In-work poverty in Western Europe. A longitudinal perspective

Employment is among the most important factors protecting individuals and their families against economic poverty. Acknowledging the close link between poverty and employment (De Graaf-Zijl and Nolan 2011; Copeland and Daly 2014), the targets of the European Union's (EU) Europe 2020 initiative use households’ low work intensity as one of the central risk dimensions in their definition of poverty. Nonetheless, a significant number of individuals experience poverty despite gainful employment – referred to as ‘in-work poverty (IWP) or, more precisely, ‘in work and at risk of poverty.’ According to Eurostat,¹ nearly one-tenth (9.2%) of EU workers between the ages of 18 and 64 were at risk of poverty after social transfers in 2016, and IWP rates in most EU countries (European Commission 2012; Eurofond 2017) had already begun increasing before the 2009 financial crisis. The chronic presence of IWP, therefore, constitutes a central challenge in Europe (Andreß and Lohmann 2008; Fraser et al.2011; Peña-Casas et al.2019).

This phenomenon has attracted substantial scholarly attention in recent years (Lohmann 2009; Giesselmann 2015; Horemans et al.2016; Tejero 2017; Lohmann and Crettaz 2018; Filandri and Struffolino 2019; Struffolino and Van Winkle 2021; Eurostat 2022; Wolf et al.2022). The growing diffusion of IWP in contemporary societies has been connected to broader economic and labor market changes and, more specifically, to globalization trends (Blossfeld et al.2011) and EU-wide institutionally-driven processes of labor market deregulation and dualization (Palier and Thelen 2010; Barbieri and Cutuli 2016; Brülle et al.2019), though relevant differences exist across countries regarding the social consequences of such processes.

While the levels and compositions of IWP have been previously documented, less is known about its dynamics and social stratification, which is where this paper makes its contribution. Gaining a deeper understanding of the dynamic relationship between poverty and paid employment within households is important, as it not only informs policy design but also enhances our comprehension of the broader processes of inequality and social stratification. This is particularly true since adverse socioeconomic conditions – with IWP being no exception – tend to persist over time and entrap the same individuals, households, and social groups, giving rise to patterns of cumulative disadvantages (Merton 1988; DiPrete and Eirich 2006; Vandecasteele 2010, 2011; Vandecasteele and Giesselmann 2018; Cutuli and Grotti 2020; Barbieri and Gioachin 2022). To design policy measures that can tackle the catalysts of poverty accumulation (Biewen 2009, 2014), an understanding of the causal mechanisms of micro-level poverty dynamics and economic disadvantage is indispensable.

Therefore, this paper examines the various drivers of IWP risk from a longitudinal perspective, assessing the importance of household employment for different social classes and examining the extent to which pre-existing poverty determines future poverty and for whom. As differences in these dynamics exist across countries, embedded in different institutional contexts (including welfare programs) and labor market settings (Giesselmann 2015; Lohmann and Crettaz 2018), we analyzed fourteen Western European countries (Esping-Andersen et al.2002; Callens and Croux 2009) and investigate whether the individual and household level drivers of IWP are universal across such contexts.

In-work poverty is defined as a combination of individual and household conditions (Lohmann 2018; Raitano et al.2019). Individuals are considered at risk of in-work poverty if they report employment as their main activity status for more than six months of the reference year and live in a household with an equivalized disposable income of less than 60% of the national median.² This implies that an individual who is employed for a considerable portion of the year can be considered working poor regardless of their individual wage, depending on their household size and the earning capacity of other household members. Moreover, individuals are not considered in-work poor if they fail to satisfy the employment criterion, due to long spells of inactivity or unemployment, regardless of their own or their household's economic conditions. Therefore, while IWP is a substantively useful concept, the combination of the employment situation of the individual and the economic situation of the household makes it analytically complex. Individuals' transitions into and out of IWP are jointly conditioned by their employment status and household income, and changes in aggregate IWP levels also depend on the composition of the workforce at a given point in time.³

The existing literature has extensively discussed the various trends and catalysts of IWP, and in recent years, there has been a substantial number of socioeconomic analyses produced on this subject (Kalugina 2013; Giesselmann 2015; Lohmann and Crettaz 2018; Lohmann and Crettaz 2018; Van Winkle and Struffolino 2018; Filandri and Struffolino 2019; Polizzi et al.2022; Vaalavuo and Sirniö 2022; Wolf et al.2022; Ratti 2022). This research has highlighted the relevance of individual, household, and macro-level contexts.

Regarding individuals’ situations, employment discontinuity (Halleröd et al.2015), ‘underemployment’, and non-standard employment situations (Horemans et al.2016; Tamayo and Popova 2021) have proven to be relevant drivers of poverty, suggesting that labor market deregulation and dualization (Emmenegger et al.2012), the spread of employment precarity (OEDC 2014; Barbieri et al.2019) have contributed to the rise in IWP. Temporary, part-time employment and low wages have also been reported as systematically overrepresented among the in-work poor (Lohmann and Crettaz 2018; Ratti 2022), though low-paid work and IWP are separate phenomena that only partially overlap since poverty includes economic resources at the household level (Andreß 2018 [2003]; Lohmann and Marx 2018). For similar reasons, gender differences in IWP have often shown counterintuitive patterns, with women, even if marginally employed, having lower IWP risks than men precisely because they are more often secondary earners (Van Lancker 2012).

Therefore, household-level characteristics are crucial for defining poverty risks, both in terms of household members and in terms of the overall labor supply. Previous literature has documented the growing importance of multiple earners (Brady et al. 2020) and the decreasing capacity of the traditional single-earner family to protect against poverty (Gioachin et al.2023), although increasing women's participation alone might not offer a way out of poverty (Filandri and Struffolino 2019). In addition to household's labor supply (Andreß and Lohmann 2008), occupational class positions also determine household poverty risks and the extent to which a second income is needed to avoid poverty. The stratification of poverty risks by social class has been well-documented (Whelan and Maître 2010; Vandecasteele 2015; Albertini et al.2020), and notwithstanding the idea that risk dynamics would spill over to the previously sheltered middle class (OECD 2019; Ranci et al.2021), recent research confirms the persistent concentration of risks in the least advantaged social groups (Brandolini et al.2018; Gioachin et al.2023). While this research has focused on poverty, still little is known about the class-based stratification of in-work poverty.

Different levels of IWP across countries have also been well-documented (Lohmann and Crettaz 2018), with previous literature highlighting the relevance of macro-level determinants in structural and institutional terms, represented by the labor market's composition, functioning, and related earning distributions (Salverda and Haas 2014; Giesselmann 2015; Stepanyan et al.2020), as well as by welfare generosity, in particular income support schemes (Van Lancker 2012; Lohmann and Crettaz 2018; Tamayo and Popova 2021). Vandelannoote and Verbist (2020) recently showed that in-work benefits could have a relevant impact on reducing poverty, though their effects differ across countries. Variation across institutional settings exists not only regarding the level and impact of different types of employment on poverty risks (Callens and Croux 2009), but also on the living standard of the poor (Wolf et al.2022).

While the incidence of IWP and its distribution have been relatively well-discussed from a cross-sectional perspective, the underlying dynamics and relevance of changes in household employment (Vandecasteele and Giesselmann 2018), the social gradient of IWP risks, and the persistent causal consequences of IWP on future economic situations (Brady et al.2020; Mussida and Sciulli 2023) have received significantly less attention. The importance of employment conditions and demographic factors has been underlined by various studies, although they have not focused on IWP (Callens and Croux 2009; Vandecasteele 2015, Polizzi et al.2022). Regarding the causal impact of previous IWP experience, Biewen (2009) emphasized the importance of considering heterogeneities and feedback effects when assessing state dependency and showed that, in Germany, about a third of poverty persistence can be attributed to causal state dependency. Tejero (2017) is among the few to provide empirical evidence for in-work poverty, confirming causal state dependence in Spain.

Building upon the work of previous studies, this paper analyzes the trends and determinants of IWP across fourteen Western European countries. It examines the dynamics of social stratification of IWP in different institutional contexts following a research tradition that aims to identify mechanisms of cumulative disadvantage as drivers of social inequalities and stratification (Nolan and Whelan 2011; DiPrete and Eirich 2006; Vandecasteele 2010, 2011). We expected most drivers of IWP-dynamics to operate at the micro-level and characteristics of the macro-institutional context to be more relevant in explaining the differences in aggregate levels of IWP in different national contexts. Therefore, our investigation centered on the individual- and household-level factors influencing IWP and we do not propose an analysis of the macro-level drivers of IWP. We limited our comparative analysis to Western European countries.

We document the incidence of IWP in various sociodemographic groups and occupational social classes and investigated the relevance of household composition, changes in families’ employment conditions, and overall labor supply for IWP. We also examined the details of individuals’ employment situations – including work contracts and the conditions of low-wage and part-time positions – and the role of these arrangements when considering household employment patterns. Our results stress the importance of both the quantity (how many household members are working) and quality (the terms of their work contracts) of households’ labor supply as crucial factors influencing risk exposure, longitudinal accumulation, and stratification consequences.

The second contribution of our study relates to the assessment of the causal effects of previous IWP exposure for current IWP risks.⁴ Understanding the distinction between causal state dependence and structural drivers leading to poverty persistence is crucial for comprehending the dynamics and ‘functioning’ of IWP and devising policies to combat poverty. Although much attention has been given to welfare transfers as a means of alleviating economic strains on families – and, indeed, net social transfers are crucial instruments in reducing poverty risk (OECD 2009; Brady et al.2010), the efficacy of these policies in reducing poverty diffusion depends on the extent to which poverty is causally driven by previous poverty experience or other structural factors related to the individual and family. We examined the longitudinal persistence of IWP and disentangled the high levels of descriptive inertia of IWP over time from the dynamics of IWP related to mechanisms of causal or: genuine state dependence (GSD) and their possible interplay with unobserved factors that may account for IWP inertia. Our results show that, for large portions of the population, causal mechanisms of poverty entrapment play only a minor role, while the reiteration of risks tends to concentrate on a small portion and is associated with (unobserved) factors at the individual and household level.

We analyzed EU-SILC longitudinal data (Eurostat 2022, release 1; Wirth and Pforr 2022) for the period of 2004 - 2019⁵ for fourteen Western European countries representing a broad set of institutional, welfare, and labor market features. These countries included Austria, Belgium, Denmark, Greece, Spain, Finland, France, Ireland, Italy, the Netherlands, Portugal, Sweden, and the United Kingdom, plus Germany for those analyses that do not control for social class. The data include information on individual and labor market-related characteristics and disposable incomes (before and after social transfers), which allowed for household features to be reconstructed in terms of composition, the presence of children, and employment patterns. Being at risk of in-work poverty was measured on the basis of equivalized net household income (HY020 in EU-SILC data), adjusted by the modified OECD equivalence-scale accounting for household size after welfare transfers (pensions included). ‘At risk of poverty’ was defined as living in a household whose income falls below 60% of the national median income. To satisfy the employment condition for IWP, the individual must have been employed for more than six months in the reference year, though not necessarily consecutively (Table A1 in the Appendix provides the descriptive statistics).

While this definition of IWP excludes those with the least labor market attachment, it still includes individuals with unstable careers in the risk set. Limiting the sample to individuals continuously employed for the entire year would underestimate IWP and hide the heterogeneity across social groups characterized by sizable differences in exposure to unemployment or spells of inactivity. In line with Eurostat, our analytical sample included men and women older than 18 who reported employment (either self-employed⁶ or wage labor) as their main activity in the reference year for at least seven months.⁷ A limit was set at age 65 in defining the number of employable adults at the household level. As we were specifically interested in IWP dynamics over time, we used longitudinal information, which was, in principle, available for up to four successive years (around 50% of the observations in the total sample belonged to units followed for the entire time window due to the functioning of rotational groups). Only observations with at least two timepoints were considered. Table A1 in the Appendix reports descriptive statistics of the analytical sample and Table A2 lists the number of cases of the analytical sample.

We began by examining the distribution of IWP risks and persistence across national contexts and over time (with longitudinal weights applied) and then analyzed the role of household labor supply for IWP risks.⁸ We did so by utilizing a set of country-specific random-effects (RE) panel regressions in the form of linear probability models (LPM). All models controlled for bi-annual period effects, gender, education (compulsory, secondary, and tertiary), age range (18–25, 26–35, 36–45, 46–55, and 56–65), number of household members, and the presence of young children (up to the age of three). Overall household labor supply was considered in terms of the number of actual workers among household members (the ‘extensive margin’).⁹ We also observed changes in the household total labor supply by considering shifts from one to two employed components between t − 1 and t (around 18% experienced some change in employment during the observational window). We then complemented household labor supply with variables detailing individual working conditions and contractual arrangements where we aggregated fixed-term contracts, part-time (either self-reported or those who regularly worked less than 30 h per week), and low-wage (below 60% of the national median) employment under the umbrella term of ‘non-standard employment’ (NSE). We had to collapse these different categories since contractual information was not uniformly available across the national datasets, and in the four countries with register-based data (Denmark, Finland, Sweden, and the Netherlands – see footnote 8), we could not properly distinguish these three states for individuals other than the principal respondent, which led to an underestimation of households with NSE in these countries. This is not ideal,¹⁰ but extending the analysis to the dynamics of ‘non-standard’ (disadvantaged) working conditions also on the household level for all countries still offers relevant insights, especially in shedding light on the negative consequences of ‘labor market dualization’.¹¹

Household employment patterns were defined according to a combination of the number of earners (one, two or more) and the presence of NSE in one or more household members over two timepoints, which yielded four distinct categories: (1) single-income households at time t (ref.cat.); (2) two workers in both timepoints; (3) a shift from one to two workers, both in ‘standard’ employment; and (4) a shift from one to two workers with at least one of them in NSE. This allowed us to assess the extent to which longitudinal variation in household labor supplies shapes IWP risk compared to alternative household work arrangements.¹²

Social class remains a powerful indicator of poverty. Households’ occupational social class position was measured based on an aggregated version of the European Socioeconomic Groups (ESeG, Rose and Harrison 2007)¹³ classification and defined according to the dominance criterion among earners, which considers the highest position present in the household. We categorized social classes as follows: (1) managers and professionals, combining EseG groups 1 and 2; (2) technicians, clerks, and skilled service occupations, combining EseG groups 3 and 5; (3) skilled industrial occupations, encompassing EseG group 6; (4) unskilled occupations, classified as EseG group 7; and (5) self-employed, comprising EseG group 4. We accounted for interactions between household employment patterns and the household's dominant social class position to examine class-specific variations of IWP risks and changes in the class gradient of IWP exposure associated with variations in household employment patterns.

Alongside a vector of observable covariates $Zit$⁠, the model controls for the initial condition of IWP $yi0$⁠, the initial conditions of time-varying covariates $wi0$⁠, and their means $w¯i$⁠. The IWP measure in the previous year $yi,t−1$ can be interpreted as the causal or ‘genuine’ state dependency under the assumption that time-constant unobserved heterogeneity in this specification is captured by controls for initial conditions – an assumption that has been widely accepted in the existing literature (Wooldridge 2005; Rabe-Hesketh and Skrondal 2013). Moreover, to better grasp the functioning of IWP inertia, the model includes the interaction between the component associated with unobserved heterogeneity (i.e. initial condition: $yi0$⁠) and GSD (i.e. IWP one year before:$yi,t−1$⁠). In this way, one can distinguish the effect of being in IWP at t − 1 for those already in IWP at the observed t0 and for those not in IWP at t0. In so doing, we could detect group-specific patterns of accumulation of IWP over time. Insofar as the initial condition at t0 captures time-constant unobserved factors associated with IWP risks (Cutuli and Grotti 2020), we read the interaction between the initial condition and GSD as a sign of multiplicative dynamics of protracted or reiterated experiences of IWP.

Regardless of the assumption concerning $yi0$ representing a proxy for time-constant unobserved heterogeneity, the positive interaction$yi0∗yi,t−1$ nevertheless provides a clear picture of heterogeneity in GSD and entrapment dynamics across population segments with processes of risk accumulation fanning out over time.

Regarding the determinants of IWP risk, we found evidence of well-established patterns concerning group differences (Lohmann and Crettaz 2018). Table 1 provides a descriptive account of IWP and its dynamics in our sample. Our results also confirm relevant cross-country heterogeneity in overall IWP levels, with Southern European countries (Italy, Spain, Greece, and Portugal) exhibiting higher rates compared to Continental (Austria, Belgium, France, Germany, and the Netherlands), Anglo-Saxon (Ireland and the UK) and Northern European (Denmark, Finland, and Sweden) countries. IWP rates were relatively stable or slightly increased over time in most of these countries (see Table A1 in the Appendix, which includes details for individual subgroups). However, these changes did not follow macro-level economic conjunctures. This finding is not surprising given the relative nature of the measure and, more importantly, the changes in the population at risk during economic downturns— as the least-advantaged lose their employment during economic downturns, IWP risks might decline. While women and young people have been shown to be generally overrepresented among low-wage earners (Maitre et al.2012; Lohmann and Crettaz 2018), they were not prevalent among the in-work poor. This seemingly contradictory but well-documented (Van Lancker 2012) observation depends on the very definition of IWP, as mentioned above. As working women are more often second earners and younger people tend to still live in their parental home, their poverty risks are limited by virtue of their household situations. In our results, the underrepresentation of women was more visible in Southern Europe, where women's employment tends to be lower. These findings highlight that when dealing with IWP, national specificities in terms of group composition can mirror diverging dynamics of (self-)selection into employment, with different gender- and age-related employment gaps occurring in distinct contexts (which may require differently targeted policies). The most relevant drivers of IWP were individuals’ contractual situations, low pay, and household-level employment patterns. Additionally, occupational social class positions were a powerful indicator of IWP risks (see also Figure A1 in the Appendix) through their relevance for general poverty (Vandecasteele 2011, Gioachin et al.2023) and income distributions (Albertini et al.2020). IWP risks were much higher among the working classes (esp. unskilled workers, group 4 – EseG 7) independent of the specific contexts, shedding doubts on the idea that the middle class had become more vulnerable due to structural changes related to globalization or technological innovations (Pressman 2007; OECD 2019; Ranci et al.2021). There were also higher risks for the self-employed, who generally form a highly diverse group; however, it is also probable that the income of self-employed individuals is underreported (Hurst et al.2014).

More interesting is the evidence from a dynamic perspective. As indicated in Table 1, the risk of entering IWP was rather moderate in all examined contexts, while relevant country differences instead emerged in the persistence of IWP over subsequent years. Even if the identification of macro-level factors and institutional determinants of IWP was beyond the scope of this article, this evidence points to the relevant role of welfare states (in terms of their decommodification capacities, targeting, and effectiveness of welfare transfers) and labor market functioning for IWP flows.

In all contexts, the amount of household labor supply played a relevant role in the dynamics of IWP. Single-earner households displayed the highest entry and persistence rates of IWP (around 6% and 60%, respectively), while dual-earner arrangements showed modest risks, in terms of reduced entry, though this was primarily through less-persistent IWP. Entrapment was lowest among those whose households increased their labor supply, moving from single to dual-earner arrangements (labeled ‘added worker’ households), and an additional worker reduced IWP risks even if one of the workers was employed in an NSE position (labeled ‘added worker 1 non-std.’).

Table 1 provides additional information regarding the longitudinal dynamics of IWP, distinguishing the above figures based on the initial condition – the IWP situation at t0. Although only a small segment of the population was observed in IWP at t0, those who were in such a condition showed much higher persistence or successive (re-)entry rates than those with the same household employment pattern but not in IWP at t0. This suggests long-lasting consequences of individual and household-level characteristics (here captured by the initial condition) that yielded higher IWP exposure over the observation period, and do so independently from other observable factors. Table A3 in the Appendix provides a more fine-grained representation of IWP trajectories, showing that more than 95% of workers not in IWP in the first available wave experienced a maximum of one spell of poverty in the four-year time window, while around 60% of workers who experienced IWP as an ‘initial condition’ tended to accumulate at least three spells of poverty in the same time frame. This evidence suggests a pattern of risk accumulation over time – which is why we analyze the interaction between the magnitude of net IWP entrapment and the initial condition in the last part of this paper.

In summary, our results show that increasing the household labor supply and its work continuity are effective means of combatting IWP and that dual-income household setups are likely capable of partially reducing the class-based stratification of social risk.

So far, we have demonstrated that IWP is a persistent and strongly stratified phenomenon, though having an additional income in the family substantially lowers risks. We now shift our focus to examining the extent to which current experiences with IWP causally increase the likelihood of experiencing poverty in the future. A scarring effect in terms of genuine state dependency dynamics could be considered a net driver of IWP stratification as well as a potential tool for policies contrasting poverty accumulation by (timely) breaking poverty chains through welfare transfers. Conversely, if the accumulation and stratification of IWP are rather due to more stable and structural characteristics of individuals and households (here captured by y_t0) and not due to previous IWP as such (here measured by y_t−1), this would cast doubt on the efficiency of passive policy measures like welfare transfers. This part of the analysis relaxed the assumption of initial unobserved heterogeneity and previous IWP exposure as pure additive factors and, instead, tested for multiplicative dynamics between (un)observed factors and poverty entrapment as potential mechanisms of cumulative disadvantage (thus inserting the interaction y_t0*y_t−1).

Looking at the large segment of the population that did not report IWP as their initial condition (y_t0 = 0), the marginal effects of previous spells of IWP – that is, the differences between predicted probabilities of IWP for those in IWP in the previous year (y_t−1= 1) and those not in IWP (y_t−1 = 0) – were negligible in the Nordic countries (Denmark, Finland, Sweden) and the Netherlands, relatively small in the Continental and Anglo-Saxon countries (Austria, Belgium, France, Germany, Ireland, and the UK), and only slightly higher in the Southern European countries (Italy, Spain, Greece, and Portugal). Put differently, at least for individuals and households with limited risk of facing IWP (i.e. not in IWP as an initial situation at t0), there seems to have been little GSD in poverty entrapment for most national contexts. For these individuals, the rare occasional spell of IWP (entry rates in IWP have been shown to be low everywhere) was indeed associated with a rather small or non-existent increase in subsequent spells of IWP. The situation changed for those reporting IWP in their first available observation (y_t0 = 1). This small and selected group of people (Table 1) not only showed a higher overall risk of IWP, as expected, but also relevant differences according to the situation one year before, suggesting carousels into and out of IWP and clarifying the presence of a causal effect of previous IWP (y_t−1 = 1) in determining actual IWP risk.

This implies that, after accounting for potential GSD, IWP persistence is significantly associated with unobserved individual (and household) attributes as well as with broader structural (seemingly time-constant) conditions, altogether captured by the initial condition at t0. Notably, despite clear heterogeneity in the level and magnitude of the interaction between IWP at t0 and t – 1 in all countries, the risk of IWP was particularly high in instances of the co-occurrence of IWP as an initial condition and previous IWP exposure.¹⁷ Therefore, stratification of IWP can be seen as a multiplicative result of the interaction between the net dynamics of GSD and pre-existing individual and household-level risk factors.

All in all, when GSD dynamics came into play, they did so especially for selected disadvantaged segments of the workforce already characterized by a higher IWP risk. Policies supporting poorer individuals and households through transfers aimed at stopping poverty chains can be effective insofar as (and limited to the workforce segments for whom) IWP persistence is largely attributable to net GSD dynamics. In a similar vein, a larger component of IWP diffusion and persistence was also associated with transitions between spells of poverty and non-poverty, especially among those with unfavorable initial conditions. This evidence carries significant policy implications, which we discuss in the following section.

In-work poverty has received increasing attention in the socioeconomic literature as well as in political debates (Peña-Casas et al.2019). While the concept of IWP is descriptively suggestive, analytically it has been a matter of debate. Furthermore, evidence on specific dynamics and causal drivers of IWP dynamics has been limited. This paper addresses this gap by providing empirical evidence to help clarify the dynamics of IWP within fourteen Western European countries for the years 2004–2019 based on EU SILC longitudinal data. Empirical evidence on the drivers and causes can, and probably should (Smith 2022; DiPrete and Fox-Williams 2021), provide useful to the task of reducing inequalities and social risks.

The dynamics and stratification of IWP turned out to be rather stable over the investigated time period and independent of business cycle changes. The cross-country comparison confirmed well-established differences in levels of IWP, especially regarding the capacities of different countries to protect single-earner households and lower social classes from poverty. However, this comparison also showed that the stratification of IWP risks and the role of household work arrangements were very similar across these different contexts. Identifying the relevance of specific macro-contextual drivers was beyond the scope of this study and is left for future research. The same applies to the extension of the geographical area beyond the focus on Western Europe. Another limitation of this study concerns the short time window of the longitudinal information included in the EU-SILC data, which precludes a more detailed analysis of duration-dependent and accumulation-driven dynamics. Nevertheless, we argue that this study still provides relevant contributions.

As indicated by our results, IWP in contemporary Western Europe is a strongly stratified phenomenon that tends to concentrate and accumulate over time among specific social groups and occupational class positions, with the working class remaining the most exposed to IWP risks. Household work arrangements that included a second income protect against IWP, especially those households from less-advantaged social classes and in contexts characterized by higher overall IWP levels, such as Southern Europe. Increasing continuous labor market participation among household members thus appears to be an effective means of combatting IWP exposure and the longitudinal accumulation of risk over time. Moreover, dual-earner household arrangements have the potential to drastically improve social equality – not only between men and women but also across social classes – by significantly reducing poverty risks among less advantaged families: the presence of an additional salary can partially compensate for the strong class-differences in IWP risk. However, increasing the proportion of dual-income households may also have implications for the overall income-distribution, which in turn may heighten the relative poverty risk faced by single-earner households, particularly single-headed households. These households structurally lack the possibility to increase labor supply, which makes other means of poverty reduction necessary (Hick and Marx 2022). Overall, however, our results show how the debate on poverty reduction might profit from a social stratification perspective with a focus on low-educated women, who often are not-employed and partnered with low-qualified working-class men. The labor market participation of these women must be raised in order to consistently keep their families and children out of poverty over the long term.

Our results partially support the notion that any additional employment is better than no employment, given that (additional) NSE situations can alleviate the risks of poverty (Gebel and Gundert 2023) – something that has been presented (and criticized) in the literature under the label of ‘equality-employment’ trade-off (Esping-Andersen and Regini 2000; Barbieri 2009). Taking a closer look, however, poverty risks are reduced once a certain continuity of employment is guaranteed for all earners, underlining the necessity of providing reliable employment conditions, ensuring decent wages, and preventing or limiting household-level accumulation of precarious labor market conditions.

A major contribution of this paper is its focus on the dynamics of IWP. Descriptive evidence indicates that IWP tends to persist over time, leading to adverse outcomes in terms of the accumulation of socioeconomic disadvantages. The magnitude of IWP inertia is different across countries, with higher degrees of persistency in contexts with higher overall IWP rates, such as in Southern European countries. We also tested for the presence of a causal effect of previous IWP experience, genuine state dependency (GDS), and its conditionality on unobserved characteristics captured by the initial condition, following a well-established approach in the literature (Wooldridge 2005). In fact, the longitudinal accumulation of IWP does not respond merely to mechanisms of GSD in large segments of the population. Rather, it stems largely from the interplay between poverty entrapment and the disadvantaged initial conditions associated with it. This comes with relevant policy implications.

Insofar as the high inertia of IWP risk can only partially be traced back to the causal mechanisms of poverty traps, welfare transfers alone cannot avoid future individual poverty exposure and dissemination. Thus, rather than focusing on GSD risk and transfer-based measures conditioned on current IWP conditions, policies should focus on supply-side endowments and demand-side factors structurally associated with individual employability and sustained work intensity.¹⁸ From the perspective of social investments, investing in education seems to be the primary viable means of improving continuous employment and access to qualified employment. Policies aimed at increasing dual-income households could help reduce the social class gradient in terms of risk exposure; adding an additional household income implies bringing more (working-class) women into ‘reasonably contingent’ and decently paid work. Labor policies fostering the demand for (female) labor as well as interventions helping to reduce structural and economic barriers to female labor-market participation are essential. Childcare services play a significant role in this regard, and accessible public childcare has recently been shown to assist low-educated mothers in accessing gainful employment (Scherer and Pavolini 2023).

More employment opportunities, better wages, and more stable career prospects would contribute to lifting families out of poverty (and, therefore, assuring better conditions for future generations). At least in the short term, employment growth might take the form of an increase in flexible (yet not casual or marginal) NSE. To protect against poverty, however, NSE must entail a certain degree of stability, offer reasonable pay levels, ensure access to reliable safety nets, and must not accumulate within households. Finally, our evidence also suggests the opportunity to organize welfare policies as follow-up procedures tailored to the workforce segments most at risk of poverty and with recent past poverty exposure, regardless of the contingent conditions of IWP at each point in time. In other words, we do not expect welfare transfers to be effective if they are not accompanied by policies aimed at reducing the structural basis of poverty risks through incentivizing individual employability and dual earners’ arrangements. Similarly, our results indicate that policies intended to prevent longitudinal accumulation of IWP should not be primarily conditioned on contingent poverty conditions but rather organized under the form of mid-term follow-up measures devoted to individuals and households with weak labor market attachments. To be clear, we do not imply that social welfare transfers would not be useful in alleviating economic strain. Previous research has documented that growing up in poverty entails severe and long-term negative consequences, and our results document adverse multiplicative effects of unfavorable initial conditions and causal effects of previous IWP, with clear consequences in terms of inequality and crystallization of social stratification. Therefore, welfare transfers are necessary to mitigate the social stratification of risk and help reduce (long-term) costs for the state. However, these transfers alone will not be sufficient to improve the mid-term socioeconomic prospects of recipients and prevent the accumulation of economic disadvantage over the life-course.

No potential conflict of interest was reported by the author(s).