Persisting class inequality?: Comparing class-specific unemployment in early working life in Britain and West Germany Open Access

A substantial body of research has found evidence for the salience of social class in structuring the opportunities and constraints of individual life-courses (e.g., Marshall et al.1988; Erikson & Goldthorpe 1993). More recent research in this tradition has considered the effect of globalisation and increasing economic uncertainty on differences between social classes. The overall conclusion of this work is that significant inequality in life chances persists between social classes (Breen 1997; Goldthorpe 2000, 2003; Mayer 2001). Economic uncertainty does not affect all classes equally: while individuals in some classes are protected from it, individuals in other classes are in a very insecure position. Class inequality persists, and may even increase over time. One concern of this paper is to consider whether or not this is the case, by empirically examining relative class inequality over time.

But are these historical processes similar across countries? A vein of cross-national research on social inequality stresses the embeddedness of social stratification and labour market processes in different national contexts (Mayer 1997; Esping-Andersen 1999; Soskice 1999). A second concern of this paper is whether class inequality varies across countries. To this end, we compare class inequality in two capitalist countries, Britain and Germany, which differ significantly in their national contexts. By this we mean primarily labour market structure, labour market institutions, the education system and the welfare system.

In this paper we select one particular indicator of class inequality, unemployment.¹ Why unemployment? As insecurity is usually taken to be a critical dimension in differentiating the employment relationship which determines classes, we argue that unemployment is a particularly pertinent indicator of the effect on social classes of recent economic developments.² For our empirical analysis, we also select a more homogeneous group of individuals for our analysis than the whole workforce, namely men in the early life course. This allows us avoid some of the problems associated with defining and recalling women's unemployment, and also minimize variations in the role of unemployment in different phases of the life course.³ However, it does mean we need to recognise the particular implications of choosing this group for our conclusions about class theory.

Specifically, the focus of the paper is the distribution of unemployment among classes from a comparative and longitudinal perspective. Our key research questions are: how does the risk of unemployment vary by social class in different cohorts and across countries, and what are the implications of this for class theory? Note that our purpose is not to try to explain the determinants of unemployment, but rather to consider the relative risk of unemployment across classes. In using the example of unemployment, this paper contributes to the literature on social class in a number of ways. Firstly, by considering the risk of unemployment among social classes over time, we consider some empirical evidence on the theoretical debate on persisting inequality. Secondly, by comparing the unemployment risk among social classes in Britain and Germany, we address the broader theoretical issue of whether class inequalities can be viewed as primarily structured by universal processes in market economies, or depend crucially on institutional structures that vary in important ways between countries.

In the next section we discuss recent work from the persisting class inequality perspective and critiques thereof, and how the arguments put forward relate to unemployment risks. In Section 3 we compare in detail both structural differences and institutional differences between Britain and Germany, and how we expect class inequalities in unemployment to differ across countries. In Section 4 we describe the data and methods used to consider some of the predictions developed earlier in the paper, in Section 5 we briefly review previous evidence and discuss the results of our own analysis, which uses a sample of young men. In conclusion we reflect on the implications of our findings for theories of inequality and the theory and measurement of class.

While there are other measures of social class,⁴ the class schema proposed by Erikson & Goldthorpe (1993) is particularly appealing because it is designed to be used in empirical research, and indeed specifically empirical research concerned, as we are here, with inequality of class outcomes. Here we discuss in more detail the basic tenets of this class theory, how classes are defined and what this class theory says about inequality between classes. In particular, we discuss the theory's predictions on the differences between the risk of unemployment among social classes, and how these are expected to change over time.

From the class perspective espoused by Erikson & Goldthorpe (1993), the basic distinction is between those who buy and sell labour (employers and employees). Among employees, it is crucial to think about type of employment contract (Breen 1997; Goldthorpe 2000). Labour contracts are easy to monitor and involve a specific exchange of work for effort, similar to a ‘spot market contract’. Service contracts, by contrast, are more difficult to monitor, and require much more autonomy. The employer makes a longer-term commitment to the employee, making the service contract much more secure than the labour contract. Workers with a service contract are shielded from market uncertainties, whereas those regulated by a labour contract are more vulnerable. There are also a group of employees who have ‘mixed’ employment contracts, combining aspects of both service and labour contract (see Goldthorpe 2000: 221–3, for further details). These are in an intermediate position regarding labour market risks. Note that while the detailed classification of occupations distinguishes between skill levels, skill itself is not an intrinsic part of the theory.

The key point of this research for us is that unemployment does not affect the different classes equally. Using the Erikson and Goldthorpe classification scheme, those on a labour contract, both manual workers (VIIa, VIIb, VI) and unskilled non-manual (IIIb) will be most exposed to the risk of unemployment.⁵ The routine non-manual workers (IIIa) and the supervisors (V) would come in an intermediate position. Those protected from unemployment are those with a service contract, I and II.

Considering change over time, Breen (1997) argues that recent rising insecurity will entail a shifting of labour market risk from employers to employees, but within employees different groups will be differentially affected by this risk. Those on a labour contract will be most affected by increasing economic risks, though the intermediate group (IIIa and V) will also suffer increasing economic risk due to technological changes and increases in monitoring by employers (Breen 1997: 481).

It should be noted that this line of research has not gone uncontested. From a quite different theoretical standpoint commentators have argued that due to changes like globalisation, societies are becoming increasingly individualised, with risks like unemployment and poverty affecting a large proportion of the population (Beck 1986; Giddens 1994). Individualisation of social inequality means that the risk of unemployment is decoupled from social class over time (Giddens 1994). Indeed the concept of class has been criticized from a diverse range of theoretical and political positions. Critics have argued that class has lost its relevance for modern societies, been replaced by new economic and political cleavages and that it is analytically bankrupt (see Lee & Turner 1996, for a summary of the debate on the concept of class and its continuing relevance).

And indeed while recent work arguing for the persistence and salience of class inequality generates interesting predictions about risks attached to different classes over time, the authors do not reflect on how these risks should vary in different institutional settings, i.e., different countries. The implicit assumption is that the employment relationship of social classes is universal. This paper takes issue with this assumption. Indeed from our comparison of unemployment risks in Britain and Germany, we draw some important conclusions for class theory. One important conclusion is that skill differences are important enough to warrant inclusion in the theory of class itself, if one aim of class analysis is to investigate the importance of class in shaping life chances. A second is that class inequalities in the risk of unemployment vary across countries, suggesting that institutional differences affect the employment security of different social classes. It is the task of this paper both to show this empirically, and to give a cogent theoretical account of why this is the case.

Why should we expect the risks attached to social classes to differ in Britain and Germany? For the purposes of conceptual clarity, we adopt a somewhat simplified perspective on the process by which unemployment risks among different social classes evolve. We treat global shocks such as technological change as ‘exogenous’ to our model. These have an impact firstly on the demand for labour in the economy. This impact may differ across countries, and hence we compare the structure of employment in Britain and Germany. We then consider how institutions influence the risk of unemployment of individuals in social class groups. The institutions we consider important for the risk of unemployment are the education and training system; the system of employment regulation and industrial relations, and the welfare system. Borrowing the term from Mills & Blossfeld (2001), we see institutions as ‘filters’, which affect how risks from globalisation are transmitted to different classes. These institutions may not influence class risks in a consistent way, but we argue that to understand class risks, it is important to consider multiple institutions and the interaction between them, like DiPrete et al. (1997). In these key institutions, Britain and Germany are shown to differ in crucial ways, which makes the cross-national comparison particularly fruitful. What follows is a stylized comparison of British and German institutions and how they have changed during the period covered by the data. Note that we limit the focus in this paper to West Germany. Our purpose in this paper is to consider long-term trends in class inequality in two capitalist societies with different institutions. East Germany was under communist rule for most of the period and classes and class inequality were rather different in nature there (Solga, 1995); East Germany is thus excluded.⁶

An important part of the story for unemployment risks is to consider some aspects of the demand for labour – the demand for various jobs in each country, and how these have changed over time. Both Britain and Germany have undergone a significant restructuring of their economies in the past 40 years. Looking at the period 1960–2000, we see a shifting balance in total employment, away from industry and agriculture towards services in both Britain and Germany. The shift from agriculture in Germany is more marked in the period, falling from 14 per cent of total employment in 1960 to 2.5 per cent in 2000 (OECD 1984, 2001). The decline in industry, particularly manufacturing, is more pronounced in Britain, falling from nearly half total employment in 1960 to 25 per cent in 2000. Conversely, in Britain we see a more sustained growth in services, from under 50 per cent of employment in 1960 to 73 per cent of employment in 2000 (OECD 1984, 2001). In West Germany this figure was just under 67 per cent in 2000, with an especially low proportion in unskilled service jobs.

How do we expect these differences in the distribution of jobs to affect the risk of unemployment in the two countries? Our initial hypothesis is a rather simple one: as the demand for manual jobs has fallen in both countries, manual workers will be much more vulnerable to unemployment than the service class. However, because we do not see such a decline in industry in Germany, and there is some evidence that structural change in the German labour market affects entry cohorts, with the jobs of existing manual workers being protected (Blossfeld 1989), we expect that manual workers will not be as much at risk of unemployment there. In summary, our hypothesis is that while manual workers will, in general, be more at risk of unemployment, this will be less the case in Germany than in Britain.

Industrial relations and labour market regulation systems can determine how easy it is for employers to dismiss employees, and who will find it difficult to get a job initially. Britain and Germany are almost always placed in different categories – whether in comparisons of industrial relations (Crouch & Streeck 1997), ‘production regimes’ (Soskice 1999) or dismissal protection (Büchtemann & Walwei 1996).

Regarding industrial relations, wages in Germany are set by collective bargaining agreements between specific industrial unions and regionally based employers’ associations. In 1995, 84 per cent of German workers were covered by collective bargaining arrangements (Bispinck 1997). British industrial relations falls into two distinct periods – 1960s and 1970s and then 1980s and 1990s. The conservative government elected in 1979 embarked on a programme of labour market reform effectively dismantling centralised wage bargaining, and reducing employment protection. Wage bargaining in Britain is currently company based and uncoordinated (Deakin & Reed 2000). On dismissal protection, German legislation and practice is more restrictive than in Britain, as seen in a wide range of measures from legislation through to employer attitude surveys (Büchtemann & Walwei 1996).⁷ It is thus much easier for employers to dismiss employees in Britain than in Germany.

How do these institutional differences affect social classes and associated labour market risks? Recent class theory proposes that employment relationships (service and labour contracts) evolve because of problems of supervision, and the specificity of skills involved in the job (Goldthorpe 2000: 213). As such this theory is an employer-centred theory, as the employer determines the conditions of the contract and this leaves no role for either (a) institutional constraints on employers (i.e., high dismissal costs for many workers) or (b) powerful trade unions. While much of the analysis of social mobility is highly sensitive to institutional features, this is not true of the theory of class. As Esping-Andersen (1993) notes ‘… orthodox class theory is nested in an institutionally ‘naked’ world, an Adam Smithian world of unfettered markets’ (Esping-Andersen 1993: 8).

We argue that employment relationships may be the product of more diverse factors like, for example, trade union bargaining and employment protection. In Germany the role of works councils is highly relevant, and institutional constraints faced by employers differ rather substantially between the two countries.⁸ The key point: the employment relationship in Germany in most cases is not negotiated as an individual contract between employer and employee, but collectively agreed. Germany thus diverges more from the orthodox class theory than Britain. If this is the case, it may imply a reworking of the basic tenets of class theory to account for institutional differences in employment security. This is a point we return to in our conclusion.

Though we should stress that while Britain never had the employee protection or the extent of collective wage agreements as Germany, we see diverging trends in the two countries, with the British system becoming more decentralised, and various legislative changes giving more power to the employer. In terms of unemployment risk, our overall hypothesis is that the differences between classes (i.e. particularly between service class and skilled manual workers) will not be as great in Germany as in Britain, and we expect the cross-national differences in class risks to become more pronounced in recent decades.

Our third main hypothesis concerns cross-national differences in the system of education and training in the two countries. In modern labour markets, the education system is an important mediator between supply and demand in the labour market. Education systems may differ substantially as to how they match their outputs with labour market demand, and this is one reason why a country's education system may play an important role in determining unemployment risks.

In particular, the extent and nature of vocational training is thought to be salient in determining the matching process in the labour market. While Britain lacks a standardised and widespread system of vocational training, there is much stronger vocational orientation in Germany, and segmentation along occupational lines (Müller et al.1998; Hillmert 2001). Vocational training may be viewed as a way of improving matching between individuals and employers by providing individuals with specific skills which they can use on the job, and by sending a very clear signal to employers about the potential productivity of a given jobseeker. Indeed many German apprentices are retained in the firm in which they did their apprenticeship, thus avoiding a potential period of unemployment completely.⁹ As many authors have noted, in this way the German system facilities a smooth transition from school to work and stable occupational careers (Hillmert 2002; Müller & Gangl, 2003).

So, those with an apprenticeship (or alternatively higher educational qualifications) are protected from unemployment. However, a very small number receive no vocational training, and they are very vulnerable to unemployment (Shavit & Müller 2000). Only those with an apprenticeship will occupy a skilled position, and many unskilled do not have an apprenticeship. Thus the key distinction in Germany is not between manual and non-manual workers, but between skilled and unskilled workers. In Britain, with its much more general education system, there is a much larger group with no formal qualifications. Unskilled workers are not such a marginal group in the labour market, or as vulnerable to unemployment. Formal skills are less salient in the British labour market: so too is the ‘skill divide’. Comparing these two countries, skill emerges as an important issue for class theory, which does not explicitly include skill differences. This is another point we return to in conclusion. Returning to a proposed hypothesis from this discussion, our hypothesis is that unskilled workers in Germany-both manual and non-manual – will have a particularly high risk of unemployment, relative to other classes; this is much less the case in Britain. In Britain while unskilled workers have a high risk of unemployment, so too do other classes (e.g., skilled workers and the intermediate class).

In the following discussion we focus on welfare provision for the unemployed, as this is the most relevant aspect of welfare for considering unemployment risks. We are primarily interested in aspects of welfare which may influence the transition to unemployment, as the focus in this paper is the risk of becoming unemployed, not the duration of unemployment.

In typologies of welfare for the unemployed, as with employment protection and industrial relations, Britain and Germany almost always fall into different categories (e.g., Esping-Andersen 1999). These typologies typically contrast the predominantly status-based insurance system (Germany) versus predominantly needs-based, means-tested system (Britain). Insurance benefits are paid to the individual, provided they satisfy the contribution conditions. In Germany most benefits are based on previous earnings, while in Britain benefit rates are not linked to previous earnings, and tend to be low. On average, though this is not true for all individuals, most authors would argue that benefits are somewhat more generous in Germany.

An influential hypothesis in both political and academic circles is that high unemployment benefits cause higher unemployment: individuals entitled to generous benefits are more likely to become (voluntarily) unemployed. The fact that German benefits aim to preserve income and British benefits are flat-rate means that we might expect a greater disincentive for higher earners (service class I and II) to become unemployed in Britain than in Germany. However, this hypothesis rests on the assumption that benefits are granted unconditionally to unemployed workers. Given that there are benefit penalties (i.e., withdrawal of benefit) for voluntary unemployment in both countries, this hypothesis seems problematic (McGinnity 2004). In addition, as there is little unambiguous previous evidence showing the effect of benefit levels on transitions into unemployment, it is not clear that this effect would be present.

The funding of unemployment benefit also differs – in Germany the principal benefit, unemployment insurance, is funded from ear-marked contributions, separately administered by the Federal Labour Office. In Britain all benefits are funded from general taxation, and the state has full control over unemployment compensation (McGinnity 2004). Some authors (e.g., Esping-Andersen 1993, 1999) argue that the higher fixed costs of labour in Germany, the result of the funding of welfare, inhibits low wage unskilled jobs, leading to more unemployment among unskilled. This supports the hypothesis based on the education system discussed above, i.e., that unemployment in Germany will be particularly high among those without an apprenticeship, namely the unskilled. Once again then, our hypothesis is that unskilled workers in Germany will be particularly vulnerable to unemployment.

As we are interested in long-term trends over time in the relative risk of unemployment among individuals, retrospective work history data provides the most suitable empirical basis for this analysis. The data on Germany are taken from the German Life History Study (GLHS) (Mayer & Brückner 1989; Brückner & Mayer 1995; Hillmert et al.2003) in which retrospective data was collected for a variety of life domains.¹⁰ In our analysis, only West German cohorts were included, containing only German citizens. The data on Britain stems from the British Household Panel Study (Taylor et al.1996).¹¹ This dataset is designed to be representative of England, Wales and the largest part of Scotland, and the analyses make use of joint work-history files which combine retrospectively collected life-history data on employment status with panel data (see Halpin 1997). As the birth cohorts in the GLHS were selected in clusters, comparable groups were also selected in the British data, as presented in Appendix Table A2 (see also Hillmert 2001). We also select the British data to cover the same period of the work histories as in the German data, i.e., the early labour market career. While this is a somewhat short observation window, we argue that our analysis considers a highly significant part of the labour market careers of these individuals.

As we use retrospective work history data, unemployment in both surveys is self-defined (i.e., as reported by the individual). This is not unproblematic. Due to memory effects (Dex 1995; Reimer 2001), it is likely that unemployment is underestimated in the retrospective reports, especially if there has been a long interval between the event and the interview. Short-term spells of unemployment are also more likely to be forgotten than long-term spells. However, we do not expect that this under-reporting will substantially affect our estimation of relative risks between classes, or the comparison between countries.

Social class was coded according to the Erikson–Goldthorpe–Portocarero (EGP) schema (Erikson & Goldthorpe 1993). The categories are presented in detail in Appendix Table A1. The class position of unemployed people is defined by the social class of the last job. To focus on a more specific part of the employment system, we do not consider self-employed and agricultural workers. For reasons of sample size, we aggregate the other classes into four groups, which are theoretically relevant for our investigation, namely: (1) upper service class, EGP I & II; (2) routine service class and supervisors, EGP IIIa and V; (3) skilled manual workers EGP VI; and (4) unskilled manual and non-manual workers, EGP VIIa and IIIb. These are very similar to the groups discussed by Breen (1997) and Goldthorpe (2000) with the important exception that we distinguish skilled and unskilled manual workers.

The class distributions for young men aged 27 are presented in Table A2. What are the overall trends? Consistent with sectoral changes described in Section 3.1, in Appendix Table A2 we see a general decline in the working class. These data are generally consistent with class distributions and trends reported in previous research (e.g., Marshall et al.1988; Müller et al.1998).

Turning to individual class sizes of interest to us, we find generally more in classes I and II in Britain. The proportion for these 27-year-olds is about the same as the entire workforce for Britain, though the proportion in classes I and II is lower for Germany, especially in the 1964 and 1971 cohorts (Marshall et al.1988; Müller 1998). We expect this may be linked to the education system in Germany, where those in higher education may not be finished by 27, whereas in Britain they are. There are also interesting differences in the manual classes between the two countries. Overall the decline of the manual class has been much less marked in Germany, consistent with the discussion of sectoral shifts described in Section 3.1. But in particular there are more skilled manual workers in Germany (class VI), and in recent years this difference has become more marked.

The class-specific risk of unemployment is measured by the following proportion:

There has been little previous research looking at the relative risk of unemployment among different social classes, and nothing comparing Britain and Germany directly. While the results do indicate class inequalities in the risk of unemployment, the evidence is not systematic, or systematically comparable. The work focuses on different time periods, and different groups at risk.

For Britain, Layte et al. (2000), looking at transitions from employment to unemployment in the period 1975–1992, find that compared to the upper service class (I & II), all classes are significantly more likely to become unemployed. The manual classes are most likely to move to unemployment, particularly the unskilled manual class. For Germany, recent evidence on transitions from the first job for the period 1984–1998, finds the higher service class (I) significantly less likely to become unemployed than the skilled manual workers and supervisors (V/VI) (Kurz & Steinhage 2001). Unskilled workers are significantly more likely to move to unemployment than skilled manual workers and supervisors. Other class differences in the transition to unemployment are not statistically significant, though note that the sample is more limited than in the British study.

Regarding changes over time, work by Gallie et al. (1998, Chapter 5) suggests that the relative risk of unemployment between the manual classes and the service class in Britain has remained stable over time, at least in the period 1979–1992. There is no similar evidence for Germany. We know of no other study so far comparing unemployment risks across classes and across cohorts in Britain and Germany.

What do we learn about relative unemployment risks using the data and methodology described in Section 4? To reduce the complexity of these various dimensions, we first ignore the focus on life-course effects and compare unemployment risks across countries, cohorts, and classes. Figure 1 presents the class-specific risk of unemployment for each cohort at a given age.

From Figure 1 we can see that in general there is a higher risk of unemployment for later cohorts, and no class is immune to unemployment. Even with an underestimation of unemployment using life course data, we still find a high incidence of unemployment, particularly in the youngest cohorts. For virtually all cohorts, the level of unemployment has been higher in Britain than in Germany.

While there have been different patterns of class-related unemployment in the two countries, they are not always clear-cut. In general the risk of unemployment is lowest for the upper service class (I & II) and highest for the unskilled workers (VIIa/IIIb), though there are exceptions to this.

One of the problems with examining each of the cohorts in detail, as in Figure 1 is that there may be considerable fluctuation due to both sampling error and short-term economic influences, which makes the overall trend difficult to disentangle. Moreover, for simplicity we have presented the unemployment risks at one particular age: the picture may be somewhat different when looking at different ages.

To minimise this fluctuation and give more systematic evidence for the ‘bigger picture’ which interests us, we combine these dimensions by estimating an overall model for the risk of unemployment, depending on country, cohort, age, and social class, using all the available data. The dependent variable is the dichotomous variable ‘experiencing unemployment within the risk period’ versus ‘not experiencing unemployment’ (operationalised as above). For each individual we have calculated the risk for three different age periods, so each individual can contribute up to three units of analysis, depending on how long we observe them for.

The results are presented in Table 1. In model 1 the focus is on the cross-national comparison, including age and class effects and how these vary between the two countries. In model 2 we also introduce terms for how the class effect varies across countries over time. The reference country in both models is Britain, thus the ‘main effects’ (except the coefficients for the cohorts) refer to Britain: the effects for Germany are represented by the main effects plus interaction terms.

For the discussion of cross-national differences in unemployment risks we consider the results from model 1. For the moment we are not interested in structural trends over time, so this model is more appropriate as it allows a general comparison of the countries for the entire period. The model shows a higher risk of unemployment for successive cohorts, and somewhat lower unemployment in Germany, both of which are consistent with the overall unemployment rate and Figure 1.

In support of the class stratification approach we find that in Britain class is a good predictor of the risk of unemployment. The service class is much less vulnerable to unemployment than either the intermediate class (IIIa/V), the skilled manual class (VI) or the unskilled classes (IIIb/VIIa). This is shown by the class main effects.

That said, we also see, by the negative and significant interaction terms for Germany that, for the most part, there is less of a difference in class risks in Germany than in Britain. This is most clearly the case for the intermediate class and the skilled manual class. These findings are consistent with our hypothesis that there will be less class differences in unemployment risks in Germany because the system of employment protection acts to protect all workers. In spite of this cross-national difference, if we sum the main effects and interaction terms, we still find class to be significantly associated with unemployment in Germany.

However, the position of German unskilled workers is somewhat different to that of other German classes. For German unskilled workers the interaction term is negative but not significant, suggesting that they are in fact in a somewhat similar position vis a vis the service class as in Britain. While in general class differences in unemployment are less in Germany, this is not the case for the unskilled. This finding fits with our discussion of the education systems, where we stress that those who have no qualifications in Germany, the unskilled, fare very badly indeed. We also suggested that the contribution-based welfare system may lead to lower job opportunities for those with no skills. The effect of an education system which particularly marginalizes the unskilled counteracts our overall hypothesis of less class inequality in Germany.

The most pronounced difference in class risks between Britain and Germany concerns the risk of unemployment among skilled manual workers. Skilled workers are a protected group in Germany, compared to Britain. We suggest that for skilled manual workers, not only differences in the employment protections system play a role, but also the less dramatic decline in industry in Germany. All of these factors led us to expect a lower risk of unemployment among manual workers in Germany, and this is indeed what we find in model 1.

Regarding cross-national differences in the distribution of risk in the early labour market career, from model 1 we find a higher risk of unemployment for the 25–29 age group in Britain, relative to the youngest age group.¹² The interaction effect for Germany counteracts this age effect for Britain, indicating that the risk of unemployment is stable for young Germans. Bear in mind that the observation period is relatively short in life course terms: a more comprehensive investigation of life course effects would require a longer observation window.

In model 2 we introduce some variables to capture how class specific risks in the two countries vary over time. In this model we specify a linear historical trend, which is the simplest, given that we have no clear assumptions about the shape of this trend.

From model 2 we can see that, for Britain, consistent with the class stratification perspective, class does not decline in salience as a predictor of unemployment risks over the period. This finding clearly contradicts the arguments that class is declining in salience. Indeed compared to older cohorts, younger cohorts of intermediate and unskilled manual workers seem somewhat more at risk of unemployment relative to the upper service class, though the coefficients are not significant. This finding tends to support the arguments of Breen (1997) that the intermediate group will also suffer increasing economic risk due to technological changes and increases in monitoring by employers.

For Germany, we do not find a reduction in inequality either: the class inequalities we found in model 1 persist throughout the period. The slight rise in inequality for the intermediate class in Britain is not evident, as the interaction term for this class group rather neutralises the trend for Britain. For other class groupings the change in risks relative to the upper service class is similar to Britain. So for Germany, too, our findings rather support the class stratification perspective (e.g., Goldthorpe 2000) and lend no support to the claims that inequalities in class risks have been diminishing over the period. Though note that the discussion here is of relative risks between classes: as seen in Figure 1, the absolute risk of unemployment rose for all classes.

Comparing the two countries over time, our hypothesis of cross-national divergence over time is given some support by the fact that inequalities in class risk are increasing for the intermediate class in Britain but not in Germany. However, clear cross-national divergence is not evident – nor is convergence. The differences in class inequality which exist and persist suggest that global shocks like technological change and resulting rising labour market insecurity have been ‘processed’ by these countries in rather different ways. Manual workers have indeed borne the brunt of rising insecurity in Britain: in Germany labour market risks have been more evenly distributed, though some inequality persists.

The key research questions in this paper were how does the risk of unemployment vary by social class in different cohorts and across countries, and what are the implications of this for class theory? The two countries chosen, Britain and Germany, differ in both key institutions and the structural change they have undergone. Unemployment is a salient and topical life course risk in both countries.

What are our main findings and what can we learn from them? Bear in mind that our empirical analysis is designed to be illustrative: we focus on a sample of young men. Firstly we find different patterns of class inequalities between countries. In our interpretation, we link these different outcomes to institutional differences between the two countries. In general we find less inequality between classes in the risk of unemployment in Germany than in Britain. This is consistent with our hypotheses following the discussion of differences in employment regulation and structural change in the two countries. However, we find little difference in the position of unskilled workers in the two countries: this is in keeping with our hypothesis following the discussion of the different education systems. Our second important finding is that of persisting inequality over time. We find no evidence of diminishing inequality in either country, in fact if anything, class inequality is rising in Britain, particularly affecting the intermediate class. This fits with the preceding account of changes in both legislation and labour demand which particularly affected Britain. So, not only do class inequalities differ cross-nationally, but these differences remain throughout this period.

This paper was written in the context of debates on persisting inequality versus the declining salience of social class. While no class is immune to rising unemployment, our evidence tends to contradict arguments proposing the declining salience of class for labour market outcomes, and to give more credence to the persisting inequality approach. Yet our main contribution to this debate is the following: when considering class inequalities in labour market risks it is crucial to consider cross-national differences and any analysis which assumes that these processes will be the same across countries is very problematic. We clearly reject the idea that there are universal trends, and that all countries change in a similar way.

We should reiterate at this juncture that our conclusions are based on a sample of young men. Including women may have resulted in rather different findings, especially as labour market participation of women of this age group varies substantially across cohorts, and so also the selectivity of the group of female labour market participants. We should also bear in mind that comparing older workers in Britain and Germany might yield rather different results. It is conceivable, for example, that a strong vocational training system is likely to reduce unemployment risks at labour market entry, but exacerbate them later in individual careers due to the difficulty of making transitions between different occupations where qualifications are highly formalised.

The discussion and results in this paper have led us to some reflections on the class schema and the theory of class underlying it for measuring inequality from a comparative perspective. These are summarised here.

The first issue is the role of skill. Class theory is based on the employment relationship. While skill plays a role in the classification used, skill is not an integral part of the theory itself (see also Marshall et al.1988). A closely related point concerning measurement issues is that as skill measurement has no ‘theoretical foundations’, its measurement may not be optimal, and in practice, it may be difficult to distinguish skilled from semi- or unskilled workers. However, even with an approximate skill division, our results for skilled manual workers in Germany show clearly that skill is important for labour market risks. We would argue that the skill differences are important enough to warrant inclusion in the theory of class itself, if one aim of class analysis is to investigate the importance of class in shaping life chances.

A second issue is the lack of focus on institutions. Our results show less class inequalities in the risk of unemployment in Germany, suggesting that institutional differences affect the employment security of different social classes. There are two possible interpretations of our findings, each of which have different implications for class theory.

One interpretation is that the theory is valid as it stands, and in some sense ‘universally applicable’. The role of institutions is in modifying class outcomes, rather than how employment relationships are determined. In this case, we can maintain a theory of class based on labour and service contract, but we need to introduce ‘institutional modifications’ when examining class outcomes like unemployment. Employment protection, for example, may affect the risk of unemployment among different classes, but not the theory of class itself. A second interpretation implies a more radical reworking of class theory. This is that in Germany both how the employment relationship is negotiated (the industrial relations system) and the institutional protection afforded to most employees fundamentally affect the employment relationship and associated risks, as suggested in Section 3.2. This interpretation suggests that class theory needs to incorporate institutional differences into the theory itself to be effective for comparative research (see also Esping-Andersen 1993). This latter argument is given further backing by measurement problems arising from the fact that the remuneration and some work conditions of many jobs in Germany are dictated by agreements between employers and trade unions. How these jobs are classified in terms of their legal and contractual status may be at odds with the kind of work performed and promotion prospects (see, for example, Brauns et al.2000). The relevance of this for cross-national research is that jobs in the same occupation may be classified in different classes, because of differences in how these aspects of the job are determined.

However, while our findings may support such an interpretation, a more rigorous test of this second interpretation would require showing that employment relationships do not evolve simply as employers’ responses to problems of workers supervision, but are affected by other factors such as the strength of trade unions (see also Breen & Rottman 1995: 460). In practice this would involve investigating the employment relationship associated with each individual job title separately, rather than relying primarily on occupational titles, as is current practice. If it were shown, for example, that many manual workers actually had the type of contracts which class theory associates with the service class, this would seriously undermine the usefulness of the class schema for cross-national empirical research, for after all, the schema ‘must be judged by the value that it proves to have in enquiry and analysis’ (Erikson & Goldthorpe 1993: 46). This investigation remains a pressing task for future comparative class analysis.

We would like to thank Karl Ulrich Mayer, Heike Solga, participants at the Spring Meeting of RC28 on Social Stratification in Oxford, April 2002 and the editor of European Societies and three anonymous reviewers for helpful comments and suggestions. Thanks also to Anja Berkes for her assistance with preparing the paper for publication.

Frances McGinnity is a research officer at the Max Planck Institute for Human Development in Berlin. Her research interests are primarily on labour market risks, specifically unemployment and temporary employment, from a longitudinal and comparative perspective.

Steffen Hillmert is an assistant professor at the University of Bamberg and a visiting researcher at the Max Plank Institute for Human Development in Berlin. His research interests are in the fields of education, training and occupational careers, also from a comparitive perspective, and research methods.