Labour market prospects of young adults in Europe: differential effects of social origin during the Great Recession

ABSTRACT

Research on the direct effect of social origin (DESO) focuses on how background influences later labour market outcomes after accounting for education. Growing up in a household of low social origin might decrease the chances of certain future outcomes; however, the extent to which this matters is contingent on the economic cycle. Using the EU Statistics on Income and Living Conditions (EU-SILC) and the European Social Survey (ESS) between 2002 and 2014, we analyse whether the gap in the DESO in terms of employment and earnings widened following the Great Recession for young adults (25-34) in France, Germany, Italy, Poland, Spain and the United Kingdom. Our results suggest that young adults of high social origin faced more disadvantages in terms of employment than young adults of low social origin in France, Spain and the United Kingdom. On the other hand, analyses show that young adults of low social origin experienced more disadvantages in terms of earnings than their counterparts of high social origin in Spain.

KEYWORDS:

• Intergenerational transmission
• Great Recession
• transition to adulthood
• DESO

Introduction

The Great Recession was the largest economic crisis since the Great Depression nearly 90 years prior, leaving deep scars across the global economy. The economic downturn in 2008 caused GDP growth to plummet, led to a surge in unemployment rates and struck not only the construction sector but also other sectors that were considered a safe haven for workers in the past decades (Grusky et al. 2011). While the economic crisis reached almost everyone, young people were among the major losers (Bell and Blanchflower 2011).

In this context, when good jobs are rare and employers are more selective, social origin may become more important, as it can mitigate hardships generated by the economic crisis (Macmillan 2014; Zwysen 2016). This means that young adults of high social origin may benefit more from their parents’ economic and social capital than young adults of low social origin, even after taking education into consideration (Breen 2004; Ballarino and Bernardi 2016). However, an opposite perspective is possible as well. Similar to the course of events during the Great Depression, an economic crisis may completely reshuffle cards (Piketty 2013; Piketty and Saez 2013). This article examines whether the direct effect of social origin (DESO) in terms of employment and earnings widened or contracted following the Great Recession.

Our article thus joins a handful of studies that try to understand how young adults fared during the Great Recession (Genda et al. 2010; Aassve et al. 2013; Cozzolino et al. 2018; Sironi 2018; Curry 2019; Schoon and Byner 2019; Salazar et al. 2020). To date, no research has investigated whether the DESO of young adults changed after the 2008 financial crisis. Substantively, this is important for two reasons: prior evidence shows that early career unemployment has long-lasting scarring effects (Gregg and Tominey 2005; Gangl 2006), and starting a career at a lower paying employer may generate less mobility thereafter (Kahn 2010; Oreopoulos et al. 2012).

The study of the DESO is well established in sociology. Even after accounting for education, adults of lower social origin tend to have lower employment levels and earnings (Hällsten 2013; Ballarino and Bernardi 2016; Passareta et al. 2018). Previous research ascribes some of these labour market disparities to two major elements. First, economic capital: financial resources can sustain young adults while looking for a job that matches their qualifications (Wiborg and Hansen 2009), a process that requires a longer search period during a financial crisis. Second, social capital: families of high social origin often possess more influential social networks than families of low social origin (Corak and Piraino 2011; Oesch and von Ow 2017).

This article uses two data sources: the EU-SILC (rounds 2005 and 2011) and the ESS (all rounds between 2002 and 2014). Our focus is on young adults aged between 25 and 34 in six large European countries, namely, France, Germany, Italy, Poland, Spain and the United Kingdom. We estimate a series of regression models separately for each country that quantify whether the direct effect of social origin changed in terms of employment and earnings between young adults of low and high social origin between 2002 and 2014.

This article first gives an overview of the literature on social mobility. Next, it reviews how young adults fared during the Great Recession. It then discusses the theoretical arguments for or against an increase in the DESO during the Great Recession. Then, it describes the disparities in economic fluctuation across countries. We then present the data and elaborate the analytical methods used in this article before showing and discussing our results.

Social mobility

In the wake of World War II, sociologists proposed several hypotheses about social mobility. The modernisation hypothesis advocates that industrial societies are becoming more equal due to educational expansion and industrialisation (Dahrendorf 1959; Lipset 1960; Bell 1973). These scholars argued that educational institutions would be influential in allocating individuals to jobs according to their expertise and skills in the labour market. They also expected that educational expansion would reach most individuals, as economic growth and democratic pressure prioritise the focus on human resources. Other hypotheses inspired by Marxism expected that the importance of class division in social mobility would not only persist but possibly grow larger over time (Braverman 1974; Wright 1989).

Within the growing body of research, two large comparative studies seem particularly noteworthy. Based on surveys collected in the 1990s, Erikson and Goldthorpe (1992) showed large differences in absolute class mobility rates among 12 industrial countries. They argued that these disparities were possibly due to the variation in the class structure across countries and its development over time. On the other hand, while they showed differences in relative mobility rates across countries and over time, these differences were small. The second main comparative study found that the patterns in absolute mobility flows became more similar among most countries in the 1990s than in the 1970s (Breen 2004; Breen and Luijkx 2004). However, they reported greater variations across countries in relative rates than did Erikson and Goldthorpe (1992). While these differences in relative rates tended to converge over time, they did not appear to have any systematic relationship with economic development (Breen and Luijkx 2004).

While the effect of origin on destination is mediated in large part through education, as shown in a previous study (Blau and Duncan 1967), there is still a residual effect of origin on destinations after controlling for educational attainment. One might expect that the direct effect of social origin would decrease following the expansion of accessible education in the past several decades, but this effect on income and socioeconomic status seems to persist in most countries over time (Ballarino and Bernardi 2016). Previous research in the United States suggested that the DESO is more pronounced among low educated individuals than among individuals with college degrees (Hout 1988). However, later evidence challenged these findings and found little or no difference in the DESO among different educational groups in France (Falcon and Bataille 2018), Spain (Bernardi and Gil-Hernández 2021), the United Kingdom (Laurison and Friedman 2016) and the United States (Zhou 2019).

While there might be some debate regarding whether the DESO is heterogeneous between low- and high-educated individuals, there is a consensus that this effect seems stable and even increases in several countries over time (Ballarino and Bernardi 2016). The focus of this article is the question of whether the DESO changed after the 2008 financial crisis in the six largest countries of the European Union: France, Germany, Italy, Poland, Spain and the United Kingdom.

Previous findings show no time trend in the DESO on socioeconomic status in Germany, Italy, Spain and the United Kingdom. In contrast, in France, the DESO increased over time (Ballarino and Bernardi 2016), while in Poland, it declined (Meraviglia 2017). With respect to the DESO on earnings, there seems to be no change in the United Kingdom over time (Ballarino and Bernardi 2016) and a U-shaped relationship in France, where this effect decreased between 1977 and 1985 and then rose (Bouchet-Valat et al. 2016).

The Great Recession and labour markets for young people

The 2008 financial crisis worsened the labour market conditions of young adults more than it did for older workers, as the former are more sensitive to the business cycle (OECD 2010). This is due to three mechanisms. First, the ‘experience trap’ is a challenge, as most young adults enter the labour market with no prior experience. As most jobs require a minimum amount of experience, this can lead to a vicious cycle of unemployment early in the career. Second, know-how in the job search favours older workers over younger workers, as the former are often more efficient in finding a job. Third, young adults have lower financial commitments (e.g. loans, child expenses) than older adults. This puts less pressure on young people to find employment, and they can also be supported for a while by their parents (Bell and Blanchflower 2011).

While the labour market conditions of young people improved in the early 2000s, this optimism disappeared following the 2008 financial crisis. Unemployment was higher among young adults than among all older cohorts. The repercussions of the economic crisis also extend to the type of employment: young workers are more likely to be faced with precarious employment than older workers. For example, part-time jobs and zero-hour contracts were more prevalent among young adults in some European countries, such as Italy and the United Kingdom, following the 2008 financial crisis than in the period that preceded it (Eurofound 2016; Prassl 2018).

Following the Great Recession, some governments implemented new schemes to enhance the difficult labour market situation of young people (OECD 2009; Bell and Blanchflower 2011). However, they did not last very long in many European countries because debt rose sharply, consumption plunged, budget deficits increased, and GDP decreased following the 2008 financial crisis. In an attempt to fight budget deficits, the EU, the IMF and the ECB pushed for an agenda that included austerity measures and tax increases. Young people were not spared from these new measures, and some governments cut support for students and increased tuition fees, such as in the United Kingdom and Spain (Theodoropoulou and Watt 2011; Antonucci and Hamilton 2014). Needless to say, these retrenchments further aggravated young individuals’ labour market conditions.

A comparative study for the period between 2000 and 2010 found that the likelihood of being poorly paid among young adults increased following the Great Recession in Spain, the United Kingdom and the United States, but this effect was less pronounced in Norway and almost absent in Germany (Sironi 2018). Other studies show either no evidence of a strong impact of economic recessions or procyclical findings – an association in which inequalities decrease when the economy declines. As noted by Pöyliö (2020), inequalities in college enrolment among young adults of different social origins decreased following the Great Recession in the United States. Another study analysed whether the Great Recession increased the co-residence of young adults with their parents due to financial difficulties (Aassve et al. 2013). Although the youth experienced more unemployment and financial hardships, co-residence notably increased in only Hungary, France and Sweden out of 24 countries.

Economic recessions and employment

There are different mechanisms at play that can influence the employment behaviour of young adults of different social origins during an economic recession. Young adults may benefit from their parents’ social capital. People tend to choose friends who resemble themselves. In this logic, one would expect people in higher class positions (e.g. managers or professionals) to befriend one another. This means that families with high social origin are more likely to know about openings of high-prestige jobs. Consequently, they can indirectly influence the process of selection to favour their children. Since one-third to half of all jobs are obtained through informal contacts in advanced economies (Granovetter 1995; Pellizzari 2010), it is expected that social capital is important to obtain a good career start. On the other hand, families of low social origin might struggle to mobilise their friends to help their children, as their network contacts are generally not in influential positions (Corak and Piraino 2011; Oesch and von Ow 2017). As much as these social dynamics are important during an expanding economic cycle, they probably matter even more during an economic recession, as employers raise their hiring standards and become more selective (Reder 1955; Devereux 2002). Subtle favouritism can also be practised by employers during that period. They can perceive direct or indirect signals from the interviewees’ behaviour, such as their accents and etiquette, that are more present among circles of high social origin (Friedman and Laurison 2019).

Economic capital may play a protective role during a financial crisis. Young adults of high social origin are less likely to quickly accept a job that does not match their skills, as they can be sustained financially by their parents (Bell and Blanchflower 2011). On the other hand, financial pressure might incentivize young adults from disadvantaged backgrounds to accept any job during an economic recession. A study in the United States found that wealth inequalities increased in favour of highly educated people following the 2008 economic recession. It shows that the median wealth of households whose heads had a BA degree fell by 2011–70% of the 2003 level compared with 19% in households whose heads had no high school education (Pfeffer et al. 2013). Even beyond the role of parents, the labour market supply could change during an economic recession if people cohabit or marry partners with similar social origin, education or social class, as can occur under the influence of homogamy (Kalmijn 1991). Consequently, if one partner experiences job loss or financial difficulties during an economic recession, the other partner might step in and become employed to support the household. Previous research supports this notion and reports that wives entered the labour market in response to their husbands’ job loss following the Great Recession in the United States (Mattingly and Smith 2010). This increase in labour market supply is likely to be more common among young adults of low social origin, as they experience more financial strains during economic recessions.

There are other counterfactual outcomes to employment during economic recessions that can be advantageous to young adults, such as education. Previous research suggests that enrolment rates in education are countercyclical (Dellas and Koubi 2003). That is, people are more likely to pursue education during economic recessions. This is likely to occur because economic downturns lead to higher job loss and decrease new employment openings. However, the opportunity cost of pursuing further education is likely to be different between young adults of different social origins. Education might include costs of tuition, housing, and books. During economic recessions, these financial burdens are more likely to hinder the enrolment – or increase the drop-out - of young adults of disadvantaged background compared with young adults of advantaged background. Moreover, parents of low social origin might face a reduction in earnings or experience a layoff at a higher rate than parents of high social origin. This could put more pressure on young adults of low social origin to find employment and contribute to family income. Empirical evidence in the United States supports this claim, where it examined the period between 1968 and 2000 and reports that the rate of college enrolment among young people was lower in households that expected to have lower incomes (Christian 2007). Other evidence from Italy reports that the drop-out probability increased for students with disadvantaged backgrounds following the Great Recession (Adamopoulou and Tanzi 2017).

Young adults of low social origin are more pressed to find a job because they have less economic capital than young adults of high social origin. On the other hand, young adults of high social origin are more likely to find a job due to their parents’ social capital. As arguments seem to support both directions, we hypothesise the following:

H1: The social origin gap in employment should have remained stable following the Great Recession.

Economic recessions and earnings

The Great Recession changed the economic structure in several countries, as some sectors, such as manufacturing and, above all, construction workers, were disproportionately hit (Redbird and Grusky 2016). In many countries, employment in the construction sector contracted after the housing bubble burst. The increase in unemployed workers put pressure on wages, especially because skilled workers who lost their jobs may have applied for positions with lower wages, as happened during earlier economic crises (Devereux 2002). This form of crowding out mechanism is likely to lead to a greater asymmetry in bargaining power between workers and employers. Generally, during and immediately after recessions, companies take advantage of this situation by cutting wages and raising hiring standards in the face of the increasing influx of workers. This situation might be more common among young adults of low social origin than among young adults of high social origin, who may be financially supported by their parents until they find jobs that match their skills or their expected salary.

The economic recessions between 1976 and 1995 affected Canadian college graduates differently. Students with low predicted wages experienced more earnings losses than students with high predicted wages (Oreopoulous 2012). Students in the United States also faced this penalty when the economy was slack. Those who graduated during the economic crisis of the early 1980s experienced approximately 20 percent more wage losses than those who graduated during an expansionary economic cycle. Moreover, the earning losses persisted over almost the entire period observed between 1979 and 1989 (Kahn 2010). Empirical studies for the period between the 1980s and early 2000s show that economic recessions tend to aggravate the employment prospects and earnings of low educated labour market entrants in Japan and the United States (Genda et al. 2010). The findings suggest that those losses are stronger for the low educated and longer lasting in Japan than in the United States. Similarly, in Germany, young adults of low social origin experienced more earnings losses than young adults of high social origin when local unemployment rates were high between 1984 and 2011 (Zwysen 2016). An American study found that having a college degree protected young adults during the Great Recession in the United States. However, this premium was experienced only by individuals with an advantaged social background and not by their peers from disadvantaged backgrounds (Curry 2019). The overview of this literature leads us to formulate the following hypothesis:

H2: The social origin gap in earnings among young adults grew following the Great Recession.

The Great Recession and countries’ differences

Although all European countries were affected by the Great Recession of 2008/09, the extent of the crisis differed across them. Schematically, one can distinguish two groups of countries. The first consists of Austria, Germany, Nordic countries, and several Eastern European countries that have close economic links with Germany. These countries recovered quickly after the crisis and resumed their previous economic growth. The second group consists of Ireland, Southern European countries and some Eastern European countries that struggled in the aftermath of the crisis. Many of them started to recover economically only at the beginning of 2014, experiencing long periods of unemployment and an increase in the level of public debt (House et al. 2017).

Figure 1 presents data on the unemployment rate for France, Germany, Italy, Poland, Spain and the United Kingdom from 2002 to 2014 (OECD 2021). This figure shows that Spain, Italy and the United Kingdom had the highest increase in unemployment rates following the Great Recession. During the same period, France had a slight surge in unemployment rate. On the other side of the spectrum, Germany and Poland¹ experienced drops in unemployment rates over the same period. The unemployment rate in Poland before the Great Recession stands out from the rest of the countries, as its unemployment rate was approximately 20% in 2002 and decreased gradually to 9% in 2007. In that sense, we analyse Poland as a country that experienced economic expansion. In sum, Figure 1 indicates that compared with the other three countries, Italy, Spain and the United Kingdom were affected the most by the Great Recession.

Figure 1. Unemployment rate between 2002 and 2014. Source: OECD Labour Market Statistics, extracted 2021.

As discussed in the previous sections, deteriorating macroeconomic conditions are likely to influence the gap in earnings between young adults of different social origins. This leads us to our third hypothesis:

H3: The social origin gap in earnings changed in the countries most strongly affected by the Great Recession: Italy, Spain and the United Kingdom.

Data and methods

Data

Our analyses are based on two surveys. First, we use two rounds from the European Union Statistics on Income and Living Conditions (EU-SILC) in 2005 and 2011 (Eurostat 2020). As our purpose is to examine the association between social origin and young adults’ earnings and employment, these two cross-sectional surveys are selected because they are the only rounds that cover a period close to the 2008 financial crisis and that include information on parental background, in the case of the EU-SILC. Second, we use seven rounds of the biennial European Social Survey (ESS) between 2002 and 2014 (Norwegian Centre for Research Data 2020). We then select six EU countries with the largest population, which account for approximately 70% of the EU’s residents: France, Germany, Italy, Poland, Spain and the United Kingdom. In addition to having demographic importance, these countries are interesting to study because they possess different welfare regimes and experienced different trajectories before and after the 2008 financial crisis. We focus only on adults aged between 25 and 34, as young people were particularly vulnerable during the Great Recession (Bell and Blanchflower 2011; Rothstein 2021).

Measures

Our first outcome variable is self-reported employment. It is a binary variable: the non-employed versus those who work part-time or full-time jobs. The non-employed include all young adults who are not working regardless of whether they are actively looking for a job or not. The second key outcome is annual earnings and is only extracted from the EU-SILC, as the ESS does not contain a continuous annual earnings variable at the individual level but provides household income only in categories. Earnings are corrected for inflation² and only concern young adults with a paid job. We use the natural logarithm³ of the Euro. The EU-SILC reports earnings in the Euro currency for all countries, even Poland and the United Kingdom, which have their own national currencies.

Our key independent variable is social origin, and we measure it using parental education. The most detailed version of parental education that we could construct across both data sources (i.e. EU-SILC and ESS) distinguishes three hierarchical categories: (1) both parents with lower secondary degree or below, (2) at least one parent with upper secondary and post-secondary non-tertiary education and (3) at least one parent with tertiary education.⁴

We replicate our analysis using parental social class instead of parental education in the robustness section. The parental social class variable is harmonised across both data sets (i.e. EU-SILC and ESS) and is divided into three hierarchical categories. We use a merged version of the scheme developed by Oesch (2006), which distinguishes three employee classes: (1) upper-middle class: managers and professionals (ISCO 1 and 2); (2) lower-middle class: technicians, clerical, service, sales, skilled (i.e. agriculture, forestry and fish) and craft workers (ISCO 3–7); and (3) working class: plant and machine operators and elementary occupations (ISCO 8 and 9). These three categories closely mirror the scheme developed by Erikson and Goldthorpe (1992) between the service relationship for occupations at the top, intermediate occupations in the middle and the labour contract for working-class occupations at the bottom.

As we aim to measure the net effect of social origin on destination, we include respondents’ education level as a control with three categories: (1) lower secondary and below (ISCED 0, 1 and 2), (2) upper secondary (ISCED 3), and (3) post-secondary and tertiary (ISCED 4 and 5). We add a binary period variable that differentiates between the periods that preceded and followed the Great Recession. For the EU-SILC dataset, the 2005 survey is considered the period that preceded the Great Recession compared with the 2011 survey. For the ESS data set, if the respondents were interviewed between 2002 and 2007, they were coded as prior to the crisis. On the other hand, if the respondents were interviewed in any period between 2008 and 2014, they were aggregated together and coded as post-crisis. We control for respondents’ sex, age, and the interaction between social origin and education. We further control for the interaction between respondents’ education and the time variable. When the employment outcome is used, we add a binary variable that distinguishes between the data sources to account for differences across them. Table A1 in the appendix provides the descriptive statistics.

Analytical strategy

Our goal is to investigate whether the social origin gap widens between young adults of low and high social origin following the 2008 financial crisis. We thus analyse whether the Great Recession had heterogeneous effects on young adults of different social origins. We use the following regression model separately for six countries: (1) ${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{\_}\mathrm{O}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{i}{\mathrm{n}}_{\mathrm{i}\mathrm{t}}+{\text{β}}_2\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}{\mathrm{d}}_{\mathrm{t}}+{\text{β}}_3\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{\_}\mathrm{O}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{i}{\mathrm{n}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}{\mathrm{d}}_{\mathrm{t}}+{\text{β}}_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{ϵ}_{\mathrm{i}\mathrm{t}}$(1) where Y stands for the logarithm of earnings or the employment status of individual i at time t. We estimate the binary variable of employment status using a linear probability model, as it makes the interpretation of the coefficients and the comparison across models easier (Mood 2010). The coefficient β₃ is an interaction between social origin and period that estimates the differences in the gap of earnings – or employment – of young adults of high social origin compared with their counterparts of low social origin before and after the Great Recession. Controls_it includes a vector of control variables, such as sex, education and age. The error term ϵ includes everything unobserved by the model, such as measurement error or luck. As our analysis of the social origin gap in earnings is limited to two modules and the number of higher-level units in our sample does not exceed the minimum number of units (30) required for unbiased estimates of the individual- and country-level effects, we do not follow a multilevel approach (Bryan and Jenkins 2016). While our study must be regarded as unearthing potential associations rather than showing pure causal effects, the presence or absence of a (possibly changing) association between social origin and labour market outcomes appears of interest per se. For replicating all analyses, we provide our R-script online (Moawad 2022).

Descriptive results

Employment

We examine the descriptive evidence for a moderating impact of the Great Recession on the employment of young adults of different social origins in Figure 2. This figure uses data from the EU-SILC and the ESS, reports the change in employment in percentage points and displays the period that preceded and followed the Great Recession. All modules from both data sources prior to 2008 were pooled together and were compared with all modules that came after that period, which were also combined jointly. Over that period, we see that following the Great Recession in Germany, employment increases by one percentage point for young adults of low social origin compared to an increase of twelve percentage points for young adults of high social origin. We observe a different pattern in France, Poland and the United Kingdom, where young adults of high social origin were more disadvantaged in terms of employment than young adults of low social origin following the 2008 financial crisis. For example, in the United Kingdom, employment decreases by one percentage point for young adults of low social origin compared with a fall of six percentage points for young adults of high social origin. In Italy and Spain, we see more or less a similar change in employment between young adults of low and high social origin following the Great Recession.

Figure 2. Change in employment in percentage points before (2002-07) and after (2008-14) the Great Recession, by parental education. Source: EU-SILC and ESS, computations by the authors.

Note:The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005 and the ESS rounds in 2002, 2004 and 2006. Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011 and the ESS rounds in 2008, 2010, 2012 and 2014.

Earnings

Figure 3 uses data from the EU-SILC and reports the change in earnings in percent among young adults of different social origins between 2005 and 2011.⁵ It shows that the earnings remained stable for young adults of low social origin compared with an increase of 6% for young adults of high social origin in Spain. In other words, young adults of low social origin experience more disadvantages in terms of earnings than their counterparts of high social origin following the Great Recession in Spain. On the other hand, young adults of high social origin experience more disadvantages in terms of earnings than their counterparts of low social origin in France, Germany, Italy and Poland following the 2008 financial crisis. Among this group, we find the most notable difference in Poland, as earnings increase by 54% for young adults of low social origin compared with an increase of 29% for young adults of high social origin.

Figure 3. Change in inflation-adjusted earnings before (2005) and after (2011) the Great Recession, by parental education. Source: EU-SILC, computations by the authors.

Note: The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005. Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011.

Results from multivariate models

We move on to analyse our multivariate models (see Tables A2 and A3 in the appendix). Figure 4 shows the coefficient β₃ in equation (1)(1) ${\mathrm{Y}}_{\mathrm{i}\mathrm{t}}={\text{β}}_0+{\text{β}}_1\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{\_}\mathrm{O}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{i}{\mathrm{n}}_{\mathrm{i}\mathrm{t}}+{\text{β}}_2\mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}{\mathrm{d}}_{\mathrm{t}}+{\text{β}}_3\mathrm{S}\mathrm{o}\mathrm{c}\mathrm{\_}\mathrm{O}\mathrm{r}\mathrm{i}\mathrm{g}\mathrm{i}{\mathrm{n}}_{\mathrm{i}\mathrm{t}}\ast \mathrm{P}\mathrm{e}\mathrm{r}\mathrm{i}\mathrm{o}{\mathrm{d}}_{\mathrm{t}}+{\text{β}}_4\mathrm{C}\mathrm{o}\mathrm{n}\mathrm{t}\mathrm{r}\mathrm{o}\mathrm{l}{\mathrm{s}}_{\mathrm{i}\mathrm{t}}+{ϵ}_{\mathrm{i}\mathrm{t}}$(1) , that is, the interaction effect of parental education with the period variable. This interaction shows whether the direct effect of social origin changed following the Great Recession by comparing young adults of low social origin with their counterparts of high social origin. The left panel is estimated using linear probability models⁶, and the right panel is estimated using OLS. Negative values of the interaction indicate that, net of education, young adults of low social origin experienced more disadvantages in terms of employment (left panel) or earnings (right panel) compared with young adults of high social origin following the Great Recession. Figure 5 illustrates the predicted probabilities of employment (left panel) and earnings (right panel) of young adults of low and high social origin before and after the Great Recession (conditional on education). The left panel in Figures 4 and 5 uses the EU-SILC and ESS data, represents the employment outcome and compares the periods of 2002–07 and 2008-14. The right panel in Figures 4 and 5 represents the log earnings outcome, compares the period of 2005 and 2011 and uses only the EU-SILC data, as the ESS does not include a continuous annual earnings variable at the individual level.

Figure 4. Change in the direct effect of social origin on employment (left) and log earnings (right) comparing low-origin adults relative to high-origin adults aged 25–34.

Note: left panel displays the interaction between parental education and period on employment. Coefficients from linear probability models. 95 percent confidence intervals are plotted horizontally. Dependent variable: employment. Models control for age, sex, respondent education, parental education, period, parental education* respondent education, respondent education*period and data source. The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005 and the ESS rounds in 2002, 2004 and 2006 (coded=0). Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011 and the ESS rounds in 2008, 2010, 2012 and 2014 (coded=1). Full table can be found in the appendix in Table A2. Sample size: France: 6501; Germany: 6280; Italy: 11396; Poland= 10708; Spain= 9817; United Kingdom= 5370. Right panel displays the interaction between parental education and period on log earnings. Coefficients from linear regression models. 95 percent confidence intervals are plotted horizontally. Dependent variable: log earnings. Models control for age, sex, respondent education, parental education, period, parental education*respondent education and respondent education*period. The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005 (coded=0). Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011 (coded=1). Full table can be found in the appendix in Table A3. Sample size: France: 3803; Germany: 2652; Italy: 7522; Poland= 6379; Spain= 5522; United Kingdom= 2698.

Figure 5. Predictive probabilities of employment (left panel) and log earnings (right panel) by social origin and period.

Note: left panel displays the predictive probabilities from linear probability models. 95 percent confidence intervals are plotted vertically. Dependent variable: employment. Models control for age, sex, respondent education, parental education, period, parental education* respondent education, respondent education*period and data source. The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005 and the ESS rounds in 2002, 2004 and 2006 (coded=0). Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011 and the ESS rounds in 2008, 2010, 2012 and 2014 (coded=1). Sample size: France: 6501; Germany: 6280; Italy: 11396; Poland= 10708; Spain= 9817; United Kingdom= 5370. Right panel displays the predictive probabilities from linear regression models. 95 percent confidence intervals are plotted vertically. Dependent variable: log earnings. Models control for age, sex, respondent education, parental education, period, parental education*respondent education and respondent education*period. The dummy period variable prior to the Great Recession includes the EU-SILC survey year in 2005 (coded=0). Following the Great Recession the dummy period variable includes the EU-SILC survey year in 2011 (coded=1). Sample size: France: 3803; Germany: 2652; Italy: 7522; Poland= 6379; Spain= 5522; United Kingdom= 2698.

Employment

The left panel in Figure 4 shows that there is a negative coefficient for Germany, suggesting that the gap in the direct effect of social origin in terms of employment changed by 9 percentage points following the Great Recession. However, this estimate is only significant at the 0.1 level (p = 0.0740). On the other hand, we see a different trend in France, Spain, and the United Kingdom, whereby the gap in the direct effect of social origin changed by 4 percentage points in terms of employment following the Great Recession. However, the estimates do not reach statistical significance at the 0.05 level in all three countries.

The model in Figure 5 predicts a value of 0.76 for employment for young adults of low social origin (grey line) and 0.7 for young adults of high social origin (black line) in Spain before the 2008 financial crisis (conditional on education). Both predicted estimates decrease following the Great Recession, as the model predicts a value of 0.71 for employment for young people of low social origin compared to 0.61 for their counterparts of high social origin (conditional on education). These predicted values translate to a change of 4 percentage points ((0.7–0.76) – (0.61–0.71) = 0.04) in the gap of the direct effect of social origin in terms of employment in Spain following the Great Recession, as shown in Figure 4. Figure 5 suggests that net of education, the likelihood of employment did fall more for young adults of high social origin than for those of low social origin in Spain following the 2008 financial crisis. Figure 5 displays a similar pattern in the United Kingdom. On the other hand, while Figure 4 also shows a positive size effect of 4 percentage points in France, Figure 5 suggests that it is driven by a different pattern. That is, net of education, the likelihood of employment increased for young adults following the Great Recession, but this increase was higher for young adults of low social origin than for their counterparts of high social origin.

Earnings

Next, the right panel in Figure 4 shows that the interaction between parental education and period on log earnings yields a large negative effect size in Spain following the Great Recession. Point estimates show that the gap in the direct effect of social origin negatively changed by 0.43 in terms of log earnings. However, this estimate is only significant at the 0.1 level (p = 0.0768). The model in Figure 5 predicts a value of 8.39 for log earnings for young adults of low social origin and 8.178 for young adults of high social origin in Spain prior to the Great Recession (conditional on education). Both predicted values increase in 2011, as the model predicts a value of 8.591 for log earnings for young adults of low social origin and 8.805 for their counterparts of high social origin following the 2008 financial crisis (conditional on education). These predicted values translate to a change of 0.43 ((8.178-8.39) – (8.805-8.591) = – 0.43) in the gap of the direct effect of social origin in terms of log earnings in Spain following the Great Recession, as shown in Figure 4. We plot the predicted probabilities for earnings⁷, rather log earnings, in Figure A2 and find that this change in the gap of the direct effect of social origin is sizable and corresponds to – 737 euros ((11977.5-11075.3) – (12544.6-10905) = – 737).

Table A3 suggests that not only young adults of low social origin faced more disadvantages in terms of log earnings in Spain following the Great Recession, but this also applies to young adults of middle social origin when compared with their counterparts of high social origin. Point estimates show that the gap in the direct effect of social origin negatively changed by 0.67 in terms of log earnings in Spain following the 2008 financial crisis (see Table A3). This estimate is statistically significant at the 0.05 level.

On the other hand, point estimates in Figure 4 suggest that the gap in the direct effect of social origin in terms of log earnings positively changed by 0.42 in Poland following the 2008 financial crisis. However, this coefficient is not statistically significant at the 0.05 level. Figure 5 indicates that while log earnings increased for both young adults in Poland following the Great Recession, this increase was higher among young adults of low social origin compared with their counterparts of high social origin. In the rest of the countries, the social origin gap in terms of log earnings remained stable following the Great Recession.

Robustness checks

A concern with our analyses is that including individuals with marginal part-time jobs may bias our findings. Thus, we discard observations with marginal part-time work, that is, individuals working less than 10 h per week. Figure A5 shows that there are two notable differences between our original findings and the robustness checks that are present in Spain and Germany. Our original findings indicate that net of education, young adults of low social origin experienced more disadvantages in terms of employment than their counterparts of high social origin in Germany following the Great Recession. On the other hand, young adults of high social origin experienced more disadvantages in terms of employment than their peers of low social origin in Spain. Our findings in Figure A5 reveal a null finding for both countries over the same period. Our robustness checks in terms of log earnings are robust, as we find similar results to our original analysis in all countries (see Figure A5 in the appendix).

Likewise, a sceptical reader might consider that social origin is better measured with parental social class. Therefore, we rerun our regression analyses on employment and earnings, as shown in Figure A6 in the appendix. The robustness checks on employment show different findings in France, Spain and the United Kingdom. While our original findings show that, net of education, young adults of high social origin faced more disadvantages in terms of employment than their counterparts of low social origin in these three countries, this is not the case if we use parental social class as a measure of social origin. This implies that this robustness check is more in line with our first hypothesis – that the social origin gap in employment should remain constant following the Great Recession – than our original results. Furthermore, our robustness checks in terms of log earnings do not hold in Spain or Poland in case parental social class is used as a measure of social origin instead of parental education.

Discussion

Amidst considerable concerns about a heterogeneous effect scenario whereby the Great Recession disproportionally affects disadvantaged groups (Redbird and Grusky 2016), our article examined the extent to which such concerns are warranted. To do so, we examined whether the direct effect of social origin in terms of employment and earnings changed among young adults of low and high social origin after the Great Recession in the six largest countries of the European Union. A first group consisting of Italy, Spain and the United Kingdom was strongly hit by the recession, in contrast with a second group including France, Germany and Poland, which was less touched by the recession.

Three main findings emerge from our empirical analysis. First, against our expectations, we find that, net of education, the chances of employment of young adults of high social origin were more negatively affected compared with their counterparts of low social origin following the Great Recession in France, Spain and the United Kingdom. This contradicts H1, which proposed that the social origin gap in employment would remain stable following the Great Recession. Our findings suggest that employment fell more for young adults of high social origin than for their counterparts of low social origin in Spain and the United Kingdom. This could happen if young adults of high social origin do not find a job that matches their skills or expected salary during an economic recession and thus extend their search period of employment. This is less likely to occur among young adults of low social origin who might have to step in and support their parents or partner in case of a financial burden or a job loss.

Second, with respect to earnings, our results indicate that young adults of low social origin seemed to be more harmed than their counterparts of high social origin following the Great Recession in Spain. Our findings point to a higher increase in earnings for young adults of high social origin compared with their counterparts of low social origin. This might be driven by the social capital of the parents. When the economy is slack, parents of high social origin are more likely to influence their children’s chances of finding employment with better pay than their counterparts of low social origin. In general, this finding provides only partial support for H3, which proposes that the social origin gap in earnings changed in the countries most strongly affected by the Great Recession. One possible explanation for the null findings in Italy and the United Kingdom may be linked to the intensity of the financial crisis. While the unemployment rate increased by 12 percentage points in Spain between 2005 and 2011, it rose by only 4 and 3 percentage points in Italy and the United Kingdom, respectively. It could be that the gap in the direct effect of social origin changes only in countries that experience a severe financial crisis.

Third, we found that in Poland, the gap in the direct effect of social origin in terms of employment remained stable following the Great Recession. On the other hand, the analysis showed that the gap in the direct effect of social origin in terms of earnings contracted following the 2008 financial crisis. These findings suggest that an economic boom is particularly helpful for the earnings of young adults of low social origin. Hence, macroeconomic policy that stimulates growth and employment seems especially effective for the earnings of disadvantaged youth. Our findings in this respect are in line with previous research showing that the disadvantaged benefit the most from economic expansion (Hines et al. 2001).

Our findings were not consistently robust and should be taken with caution. This is the case if parental social class is used instead of parental education. Parental education and parental occupation have been used interchangeably in intergenerational research. After all, parental education precedes parental occupation and highly influences it. While both indicators can be used to measure social origin, scholars also argue that they can operate – to some extent – through different mechanisms (Jaeger 2007; Bukodi and Goldthorpe 2013; Erola et al. 2016). Parental education is more likely to influence children’s outcome through skills, traits and cultural capital. On the other hand, parental social class is more likely to affect children’s outcome through status and prestige (Erola et al. 2016). However, it is important to note that these different mechanisms of parental education and parental social class are not completely independent from each other and can largely overlap. Our findings suggest that parental education seems to matter more than parental social class with respect to the change in the direct effect of social origin between young adults of low and high social origin. This is in line with previous research which shows that parental education is more important than parental social class for young adults’ socioeconomic status (Erola et al. 2016). Alternatively, the differences in our findings between parental education and parental social class could be driven by the lack of detailed data on parental occupation. We could only construct a parental social class using one digit of precision.

Our study presents several limitations. First, earnings are self-reported by the respondents, which likely leads to some measurement errors. Second, our findings with respect to earnings capture only two time points, which could be outliers of a general trend. To check whether this is the case, we plot the overall trends of earnings inequality in our countries in Figure A7 and find that the two time points in 2005 and 2011 are not outliers (World Inequality Database 2021). Finally, we capture only whether the direct effect of social origin changed in the short term following the Great Recession. It would also be interesting to measure the long-term repercussions, as the 2008 economic crisis may have led to structural changes in the labour market. Detailed data on single countries might be used to address this possibility, but data that cover multiple countries face more constraints.

New concerns regarding social mobility in OECD countries have recently been raised (OECD 2018). Our analysis of the aftermath of the Great Recession may also be relevant to the recent economic crisis unfolding after the COVID pandemic. As our findings suggest that only severe economic recessions could disproportionally affect the earnings of disadvantaged groups, there may be no strong reason for concern, as the economic recession following the COVID crisis is less severe than the Great Recession thus far.

Acknowledgment

For helpful discussions or relevant feedback to this article, I thank L. Iván Canzio, Alessandro Di Nallo, Michael Grätz, Patrick McDonald and the three anonymous reviewers at European Societies. Special thanks to Daniel Oesch for extensive feedback.

Disclosure statement

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

Additional information

Notes on contributors

Jad Moawad

Jad Moawad is a PhD candidate at the University of Lausanne. His core research interest lies in the field of social stratification and mobility.

Notes

1 Even though Poland experienced a currency devaluation at the onset of the Great Recession, the literature suggests that Poland was one of the European countries that was least affected by the 2008 financial crisis (Polanski 2014; Allington and McCombie 2016; Giugni and Grasso 2018).

2 The EU-SILC data are harmonised and suitable for comparison; however, for Italy and Spain, there is information only on net earnings in both modules, while Germany and the United Kingdom are missing information on net earnings in 2011. The rest of the countries include information on gross and net earnings. The Spearman correlation between the two indicators reveals that they are closely linked, with a correlation of 99%. Therefore, we use net earnings in Italy and Spain and gross earnings in the rest of the countries.

3 We recode zero earnings to one before taking the natural logarithm.

4 In case of missing information on one of the parent’s education, the educational attainment of the other parent was selected.

5 Earnings are measured in Euro in the EU-SILC. The large decrease in earnings of young adults in the United Kingdom is strongly influenced by the devaluation of the British Pound relative to the Euro between 2005 and 2011.

6 We plot the predicted probabilities of the logistic regression models in Figure A1. The results suggest that the use of linear probability models or logistic regression models leads to identical substantive findings.

7 To check whether our findings change substantively in case earnings are used instead of log earnings, we compare the change in the gap of the direct effect of social origin for both outcomes in Figure A3. The figure suggests that there is no substantive difference between both outcomes in all countries except for Poland. This divergence in Poland seems to be influenced by the highly negative skewness in the earnings variable. This is clear in Figure A4 where we plot the density of earnings and log earnings. Figure A4 also shows that the tail of the distribution of log earnings is less skewed than the one of earnings in all countries. This is not surprising as log earnings are used to avoid issues with non-normality and growing error variance (see Ermini and Hendry 2008).