty rate is 16 percent and among the peripheral self-employed, the poverty rate is 15 percent. Adding these groups together, we can see that more than one-third of all working poor are self-employed, which corresponds with findings from previous analyses of in-work poverty (Crettaz, 2011; Goerne, 2011; Halleröd and Larsson, 2008a).

Table 3.

Distribution of LMT clusters, poverty rates and distribution of the poor.

	LMTs frequency	LMTs percent	In-work poverty rate	Distribution of in-work poverty
Core labour force	53326	57.2	2.8	24.8
Core self-employed	8931	9.6	16.0	23.6
Into core labour force	7929	8.5	9.3	12.2
Marginalized peripheral labour force	2505	2.7	17.0	7.0
Peripheral labour force	15978	17.2	8.2	21.5
Peripheral self-employed	4509	4.8	14.6	10.9
All	93178	100.0	11.3	100.0

Mixed model analysis

Table 4 shows the results from the regression models. The null-model provides estimates for the fixed intercept (overall mean) and the random country intercept, that is, an estimate of how much each country deviates from the overall mean. The random intercepts are shown by the solid dots in Figure 1. We find the highest figures in Greece followed by Poland, and thereafter three additional Southern European countries. The only Southern European country below the overall mean is Cyprus. However, we also find all the Nordic countries, except Denmark, above the overall mean. The results for the two Anglo-Saxon countries differ, with the United Kingdom above the mean and Ireland clearly below the overall mean. All three continental countries are found below the overall mean. The Eastern European countries are very heterogeneous, with Poland at one end of the distribution and the Czech Republic at the other.

Table 4.

Mixed model in-work poverty estimates – odds ratios.

	Null-model	Model 1	Model 2	Model 3
Fixed effects
Intercept	0.05***	0.02***	0.08***	0.09***
LMTs – (Core labour force: ref)		1.00	1.00	1.00
Core self-employed		7.49***	6.73***	6.53***
Into core labour force		3.45***	2.92***	2.48***
Marginalized peripheral labour force		6.55***	4.48***	4.72***
Peripheral labour force		2.77***	2.41***	2.35***
Peripheral self-employed		6.34***	5.74***	5.62***
LMTs interactions – male
Core self-employed × male		0.71***	0.79**	0.87
Into core labour force × male		1.00	0.97	0.91
Marg. peripheral labour force × male		1.14	1.28*	1.49#
Peripheral labour force × male		1.23*	1.27**	1.27
Peripheral self-employed × male		0.79*	0.87	0.9
Gender (male)		1.30***	1.13*	0.98
Age (z-scores)			0.88***	0.88***
Age square (z-scores)			0.96**	0.95**
Children per adult in household			0.98#	0.97**
Number of children			1.20***	1.21***
Proportion of adults in employment			0.31***	0.31***
Household type (couple without children: ref)			1	1
Couple with children			0.90#	0.89*
Single adult with children			1.62***	1.65***
Single adult without children			1.19***	1.20***
Education (Primary: ref)			1	1
Lower secondary			0.68***	0.67***
Upper secondary			0.41***	0.41***
Tertiary			0.21***	0.21***
Random effects:
Intercept (variance)	0.29	0.21	0.19	0.06
Gender (variance)				0.14
Core labour force (variance)				0.19
Core self-employed (variance)				0.54
Into core labour force (variance)				0.33
Marginalized peripheral (variance)				0.12
Peripheral labour force (variance)				0.21
Peripheral self-employed (variance)				0.34
N	93178	93178	93178	93178
Countries	22	22	22	22

Significance: ***p < 0.001, **p < 0.01, *p > 0.05, #p > 0.1.

Figure 1.

Mixed model in-work poverty estimates – random labour market trajectories effects, model 2.

In Model 1 (Table 4), we include fixed estimates for LMTs and the interaction between gender and LMTs. The odds-ratio varies greatly between LMTs, basically reproducing the result from Table 3. The estimate for gender shows that the in-work poverty risk is higher among men, especially among men in the peripheral labour force, but less so among peripheral self-employed. The variation in the random intercept decreases somewhat, which is reflected by the estimates of random intercepts (rings) in Figure 1, which, compared with the null-model, are generally closer to the fixed intercept. The greatest changes are seen for Greece, probably because of the large share of self-employed, and in Finland, probably reflecting the small size of the core labour force. However, the overall picture is that the internal ordering of countries remains relatively intact.

In the third step, we estimated the random LMT effects (not shown, can be requested from the author). To our surprise, few estimates were significant, that is, most estimates did not deviate significantly from the fixed effects in Model 1. Further, looking at the significant estimates, it is difficult to find any systematic support for the expected labour market regime differences. For example, we did expect that, in the Nordic countries, the in-work poverty risk would be low among the core labour force and high among those who are transitioning into the labour force. But, this seems to be the case only for Finland and Norway, not for Denmark and Sweden. The poverty risk is also relatively high for the core labour force in Italy, but not in the other Southern European countries. Neither does the analysis reveal any systematic gender difference; the country pattern looks largely the same for men and women – out of 132 estimates for gender differences per employment type and country, only nine significantly differed from zero.

In Model 2, we also include a set of control variables, which, as expected, additionally decreases the fixed effects of LMTs. The impact of LMTs is nevertheless highly significant, that is, in-work poverty is still related to peripheral labour market positions and self-employment. The addition of controls only marginally decreases the random intercept variance. In Model 3, we extend the model to include random LMT estimates (random slopes). This improves the model, and the random intercept variance is decreased. Also, if we inspect the random effects (not shown), it turns out that few, in fact only 33 out of 110 random slope estimates, are significant and that it is very difficult to find any systematic, interpretable pattern that can be related to our a priori assumptions.

Conclusion

In-work poverty is a much-discussed, but we wish to argue, ill-defined social phenomenon. We have tried to address this lacuna using monthly information about labour market positions from a 36-month period in order to construct a set of specific LMTs. We argue that this provides a more detailed and diverse picture of the type of labour market positions people experience as well as a better understanding of the relationship between labour market positions and poverty. We have used longitudinal EU-SILC data from 22 countries.

The analysis shows that in-work poverty in Europe is mainly an unemployment problem and a problem among the self-employed. All over Europe, the poverty risk among the core labour force – that is, individuals who were full-time employed without interruption during the whole 36-month observation window – is very low. The only country that significantly deviates from the overall European level is Italy, where the poverty risk in the core labour force is relatively high, making Italy an interesting country case that needs to be explored in more detail in further research. Otherwise, in-work poverty is mainly a problem for people establishing themselves on the labour market and for those in a more permanent precarious labour market situation, mixing periods of employment with unemployment and activities outside the labour market. We also find, as expected, high in-work poverty rates among the self-employed, both among the core group who were self-employed during the entire 36-month period and those who have a more peripheral position as self-employed.

There are substantial differences in the in-work poverty rate between European countries. To some degree, these differences follow different regime types. We find high in-work poverty rates in most of the Southern European countries and in the United Kingdom, but also in the Nordic countries, except for Denmark. In the more heterogeneous transitional model characterizing Eastern European countries, it is hard to find a common pattern of in-work poverty. We did expect to find some systematic country differences that corresponded to the classification of labour market regimes. However, when looking for systematic country differences, most effects were insignificant and an overall pattern was difficult, if not impossible, to find. Thus, the reason why LMTs explain a substantial part of the variation in in-work poverty between European countries is not that LMTs have significantly different effects in different countries, but because the size of various LMTs varies between countries.

As in most other studies, we find that the self-employed suffer from a very high in-work poverty risk (Crettaz, 2011; Fraser et al., 2011; Lohmann and Marx, 2008). Again it can be concluded that few countries deviate from this pattern, even though the risk seems to be somewhat higher in some Southern European countries. To what degree this result is driven by actual hardship among small-scale self-employed, farmers, for example, or by difficulties collecting reliable income data from the self-employed still needs to be investigated. It nevertheless seems as if a substantial share, almost 40 percent, of Europe’s working poor is self-employed.

Relating our results to previous research on in-work poverty (e.g. Andreß and Lohmann, 2008; Crettaz, 2011; Crettaz and Bonoli, 2011; Fraser et al., 2011; Marx and Nolan, 2014), we conclude that it is time for both policy-makers and social scientists to reconsider the concept of in-work poverty, at least within a European context. The concept leads us to believe that measures of in-work poverty are related to, or caused by, low hourly wage. However, research has repeatedly shown that the link between low pay and in-work poverty is weak, while the link between lack of employment and in-work poverty is strong. What this suggests is that we need to focus on the actual labour market conditions that lead to poverty, that is, we need to focus on unemployment and insecure short-term employment conditions and the level, duration and conditionality of our social protection schemes, not least among them, unemployment benefits. Recent analyses of household work intensity and household joblessness (Corluy and Vandenbroucke, 2014; De Graaf-Zijl and Nolan, 2011) would seem, from this perspective, to be a more fruitful and accurate way of proceeding when examining the relationship between employment and poverty. Taking one’s point of departure from household work intensity also has the appealing feature of bringing the job situation within the household to the fore. However, we do believe that it is important to continue to analyse whether working people are paid enough to avoid poverty, but these analyses should focus on wages by separating low hourly wages from few working hours. This is an important aspect, because even though our analysis shows that, across Europe, the large majority of the employed working poor has a precarious labour market position involving a long-term mix of employment, unemployment and inactivity, we do not know whether all of them or even the majority has an hourly wage that would lift them above the poverty line even if they were to work full time, full year. To shed light on this question, additional research is needed. Finally, we again wish to highlight the fact that a large share of the working poor are self-employed, which points to the need to gain a deeper understanding of economic conditions among the self-employed.

The EU 2020 targets stipulate that the EU will have at least 20 million fewer people in or at risk of poverty and social exclusion by the year 2020. The way to reach that goal is not only to reduce the number of people who are at risk of (income) poverty and/or social exclusion, but also to decrease the number of people living in households with low job intensity. Now, a low work intensity household is a household in which working-age members worked less than 20 percent of their potential during the past year. In cases where there are, for example, three working-age household members, two unemployed and one working part time, there is a possibility that in-work poverty and low work intensity overlap, that is, that the same individuals are covered by both definitions, which is yet another indication of the problems surrounding the concept of in-work poverty. However, using the 6-months work criteria when defining in-work poverty will most probably lead to a situation where, in most cases, two different populations are identified. That is, most working poor are not living in a low work intensity household. Thus, our results indicate that lifting people just above the threshold for low job intensity will probably have a marginal impact on reducing the risk of poverty. The low job intensity measure needs to be more ambitious. But, we will also argue that, as an indicator of poverty and social exclusion, low job intensity is an anomaly. The interest devoted to in-work poverty is related to the strongly endorsed assumption that paid work will lift people out of poverty. From this perspective, using household joblessness as an indicator of poverty and social exclusion, as is the case in the EU targets, is problematic. What we need is knowledge about how much members of a household need to work to avoid poverty and social exclusion as well as about what types of labour market situations, that is, LMTs, lead to poverty and social exclusion. Hence, unemployment and joblessness should be treated as a cause of poverty and social exclusion, not as an indicator of these phenomena.

In line with previous studies, our analysis shows that the risk of in-work poverty is relatively large among households with children, especially among households with many children and among single-parent households. Because this is a consistent finding, it calls for improved family policies, which most likely imply a redistribution of economic resources from households without children to households with children. Finally, our main conclusion is that a more nuanced operationalization of individual LMTs shows that in-work poverty is mainly an unemployment problem. Very few of those who are fully integrated on the labour market are poor, and it is mainly the existence of a peripheral labour market that causes in-work poverty. This calls for policies that first and foremost fight unemployment, promote employment security, and that, in the event of unemployment, guarantee unemployment benefits that lift the unemployed above the poverty line.

Footnotes

Appendix

Appendix 1.

Percent in work, in-work poverty rates and LMTs.

	In-work	In-work poverty	Core	Peripheral	Into core	Core self-emp.	Peripheral self-emp.	Marginalized
DK	82.9	2.9	66.0	17.7	6.6	4.0	4.6	1.1
FI	72.8	6.3	42.4	17.9	10.7	19.0	7.7	2.3
NO	77.9	6.9	58.9	24.4	10.0	2.8	3.0	0.9
SE	82.0	5.7	55.9	26.0	8.2	4.0	5.0	0.9
Nordic	78.9	5.5	55.8	21.5	8.9	7.5	5.1	1.3
IE	63.2	2.2	52.6	21.2	5.6	13.4	4.6	2.6
UK	73.2	6.0	61.4	20.3	5.1	8.0	3.3	1.8
Anglo-Saxon	68.2	4.1	57.0	20.8	5.4	10.7	4.0	2.2
AT	69.9	5.7	56.5	21.9	6.3	7.1	4.7	3.5
BE	63.6	3.3	61.1	18.4	7.5	7.0	3.5	2.6
NL	75.3	3.3	60.1	27.5	2.4	5.1	2.8	2.1
Continental	69.6	4.1	59.2	22.6	5.4	6.4	3.7	2.7
CY	67.4	4.1	61.1	15.1	7.5	8.8	5.7	1.7
ES	64.8	9.9	49.9	20.0	8.7	11.4	5.9	4.1
GR	62.4	11.8	39.6	12.6	5.0	30.9	8.2	3.6
IT	60.0	8.7	52.0	15.3	6.0	16.3	7.0	3.3
PT	67.3	8.3	59.8	10.9	7.2	14.0	5.6	2.6
South European	64.4	8.6	52.5	14.8	6.9	16.3	6.5	3.1
CZ	67.8	1.5	67.6	10.1	7.8	8.8	3.6	2.1
EE	67.7	5.8	62.2	14.3	13.0	4.3	3.8	2.5
HU	59.2	4.2	58.5	19.1	10.5	4.5	4.2	3.3
LT	69.6	5.3	66.1	11.4	9.7	6.3	3.6	3.0
LV	68.8	4.7	63.9	15.3	11.3	2.8	4.2	2.5
PL	56.6	10.6	47.8	12.5	12.0	16.0	6.3	5.2
SI	61.6	3.6	70.6	9.8	9.2	6.6	1.9	2.0
SK	67.2	3.0	67.0	11.3	10.4	3.4	5.6	2.4
East European	64.8	4.8	63.0	13.0	10.5	6.6	4.2	2.9

DK: Denmark; FI: Finland; NO: Norway; SE: Sweden; IE: Ireland; UK: United Kingdom; AT: Austria; BE: Belgium; NL: The Netherlands; CY: Cyprus; ES: Spain; GR: Greece; IT: Italy; PT: Portugal; CZ: Czech Republic; EE: Estonia; HU: Hungary; LT: Lithuania; LV: Latvia; PL: Poland; SI: Slovenia; SK: Slovakia.

Funding

This research was supported by The Swedish Research Council (Grant 2007-1752). Previous versions of the manuscript has been presented at the FISS conference, Sigtuna, Sweden. We thank Björn Gustafsson, other participants at the FISS conference, and colleagues for helpful comments.

Notes

References

Aidukaite

(2011) ‘Welfare Reforms and Socio-Economic Trends in the 10 New EU Member States of Central and Eastern Europe’, Communist and Post-Communist Studies 44(3): 211–9.

Andreß

H.-J.

Lohmann

(eds) (2008) The Working Poor in Europe: Employment, Poverty and Globalization. Cheltenham: Edward Elgar.

Atkinson

A.B.

(1984) ‘Manpower Strategies for Flexible Organisations’, Personnel Management 16(115): 21–31.

Bardone

Guio

A.-C.

(2005) ‘In-Work Poverty’, in Statistics in Focus – Population and Social Conditions. Luxembourg: Eurostat, available at http://ec.europa.eu/employment_social/social_inclusion/docs/statistics5-2005_en.pdf.

Barron

D.R.

Norris

G.M.

(1976) ‘Sexual Divisions and the Dual Labour Market’, in Barker

D.L.

Allen

(eds) Dependence and Exploitation in Work and Marriage, pp. 47–69. New York: Longman.

Bengtsson

Berglund

(2012) ‘Labour Market Policies in Transition: From Social Engineering to Standby-Ability’, in Larsson

Letell

Thörn

(eds) Transformations of the Swedish Welfare State: From Social Engineering to Governance? pp. 86–103. Basingstoke: Palgrave Macmillan.

Bradshaw

Mayhew

(2011) The Measurement of Extreme Poverty in the European Union. Brussels: Directorate-General for Employment, Social Affairs and Inclusion.

Copeland

Daly

(2012) ‘Varieties of Poverty Reduction: Inserting the Poverty and Social Exclusion Target into Europe 2020’, Journal of European Social Policy 22(3): 273–87.

Corluy

Vandenbroucke

(2014) ‘Individual Employment, Household Employment, and Risk of Poverty in the European Union: A Decomposit Analysis’, in Cantillon

Vandenbroucke

(eds) Reconciling Work and Poverty Reduction, pp. 94–130. Oxford: Oxford University Press.

10.

Crettaz

(2011) Fighting Working Poverty in Post-Industrial Economies: Causes Trade-Offs and Policy Solutions. Cheltenham: Edward Elgar.

11.

Crettaz

Bonoli

(2011) ‘Worlds of Working Poverty: National Variations in Mechanisms’, in Fraser

Gutiérrez

Peña-Casaan

(eds) Working Poverty in Europe: A Comparative Approach, pp. 46–72. Basingstoke: Palgrave Macmillan.

12.

De Graaf-Zijl

Nolan

. (2011) ‘Household Joblessness and Its Impact on Poverty and Deprivation in Europe’, Journal of European Social Policy 21(5): 413–31.

13.

De La Porte

Jacobsson

(2012) ‘Social Investment or Recommodification? Assessing the Employment Policies of the EU Member States’, in Morel

Palier

Palme

(eds) Towards a Social Investment Welfare State? Ideas, Policies and Challenges, pp. 117–52. Bristol: Policy Press.

14.

Ellingsaeter

A.L.

(1998) ‘Labour Market Restructuring and Polarization Processes: The Significance of Political-Institutional Factors’, Economic and Industrial Democracy 19: 579–603.

15.

Esping-Andersen

(1991) The Three Worlds of Welfare Capitalism. Oxford: Polity Press.

16.

Eurostat (2010) In-Work Poverty in the EU. Luxembourg: Eurostat.

17.

Fraser

Gutiérrez

Peña-Casa

(eds) (2011) Working Poverty in Europe: A Comparative Approach. Basingstoke: Palgrave Macmillan.

18.

Gal

(2010) ‘Is There an Extended Family of Mediterranean Welfare States?’ Journal of European Social Policy 20(4): 283–300.

19.

Gallie

(2007) Employment Regimes and the Quality of Work. Oxford: Oxford University Press.

20.

Gallie

White

Cheng

Tomlinson

(1998) Restructuring the Employment Relationship. New York: Oxford University Press.

21.

Goerne

(2011) ‘A Comparative Analysis of in-Work Poverty in the European Union’, in Fraser

Gutiérrez

Peña-Casaan

(eds) Working Poverty in Europe: A Comparative Approach, pp. 15–45. Basingstoke: Palgrave Macmillan.

22.

Guio

A.-C.

Gordon

Marlier

(2012) ‘Measuring Material Deprivation in the EU: Indicators for the Whole Population and Child-Specific Indicators’, Eurostat Methodologies and Working Papers, Office for Official Publications of the European Communities (OPOCE), Luxembourg.

23.

Halleröd

(1995) ‘The Truly Poor: Indirect and Direct Measurement of Consensual Poverty in Sweden’, Journal of European Social Policy 5(2): 111–29.

24.

Halleröd

(2004) ‘What I Need and What the Poor Deserve: Analyzing the Gap between the Minimum Income Needed for Oneself and the View of an Adequate Norm for Social Assistance’, Social Forces 83(1): 35–59.

25.

Halleröd

Ekbrand

(2014) ‘Labour Market Trajectories and Young Europeans’ Capabilities to Avoid Poverty, Social Exclusion and Dependency: A Comparative Analysis of 23 European Countries’, in Otto

H.-U.

Atzmüller

Berthet

Bifulco

Bonvin

J.-M.

Chiappero-Martinetti

Egdell

Zielenska

. . . M.

(eds) Facing Trajectories from School to Work. Cham: Springer International Publishing Switzerland.

26.

Halleröd

Larsson

(2008a) ‘In-Work Poverty in a Transitional Labour Market: Sweden, 1988–2003’, in Andreß

H.-J.

Lohmann

(eds) The Working Poor in Europe: Employment, Poverty and Globalization, pp. 155–78. Cheltenham: Edward Elgar.

27.

Halleröd

Larsson

(2008b) ‘Poverty, Welfare Problems and Social Exclusion’, International Journal of Social Welfare 17(1): 15–25.

28.

Halleröd

Westberg

(2006) ‘Youth Problem: What’s the Problem? A Longitudinal Study of Incomes and Economic Hardship among Swedish Youth’, Acta Sociologica 49(1): 83–102.

29.

Halleröd

Örestig

Stattin

(2013) ‘Leaving the Labour Market: The Impact of Exit Routes from Employment to Retirement on Health and Wellbeing in Old Age’, European Journal of Ageing 10: 25–35.

30.

Hox

J.J.

(2010) Multilevel Analysis: Techniques and Applications. New York: Routledge.

31.

Jørgensen

(2009) ‘From a Beautiful Swan to an Ugly Duckling: The Renewal of Danish Activation Policy since 2003’, European Journal of Social Security 11(4): 337–67.

32.

Junestav

(2011) ‘Promoting Employment or Employability? The Move from Active Labour Market Policy to Workfare’, in Thörnquist

Engstrand

Å.-K.

(eds) Precarious Employment in Perspective: Old and New Challenges to Working Conditions in Sweden, pp. 221–48. Bruxelles/New York: P.I.E. Peter Lang.

33.

Kalleberg

A.L.

(2003) ‘Flexible Firms and Labor Market Segmentation: Effects of Workplace Restructuring on Jobs and Workers’, Work and Occupations 30: 154–75.

34.

Klein

B.W.

Rones

P.-L.

(1989) ‘A Profile of the Working Poor’, Monthly Labor Review 112: 3–13.

35.

Kohl

Platzer

H.-W.

(2007) ‘The Role of the State in Central and Eastern European Industrial Relations: The Case of Minimum Wages’, Industrial Relations Journal 38(6): 614–35.

36.

Korpi

Palme

(1998) ‘The Paradox of Redistribution and Strategies of Equality: Welfare State Institutions, Inequality, and Poverty in the Western Countries’, American Sociological Review 63(5): 661–87.

37.

Korpi

Palme

(2004) ‘New Politics and Class Politics in the Context of Austerity and Globalization: Welfare State Regress in 18 Countries 1975–1995’, American Political Science Review 97(3): 425–46.

38.

Larsson

Halleröd

(2011) ‘Sweden: The Impact of Policy and Labour Market Transformation’, in Fraser

Gutiérrez

Peña-Casas

(eds) Working Poverty in Europe: A Comparative Approach, pp. 112–32. Hampshire: Palgrave Macmillan.

39.

Lohmann

(2009) ‘Welfare States, Labour Market Institutions and the Working Poor: A Comparative Analysis of 20 European Countries’, European Sociological Review 25(4): 489–504.

40.

Lohmann

Marx

(2008) ‘The Different Faces of in-Work Poverty across Welfare State Regimes’, in Andreß

H.-J.

Lohmann

(eds) The Working Poor in Europe: Employment, Poverty and Globalisation, pp. 17–46. Cheltenham: Edward Elgar.

41.

Maechler

Rousseeuw

Struyf

Hubert

Hornik

(2014) Cluster: Cluster Analysis Basics and Extensions, R package version 1.15.3. Vienna: R Foundation for Statistical Computing.

42.

Maitre

Nolan

Whelan

C.T.

(2012) ‘Low Pay, in-Work Poverty and Economic Vulnerability: A Comparative Analysis Using EU-SILC’, Manchester School 80(1): 99–116.

43.

Marx

Nolan

(2014) ‘In-Work Poverty’, in Cantillon

Vandenbroucke

(eds) Reconciling Work and Poverty Reduction: How Successful Are European Welfare States? pp. 131–156. Oxford: Oxford University Press.

44.

Marx

Vanhille

Verbist

(2012) ‘Combating in-Work Poverty in Continental Europe: An Investigation Using the Belgian Case’, Journal of Social Policy 41: 19–41.

45.

Mosisa

A.T.

(2003) ‘The Working Poor in 2001’, Monthly Labor Review 126: 13–9.

46.

Nightingale

D.S.

Fix

(2004) ‘Economic and Labor Market Trends’, Future of Children 14: 49–59.

47.

Nikolai

(2012) ‘Towards Social Investment? Patterns of Public Policy in the OECD World’, in Morel

Palier

Palme

(eds) Towards a Social Investment Welfare State? Ideas, Policies and Challenges, pp. 91–116. Bristol: Policy Press.

48.

Palier

Martin

(2007) ‘Editorial Introduction – From ‘A Frozen Landscape’ to Structural Reforms: The Sequential Transformation of Bismarckian Welfare Systems’, Social Policy and Administration 41(6): 535–54.

49.

Peña-Casas

Latta

(2004) Working Poor in the European Union. Luxembourg: European Foundation for the Improvement of Living and Working Conditions.

50.

Polanyi

(1968) The Great Transformation. Boston, MA: Beacon Press.

51.

Scruggs

Allan

(2006) ‘Welfare-State Decommodification in 18 OECD Countries: A Replication and Revision’, Journal of European Social Policy 16(1): 55–72.

52.

Standing

(2011) The Precariat: The New Dangerous Class. London: Bloomsbury Academic.

53.

Van Lancker

. (2012) ‘The European World of Temporary Employment Gendered and Poor?’ European Societies 14(1): 83–111.

54.

Visser

(1996) ‘Traditions and Transitions in Industrial Relations: A European View’, in Van Ruysseveldt

Visser

(eds) Industrial Relations in Europe: Traditions and Transitions, pp. 1–41. London: Sage.

55.

Visser

Beentjes

Van Gerven

Di Stasio

(2009) ‘The Quality of Industrial Relations and the Lisbon Strategy’, in European Commission (ed.) Industrial Relations in Europe, 2008, pp. 45–72. Brussels: European Commission.