External labour market flexibility and social inequality: Temporary employment in Germany and the UK

Persistently high unemployment rates and sluggish economies in almost all European countries have led many commentators to demand higher labour market flexibility in order to fight ‘eurosclerosis’ (e.g., OECD 1986, 1994; Burgess 1992; Siebert 1997; Chen, et al. 1999). One way to increase labour market flexibility is to improve employers’ opportunities to establish ‘atypical’ employment relationships.¹ Through atypical employment employers are given the chance to adapt the number of employees and/or the number of hours worked to current demand levels, thereby avoiding costly overcapacities in personnel. However, the particular forms of atypical employment are different in respect to changing the institutional foundations of the standard employment relationship and may thus have different impacts on the system of social inequality. The common distinction of ‘internal’ and ‘external’ flexibility (Dragendorf, et al. 1988; Matthies, et al. 1994) takes these differences into account. Internal flexibility refers to changes in the length and the location of working time, leaving continuance of employment relations unchanged. In contrast, external flexibility encompasses strategies, which undermine the continuance of employment relationships. In this paper we focus on one type of external flexibility, namely that of temporary employment.²

By establishing temporary jobs, employers are able to circumvent dismissal restrictions stemming from legal regulations and/or collective bargaining agreements. Due to certain institutional settings (e.g., laws on employment protection) it can be rather costly or even impossible for employers to dismiss employees. For this reason, the a priori termination of employment relationships through the use of temporary employment can be an effective way to increase a firm's flexibility. When the economy is doing well, employers can hire additional employees for the time they expect to need them. When the economy slows down, employers can easily adjust their number of employees by not renewing temporary contracts, thereby avoiding any firing costs.

But what are the socio-economic consequences of temporary employment for employees? Do employees systematically earn less in temporary jobs than in permanent jobs? Do temporary jobs foster unstable working careers by leading to further temporary employment or even to unemployment? In this paper we try to answer these questions both theoretically and empirically. In the theoretical part (Section 2) we discuss potential consequences of temporary employment and derive some general hypotheses about socio-economic consequences of temporary employment. As was shown by previous empirical research on career consequences of temporary employment in Germany, temporary jobs do increase the risk of unstable working careers; at the same time, they seem to lower income prospects (Bielenski et al. 1994; Groß 2001; Hagen 2002; Mertens & McGinnity 2002; Giesecke & Groß 2003).

If the assumption holds that temporary jobs allow a higher flexibility when institutional settings establish a ‘rigid’ labour market the effects of temporary employment on income and mobility can be expected to vary between labour markets that differ in their degree of openness. We test this idea by comparing Germany and the UK, two countries which show considerable differences with respect to dismissal restrictions. Section 3 shortly describes the institutions causing these restrictions. In the empirical part (Section 4) we evaluate our hypotheses using data for German and British employees. We present models showing the effect of temporary employment on income as well as on individual working careers.

In the next sections, some labour market theories are presented which help us to develop hypotheses about the consequences of temporary employment on income inequality and mobility. We start by discussing income effects of temporary employment based on Sørensen's theory of closed positions (Sørensen 1983, 2000) and proceed to a discussion of mobility effects.

Sørensen stresses the distinction between ‘open’ and ‘closed’ positions: ‘… positions will be referred to as closed when they are available only when vacated by the previous incumbent …  In contrast, incumbents of positions in open position systems can be replaced at any moment in time …’ (Sørensen 1983: 109). A company may create or eliminate open positions without facing restrictions. In contrast, closed positions cannot easily be eliminated. Thus they are characterized by a greater permanence. It is this very characteristic of closed positions that allows them an existence independent of individuals occupying the positions.

Wages in closed positions are not directly determined by the performance of a particular worker, but by the position itself. In long-term labour contracts, the employee's salary is set for a substantial period of time. The longer the duration of the contract, the smaller the chance that performance-related incentives will be included in the contract. Additionally, employees cannot be fired due to poor performance because they decide themselves when they want to terminate their contracts. Since wages are fixed according to the position and employees control the occupancy of their positions, productivity and salary frequently do not correspond with one another (Sørensen 2000). In other words, incumbents of closed positions are able to acquire ‘positional rents’, meaning that they receive wages that exceed the equivalent of their productivity.

One way to lower the risk of workers’ productivity in closed positions not equalling their wages is to accurately judge the worker's future productivity at the time of hiring. ‘Signals’ of productivity may be used to gauge whether the employee's performance will fulfil employers’ expectation. Aside from sex and ethnicity, educational credentials are the most important indicators employers use to estimate the future productivity of job candidates. Given that a candidate's ascribed characteristics are very important for filling closed positions, these characteristics become important determinants of wage setting and career mobility.

According to Sørensen's argument, employees in closed positions are privileged in that they can acquire ‘rents’. Employees in open positions are then disadvantaged as they cannot acquire such rents – meaning that their wages tend to be lower compared to their counterparts in occupying closed positions.

Temporary employment relationships can be characterised as being more open than permanent ones because employees cannot control the termination of the employment relation. Thus we can put forward our first hypothesis:

Hypothesis I-1: Wages in temporary employment are, ceteris paribus, lower in temporary employment than in permanent employment.³

In addition, we believe that individual characteristics operate differently in permanent and temporary employment. In particular, educational titles will work differently in open and closed positions. This assumption is based on Sørensen's assertion that educational credentials are used as screening devices when hiring people for closed positions. If this holds, we can expect a close correlation between educational titles and wages in systems of closed positions, meaning that educational credentials are important for acquiring positional rents in those positions. In contrast, the power of educational titles to acquire rents will be much weaker in open positions. Thus we come to

Hypothesis I-2: Returns to educational titles are lower in temporary jobs than in permanent jobs.

In addition to the discussed income effects of temporary jobs we expect temporary employment to have important consequences for the individual career mobility.

Three aspects of the consequences of temporary employment for intragenerational mobility are important. First, temporary employment raises the level of intragenerational mobility. As the share of temporary employment increases job duration shrinks and the number of job shifts grows larger on average. Second, the question arises of who receives temporary jobs. Is the risk of becoming temporarily employed the same for all employees or do some groups face a greater risk than others? Third, what consequences does the fact of being in temporary employment have for the working careers of individuals?

Regarding the first point it seems quite obvious that temporary employment tends to raise the level of intragenerational mobility. However, the process of allocating people to temporary jobs and the mobility consequences of temporary employment have to be discussed in more detail.

Employers are interested in binding those employees to the firm, who show a high productivity. The employers’ interest in stable and long-term employment relationships is especially high when specific human capital is needed. This kind of human capital is useful only in a specific firm and useless in other firms. It is costly for both employers and employees to build up this kind of human capital, because it has to be learned on-the-job. Training periods are costly for employers because employees are not fully productive when being trained. They might be costly for the employees as well because employees may not receive full wages while in training programs. If an employee changes companies he/she, as well as his/her former employer, loses the returns from the investment in specific human capital. Also, the costs of the specific training cannot be compensated for. Thus, both employers and employees are interested in stable employment relationships as long as investment in specific human capital is needed.⁴

Things are different for people with no or only low qualifications. They can be recruited easily from the external labour market and need no costly training to perform their jobs. Thus, employers are less interested in binding such employees to the firm. Consequently, it is this group of employees who employers want to dismiss first when economy slows down. This is also true for employees who are qualified, but have general qualifications, i.e., qualifications, that can be used in a wide range of firms. Employees with general qualifications do not lose their investment in human capital when changing firms. Working in different firms even may raise their occupational experience, thereby heightening their human capital. At the same time, employers do not invest in general human capital of their employees, i.e., they do not lose any investments when these employees leave the firm. Also, they can recruit this type of employees from the external labour market without the need of extensive and specialised training. Thus, employees with general qualifications as well as their employers are less interested in stable employment relationships as compared to employees with specific human capital (as well as their employers), who will be more interested in stable employment and long-term relationships. Thus, we can put forward

Hypothesis M-1: The risk of getting a temporary job varies with employees’ qualifications. People with low qualifications or general human capital have a higher risk of getting a temporary job as compared to people with specific human capital.

In the last section we argued that the allocation process to temporary jobs differs between different groups of employees, depending on their qualifications. But can we expect some ‘allocational power’ of temporary jobs? Does the fact of having a temporary job influence the process of career mobility? That is to say, does the fact of having had a temporary job have an effect on the probability of receiving a temporary job again or of becoming unemployed?

Some segmentation approaches (e.g., Kalleberg 1988) suggest such an effect to be observable. According to these approaches, the labour market is divided into different regions with different income determination processes and different mobility regimes. While some regions (e.g., the ‘lower primary labour market’ as Doeringer & Piore (1985 [1971]) define it) consist of ‘good’ jobs, yielding stable, well-paid employment, other regions (e.g., the ‘secondary labour market’) consist of ‘bad’, unstable, and low-paid jobs.⁵ And, most importantly for our investigation, these regions are divided by barriers to mobility that cannot easily be crossed. It is harder for an employee in the secondary labour market to get a job in the primary labour market than it is for an employee with the same individual characteristics who already has a job in the primary market.⁶

What can be concluded from these segmentation approaches as to the effect of temporary jobs on career mobility? If temporary jobs are indicators of unstable employment – meaning that they are mostly to be found in the upper primary or in the secondary labour market – we can expect employees in temporary jobs to face a higher risk of getting a temporary job again as compared to permanently employed employees. At the same time, they would also face a higher risk of becoming unemployed.

Employment stability or security is the key to understanding the effects temporary employment has on mobility and income. Stable positions are not only linked to higher wages, they also enable employees to avoid unemployment as well as employment relations contributing to unstable working careers. Unstable temporary employment relations in turn have negative income effects and may lead to further unstable employment.

As the degree of employment security shows remarkable differences between countries cross-national comparisons of the socio-economic effects of temporary jobs can help to shed some more light on this relationship. Furthermore, in addition to institutions regulating employment security other institutional settings may influence the way temporary employment is affecting individual labour market chances.

We expect country-specific consequences of temporary work for two reasons. First, mobility regimes differ between countries due to differences in educational systems and labour market structures. But if mobility regimes differ between countries, then the effects of temporary employment on mobility structures (and resulting from that, on inequality) may also differ. Second, institutions regulating temporary employment – legal rules and collective agreements between employers and unions – as well as the general degree of employment protection differ between countries, having different effects on distribution and on the consequences of temporary employment.

In order to test these ideas we compare the socio-economic effects of temporary employment in two countries which considerably differ with respect to these institutional settings: Germany and the UK.⁷ These institutional differences and their presumed impact on the consequences of temporary work will be more closely described in the following.

A number of studies point to the fact that differences in the educational systems lead to differences in the patterns of intragenerational mobility (e.g., Allmendinger 1989; Brauns, et al. 1997; Müller & Shavit 1998). Thus we have to take a look at the educational system of the UK and Germany if we want to explore the mobility regimes of these two countries.

The German and the British educational systems differ greatly in various respects. The German educational system is much more standardized than the British one as nationwide curricula and similar standards govern the education and training processes at all levels. In Britain, attempts to establish common curricula have been started lately and so far not been as successful as common curricula in Germany. While in Britain the widespread comprehensive school diminished the degree of stratification of the educational system, the German tripartite system leads to a tight connection between corresponding levels of the educational and the occupational system, meaning that the degree of stratification is rather high (Brauns et al. 1997).

The most important difference between the two educational systems concerns the degree of differentiation. In Germany a ‘dual system’ of apprenticeship training exists (Münch 1985), meaning that trainees are partly educated in vocational schools, partly trained on-the-job in firms. An important point is that both the education in the vocational schools as well as the training in the firm is standardized. At the same time, apprenticeships provide specific, production-related knowledge needed for a given occupation. Moreover, this knowledge is tied to a certain educational title. As a result, these titles signal that their owners have the qualifications needed to fulfil all tasks of a certain occupation.

Training on-the-job, which provides knowledge needed for the required tasks in a given firm, is of less importance in Germany. Since holders of apprenticeship certificates already have a broad range of knowledge needed for their occupational tasks, only minor training courses have to be offered to adapt this broad knowledge base to the specific requirements of the firm. Employees with such vocational qualifications can easily change firms without having to fear the loss of their human capital. Moreover, since employees with apprenticeship certificates are viewed as highly educated and as possessing valuable human capital, employers are interested in binding these employees to their firms.

In Britain, the degree of differentiation of the educational system is much lower. The system of ‘further education’ provides a wide range of educational titles, each of which stands for a very special, narrow range of knowledge. Therefore, if the occupational tasks require further qualifications, which are not related to the educational title, these qualifications have to be gained on-the-job. Thus, training on-the-job is much more common in the UK as compared to Germany (Glowka 1996).

The difference between the German and the British educational systems results in important differences in the patterns of intragenerational mobility of both countries. First, since training on-the-job is of greater importance in the UK, employees often have to change jobs within or between firms to acquire knowledge important for a given occupation, meaning that the degree of mobility is higher in Britain than in Germany. Secondly, educational titles tend to have a greater impact on career mobility in Germany than in Britain.⁸ This particularly holds for the apprenticeship certificates, resulting in the greater importance of occupational labour markets as compared to firm internal labour markets (Blossfeld & Mayer 1988).

For our study, we would expect two effects resulting from these differences in the educational systems. First, we expect that in Germany vocational education protects employees from getting temporary jobs because employers are interested in binding people with valuable qualifications to their firm. In Germany, apprenticeship certificates partly take over the effects that we expect of specific human capital, because these certificates guarantee a certain level of production-related knowledge. In Britain, educational titles play only a minor role in the risk of receiving a temporary job. Secondly, we expect the differences in returns to educational titles between permanent and temporary jobs to be larger in Germany than in the UK because the power of educational credentials to acquire positional rents is larger in Germany. Thus we get two further hypotheses, one on mobility and one on income effects of temporary employment.

Hypothesis M-3: The effect of educational titles on the probability to receive a temporary job is greater in Germany than in the UK. In particular, vocational educational titles reduce the probability of getting a temporary job.

Hypothesis I-3: The loss of income returns to educational titles in temporary jobs is larger in Germany than in the UK.

In Germany, legal regulations of unfair dismissal can be seen as the most important source of employment protection. Collective bargaining agreements at both the industry and the firm level were always secondary to ensuring employment protection (Birk 1993). Legal regulation of unfair dismissal is basically based on two laws: the Dismissal Protection Act (Kündigungsschutzgesetz) of 1951,⁹ which enhanced the rights of individuals, and the Amended Works Constitution Act (Betriebsverfassungsgesetz) of 1972, which established mandatory consultations with the relevant works council before each dismissal. In the German law against unfair dismissal the concept of ‘just cause’ is of central importance. Only if the employer can show just cause are terminations of employment relationships legal.¹⁰ Otherwise terminations of employment relationships are regarded as being invalid, thus giving the employee the right to either continue working or to dissolve the employment contract and receive a compensation payment. Importantly, in Germany, additional employment protection is provided to employees working in the public sector. Public servants (Beamte) are subject to a special law (Bundesbeamtengesetz) granting them life-long employment (after a certain probation period). Employees in the public sector who are not public servants do not receive a life-long employment guarantee, but their employment relationships are also very secure. Therefore, employment security is particularly strong in the German public sector. In short, the German employment protection regime shows a substantial degree of restrictiveness vis-à-vis terminations of employment relationships.

In Britain a different picture emerges. Here collective bargaining agreements are more important for employment protection than legal regulations on job security. The British law contains some redundant and unfair dismissal legislation (Bercusson 1993). However, these regulations were cut back or even abolished by the Conservative Government during the 1980s and much of the 1990s,¹¹ so that the British dismissal-protection regime is seen as being rather unrestrictive (OECD 1999). The British Conservative Government also pushed back the influence of the trade unions, which traditionally were an important source for employment security (Hyman 1993; Towers 1993; Kastendiek 1998) and strongly decreased employment protection in the public sector (Johnson 2001), making present employment patterns in the public sector quite similar to those in the private sector.¹² In sum, the British labour market seems to be rather open and does not show a division between public and private sector with respect to the level of employment protection as is the case in Germany.

The regulations governing the use of temporary employment also differ substantially between the two countries. In Germany, following the jurisdiction of the Federal Labour Court (Bundesarbeitsgericht), the use of temporary employment needed a specification of factual reasons.¹³ This was changed with the Employment Promotion Act (Beschäftigungsförderungsgesetz) of 1985, which provided the right to establish temporary jobs without justification,¹⁴ allowing temporary contracts with a maximum duration of 18 months (this number was increased to 24 months in 1996). Renewal of temporary contracts is possible if the new contact is covered by factual reasons, but reasons for renewed contracts are subject to closer scrutiny, thus making it harder to repeatedly renew temporary contracts.

Until 2002 there were no statutory restrictions on the use of temporary contracts in the UK. Consequently, there were no restrictions on the length and the renewal of temporary contracts. Thus ‘chains’ of temporary employment relationships were easily possible. However, as some commentators argued, the 2-year qualification period for legal employment protection may have served as a functional equivalent for temporary employment (Schömann et al. 1998; Robinson 1999). With the Fixed-term Employees (Prevention of Less Favourable Treatment) Regulations 2002 there is now some sort of protection against permanent renewal of temporary contracts. Renewals are illegal if the employee was continuously employed for a period of 4 years or more and the fixed-term contract was not justified on objective reasons.

What follows from these institutional differences between the UK and Germany for the effects of temporary employment on social inequality? Temporary jobs as a means of opening closed positions are only needed in systems of closed positions. Establishing temporary jobs in order to undermine employment security makes no sense if the employment security of permanent jobs is already low. Moreover, if the employment security of permanent jobs is low we would not expect income disadvantages resulting from temporary jobs. If permanent jobs can be terminated at will by employers, they are open by definition. Thus their incumbents cannot acquire positional rents, and we cannot expect any difference in wage levels to temporary jobs.

Thus, we expect the effects of temporary employment on income to be larger in Germany than in the UK, because the German labour market is characterized by a higher employment security. Moreover, the public sector in Germany is more closed than the private sector, because employment security is especially high in this area of the German labour market. The fact that the German labour market, and the public sector in particular, is more closed than the British one may affect all consequences of temporary employment on income and mobility, but we will focus on three additional hypotheses:

For our analyses we use data from the German Socio-Economic Panel (GSOEP)¹⁵ and from the British Household Panel Study (BHPS).¹⁶

The GSOEP is an annually repeated representative study of private households in Germany, first conducted in 1984. We use 16 waves (1984–1999) of this data set for our analyses. The GSOEP consists of five sub-samples, two of which we use for our mobility analyses. Sample A (start: 1984, West Germany, Germans and foreigners, excluding those foreigners in sample B) and sample B (1984, households whose head is Turkish, Spanish, Italian, Greek or a former Yugoslavian). We limit our analyses to these two samples because the remaining ones have different starting points and because there are analytical problems when labour market processes in West Germany are compared with those in East Germany (Sample C).¹⁷ These analytical problems should be addressed by running separate models for each region rather than by combining the data. Thus, the analyses in this paper are confined to West Germans. Respondents who moved from West to East Germany are not excluded from the sample.¹⁸

The second data set, the BHPS, is an annually repeated representative study of private households in the UK, first conducted in 1991. We use nine waves (1991–1999) of this data set.

While in the income analyses cross-sectional data is used (wave 1998 of both the GSOEP and the BHPS), the mobility analyses are based on longitudinal data. Here, the units of analysis are job and unemployment spells, not individuals. For each individual there are several of such spells, which can be characterized by information stemming from the individual waves (e.g., information about the current job, but also information about the previous job). Besides the information about the current and last jobs and their characteristics, each wave also contains detailed data about the respondents’ working activities during the past year, including information about periods of unemployment. By combining these two types of information we are able to analyse individual working careers. In the case of the BHPS we use the Combined Work-Life History Data,¹⁹ which combines the information in the same way as we did for the GSOEP.²⁰

In both the income and the mobility analyses we excluded self-employed persons, apprentices, and students.²¹ Furthermore, we confined the sample to persons aged 16–65. Given this definition of the working population, the incidence of temporary employment ranges between 5 and 9 per cent of all employment relationships.²² Interestingly, there does not seem to be a big difference in the usage of temporary employment in Germany and the UK, which may be explained by the relatively strict regulations on temporary employment in Germany. Furthermore, in both countries women are more likely than men in temporary jobs, but the gender gap is larger in the UK.²³

The variables used try to capture important individual as well as job characteristics. Individual characteristics are: education, age, nationality, the number of children, and the marital status. Education is measured on the basis of the CASMIN classification (Brauns & Steinmann 1999) in order to establish a comparable indicator of education. We combined the nine original categories into four categories: Casmin low (being the reference category), Casmin middle vocational, Casmin middle general, and Casmin high.²⁴Age is measured in years;²⁵age squared is used for checking the non-linear effect of age. Nationality distinguishes German (0) and Non-German (1) nationality in case of the GSEOP, and British (0) and Non-British (1) in case of the BHPS. Moreover, we use the number of children in the household and the marital status distinguishing married persons (reference category) from singles, divorced, and widowed respondents.

Beside these variables capturing individual characteristics, a range of labour market variables is used in the models. Labour income is measured in log gross hourly wages. This is the dependent variable for the income models. Sector is coded 0 for the private and 1 for the public sector. Firm size distinguishes large enterprises (more than 200 employees), middle-size companies (25–199 employees for the BHPS, 20–199 employees for the GSOEP), and small firms. Several variables capture the character of the current or last job. Most importantly, type of contract is coded 1 if temporary, and 0 if permanent. Part-time distinguishes marginal employment (less than 15 hours a week) from regular part-time (less than 35 hours) and full-time jobs (being the reference category).²⁶

For the mobility analyses we use two additional variables: the number of previous periods of unemployment coded 0 for none, 1 for one, and 2 for two and more²⁷ and the variable unemployed, distinguishing unemployment from other states of activity. The latter variable is used as the dependent variable in the models which investigate the effects temporary jobs have on the risk of becoming unemployed after the termination of a temporary job.

In the following empirical analyses all models control for these individual and labour market variables, but only the effects which are relevant for our hypotheses are discussed and shown in the tables.²⁸

The purpose of this section is to investigate the income effects of temporary employment. The presented models are based on regression models (dependent variable: log gross hourly wages) using cross-sectional data stemming from GSEOP and BHPS waves 1998. All models contain the common covariates as education, age, age squared, citizenship, part-time/full-time employment, sector, firm-size, marital status, and number of children in the household. In order to keep this paper readable, we are not going to interpret the effects of these covariates, nor are we showing them in the tables.

The results of the income analyses for German men and women are presented in Table 1. Men in temporary employment earn about 18 per cent less than men in permanent employment. It is important to note that this loss in income cannot be explained by individual characteristics of employees holding temporary contracts, because these variables are included in the model. Nor can the loss be explained by other job characteristics, which are also controlled for. Thus, the income loss is a consequence of the temporary character of the job itself. This strongly supports the assumption that temporary jobs bear structural disadvantages resulting from the loss of employment security.

This conclusion is underlined by the interaction effects of temporary employment with education and sector which are included in the model (see second and third columns of Table 1). In the second model we can see that the difference between permanent and temporary employment cannot be found among the less educated people (the main effect of temporary employment disappears when the interaction effects are included in the model), but mainly among the highly educated employees. Only employees with tertiary educational certificates earn significantly (and with nearly 28 per cent substantially) less than people with the same certificates in permanent employment.²⁹ This confirmation of hypothesis I-2 clearly indicates that educational credentials cannot be used in open positions for acquiring positional rents as Sørensen's theory of closed positions suggests. Finally, the last model (third column) shows that the disadvantages of temporary employment are much greater in the public sector than in the private sector. While people in temporary jobs in the private sector earn about 12 per cent less than people in permanent employment, the disadvantage of employees in temporary employment in the public sector raises by another 19 per cent points.

Looking at Table 1 we see that our hypotheses are not as strongly supported by the results for the German women. The main effect of temporary employment is substantial and significant, but smaller than the effect for men. As for men, the income loss rises with education, with the only significant income difference being that of women with tertiary education. However, in contrast to men, the interaction effect of temporary employment with public sector is not significant, meaning that for women there is no additional income loss for temporary jobs in the public sector.

All three models investigating the income effects of temporary employment seem to support the hypotheses about the income effects that were developed based on Sørensen's theory of closed positions. Temporary employment leads to an income loss because employees in open positions do not have the power to acquire positional rents. This loss depends on education, since educational credentials have the power to acquire such rents – but not in open positions. And finally, in Germany the difference between temporary and permanent employment is largest in the public sector because this sector is in general more closed than the private sector, though this only holds for German men.

In Table 2, the effects of temporary employment on wages in the UK are presented. For men in temporary employment is connected with a substantial wage reduction (about 21 per cent). This contradicts our hypothesis, since we assumed that the effect of temporary employment on income would be larger in the more-closed German labour market. However, the British data set allows us to distinguish between two types of temporary jobs: fixed-term contracts and seasonal or casual work. We calculated some additional models to control for this distinction (results not shown here). In fact, only the latter type of employment leads to a lower income. Fixed-term contracts do not show any negative income effects after controlling for the covariates used in our models. Thus, it seems that the negative effect of temporary employment in Britain is not as serious as first thought.³⁰

Furthermore, as can be seen in Table 2, the interaction effects of education and temporary employment are not present. That is to say that the disadvantage of temporary jobs is about the same for all educational groups. This is in line with our hypotheses, in which we assumed that in the UK educational titles can far less be used as credentials that acquire positional rents because the British educational system is less standardized and less stratified than the German one. At the same time there is no difference between wages of temporary jobs in the public vs. the private sector (no significant interaction term in the third column), which we expected because the degree of closure of the British public sector does not differ from that of the private sector (Hypothesis I-5).

The assumption that temporary work is less disadvantageous in the UK than in Germany is underlined by the results for women, because we have not found any significant (main or interaction effects) of temporary jobs for British women.

In Germany we have identified three clear effects of temporary employment on income. On average, wages are lower in temporary employment than in permanent positions. The disadvantage of temporary employment rises with the level of education. Furthermore it is greater in the public than in the private sector. Thus, these results support our assumptions that in Germany temporary employment is a means of opening closed positions (particularly in the public sector), thereby lowering the power of credentials to acquire rents. However, the effects are stronger for men than for women.

In Britain, we can also find an income loss for men in temporary jobs, though this effect mainly stems from casual or seasonal work and not from fixed-term contracts. That is to say only a certain type of temporary jobs leads to lower wages. Moreover, the effect of temporary employment does not differ with educational credentials, and there is no difference in the effects between public and private sector. Additionally, we do not find any effect of temporary jobs on wages for British women. Thus, our results indicate that the wage effects of temporary employment in total are weaker in Britain than in Germany.

In our mobility analyses three different aspects of temporary employment are evaluated. First, we want to investigate who is employed in temporary jobs. This is done by analysing the influence of the various independent variables on the probability of getting a temporary job (versus permanent jobs) by using logistic regression. Second, we explore the effect temporary employment has on the risk of being temporarily employed again. Here, the models focus on the risk of getting a fixed-term contract after experiencing a job shift. Third, we evaluate the relationship between temporary employment and subsequent unemployment.

For all models, every spell that could be identified as being a temporary or permanent job or as being a period of unemployment was included in the regression. To control for the effect of correlation between spells belonging to the same person, we used random-effects logit models (cf., e.g., Conway 1990), which permit the use of logit models for panel data. In contrast to simple logit models, a random-effects model takes the panel information into account in order to estimate the regression coefficients. As in the income analyses we calculated separate models for men and women.

The risk of getting a temporary job (versus getting a permanent job) is analysed in Model I (see Table 3). Here, three results are important.

First, it seems that temporary employment is used more often in the public than in the private sector. This result particularly holds for German men, who bear a substantially higher risk than women. Thus our hypothesis M-4 is supported by the data. Obviously, temporary employment is used to open the rather closed positions in the public sector in face of the prevalent saving constraints in this area.

Second, when proceeding to the human capital variables it becomes obvious that education has a significant impact on the risk of holding a temporary job. People with only limited education as well as people with general (higher) qualifications are more likely to be found in jobs with temporary contracts, while vocational qualifications decrease this risk. This gives support to hypothesis M-1. Interestingly, women with higher education face an even higher risk of holding a temporary job in comparison to women with only limited education. The second human-capital variable also indicates the expected effect. The highly significant coefficients of age and age squared both for men and women indicate a non-linear impact of age on the risk of having a temporary contract.³¹ The non-linear function corresponds to a U-shaped curve, which reaches its minimum at approximately 42 years for men and 38 years for women, thus suggesting that both young and old people face the highest risk of holding a temporary job when changing jobs. Specific human capital, which is highest in a person's middle age, tends to protect employees from being in temporary employment. Thus, this result also supports hypothesis M-1.

Thirdly, previous periods of unemployment increase the risk of being temporarily employed. Compared to people without or with only one period of unemployment, people with two or more of such periods face the highest risk of having only a temporary contract. These results clearly indicate that formerly unemployed people have fewer chances of getting permanent and secure jobs when re-entering the workforce,³² which is in line with hypothesis M-2.

In Model II (Table 4) only spells of people who changed their job at least once were analysed. Thus, people without a shift in jobs were excluded from the analysis. Additionally, the first job spell of people experiencing a job shift must be excluded from the analysis simply because there is no information about the type of contract of the last job. These necessary restrictions result in a smaller sample size compared to Model I. The dependent variable and the independent variables are the same as in Model I except for an additional variable measuring whether the last job was temporary.

The results indicate that people have a lower chance of finding a new job that is permanent, when their most recent job was a temporary one. This holds net of the influence of many important individual and labour market variables, which are controlled for in the model. Thus there seem to be an impact of temporary work itself on the individual working career. Moreover, for men temporarily employed in the public sector, the chances of finding permanent positions are even worse compared to men temporarily employed in the private sector, as can be seen in the significant interaction term (see second column of Table 4). However, this effect cannot be found for women (fourth column).

The results of Model II clearly show that there are ‘chains’ of temporary employment. This risk is highest for jobs in the public sector, but also temporary jobs in the private sector are followed by new temporary jobs. Considering the risk of being employed in ‘chains’ of temporary jobs, it seems that temporary work can actually turn out to be a ‘trap’ (Büchtemann & Quack 1989) for the employees by making it more difficult to find permanent jobs.

In Model III (Table 5) we examine the impact of temporary work on the risk of becoming unemployed. We coded a spell as ‘unemployed’ when there was registered unemployment for at least 3 months. This was done in order to exclude unemployment periods of search-unemployment. The reference category of the dependent variable is ‘not unemployed’. This category encompasses periods of employment, but also periods of neither employment nor unemployment (e.g., maternity leaves) followed by a period of employment.³³ Because many quite different activities are counted, this category is a very heterogeneous one.

In Model III two points are most interesting with respect to our research question. First, as can be seen from the table, the public sector strongly decreases the risk of becoming unemployed. This result underlines the structural differences between jobs in the public and the private sector in Germany. Obviously, jobs in the public sector are better protected compared to jobs in the private sector. Secondly, we can find an effect of temporary jobs. By looking at the coefficient of temporary work it becomes clear that temporary contracts significantly increase the risk of becoming unemployed. However, when taking the interaction of public sector and temporary employment into account it becomes apparent that the risk of becoming unemployed after the termination of a temporary job is only increased for jobs in the public sector. The results hold for men and women and underline the relatively strong divide between the public and private sector in Germany. On the one hand, the German public sector offers rather closed and thus stable positions, thereby protecting incumbents from becoming unemployed. On the other hand, it also offers rather open positions (temporary jobs) which are far less stable, resulting in an increased risk of becoming unemployed for the incumbents of such positions.

As Model I did for Germany Model IV investigates the risk of getting a temporary job (versus getting a permanent job) for the UK. Table 6 highlights three findings.

First, when looking at the sector variable, we find that jobs in the public sector are more often on a temporary basis than positions in the private sector. This effect is slightly higher for women than for men.

Second, regarding the impact of human capital variables, it appears that education has a much weaker effect on the risk of holding a temporary contract in Britain as it has in Germany. Only the highly educated women face an increased risk of being in temporary employment, while the remaining educational groups show no difference in the risk of getting a temporary job. This implies that in Britain, vocational qualifications do not protect employees from being temporarily employed. Obviously, British vocational qualifications do not signal the existence of production-related knowledge that has to be protected by long-term employment relationships, which is in line with hypothesis M-3. For age we find a non-linear effect that resembles the effect we found for Germany. The minimum of the corresponding quadratic function is reached at approximately 38 years (men) and 34 years (women). Again, this indicates an increased risk of getting a temporary job for younger and older employees when changing their jobs. This result seems to give further support to our hypothesis that specific human capital, which is highest in a person's middle age, tends to protect employees from being in temporary employment.

Thirdly, periods of previous employment strongly affect the probability of being in temporary employment. For men, this effect is substantially higher than for women.

Model V (Table 7) investigates the probability of receiving a temporary job after a job shift has occurred. As in the German case we find a significant effect of the previous job being temporary, meaning that employees holding temporary contracts are more likely to be in new temporary employment when changing jobs compared to their colleagues in permanent positions. This particularly holds for British women, for whom the effect is slightly higher compared to men. However, while for Germany we found a significant interaction effect of the type of contract of the previous job and the public sector, we cannot find such an interaction for the UK. There seems to be no difference in the impact of temporary employment for the public and the private sector, a result which supports hypothesis M-4.

Finally, Model VI (Table 8) investigates the risk of becoming unemployed. As in Model III for Germany we are mainly interested in the effects of two variables: the public sector and the type of job contract.

As indicated by the results temporary employment increases the risk of becoming unemployed, both for men and for women. However, this risk is not different for the public and the private sector, as can be seen from the coefficient of the interaction term, which is not significant. This result differs from the findings for Germany, where we observed an increased unemployment risk of temporary jobs only for jobs in the public sector.

The patterns of allocation to temporary jobs are in some respects quite similar for Germany and the UK. We found that certain job characteristics, such as for example jobs in the public sector, increase the risk of holding a temporary contract. Furthermore, individual characteristics, such as age and unemployment experience, are also related to the type of job contract a person holds. But we find also some differences between the allocation patterns of the two countries. The most interesting difference can be found for the effect of education. In Germany vocational qualifications tend to protect employees from being in temporary employment, while in Britain these qualifications do not protect employees. This result can be explained by the higher degree of standardization of the German educational system, which enables employers to rely on vocational educational titles as being signals for employees’ productivity.

Regarding the impact of the last job being temporary on the probability of getting a temporary contract we found that temporary employment increases the risk of getting a temporary job again in both countries. However, there is a notable difference between Germany and Britain. While for the UK this effect is equal for jobs in the private as for jobs in the public sector, we found that for German men temporary jobs in the public sector show significantly stronger effects compared to jobs in the private sector.

Finally, we found that temporary employment increases the risk of becoming unemployed. In Britain this risk is the same for temporary jobs in the private and the public sector. In Germany people temporarily employed in the public sector have a higher risk of not finding a new job in a short period of time compared to employees temporarily employed in the private sector.

Our results strongly support the assumption that temporary employment is a means to open closed positions. This has two important consequences for the structure of social inequality. First, stability of individual working careers is decreased. Temporary employment leads to ‘chains of uncertainty’ by increasing the risk of being temporarily employed again or by becoming unemployed. Secondly, temporary jobs have consequences for income inequality. They are not only less paid, but they also change the process of income determination. In temporary jobs educational titles have a lower impact on wages than in permanent jobs. Importantly, the results reveal that the risk of holding a temporary job is strongly dependent on human capital. Those employees with low human capital or with high general human capital face a higher risk of being temporarily employed as compared to people with specific human capital, who are protected from being employed on temporary contracts.

However, these consequences for social inequality depend on the basic structure of the labour market in a given country. Though the gross effects of temporary employment on income seem to be quite similar in Germany and in the UK, we found considerable differences in the returns to education in temporary jobs. While in Germany, educational titles tend to lose their power to acquire positional rents in temporary jobs, we cannot observe such an effect in the UK. Moreover, the income effect differs within Germany between the private sector and the extremely closed public sector, whereas in the UK the income effect is the same for the two sectors. The relatively strong divide between the public and the private sector in Germany is also underlined by the results of the mobility analyses. ‘Chains of uncertainty’ can mainly be found for employees holding temporary jobs in the public sector in Germany. In the UK no such sectoral differences in the mobility effects can be found.

Thus, our results point to the fact that Germany's and the UK's different ways of coping with the demand for labour market flexibility lead to different effects of temporary employment on social inequality. In Germany, employment protection regulations of regular jobs remained more or less unchanged and flexibility is mainly based on the use of atypical employment. Consequently, we can clearly distinguish two groups of employees: on the one hand those employees who can be found in standard employment relations, which are well paid and offer good and stable career prospects (especially in the public sector). On the other hand we can find those employees working in temporary jobs, which are less paid and have worse career prospects. Thus, the costs of labour market flexibility are mainly borne by employees in atypical jobs, while employees in standard employment relations are to a large extent protected from the negative effects of increased labour market flexibility. In the UK, which adopted a strategy that made the labour market in general more open, the divide between standard and atypical jobs is less emphasized – the costs of flexibility are more evenly spread over all employees.

These results are pretty much in line with the labour market theories we discussed in Section 2. However, there are some effects that need further exploration. Most noteworthy, we found interesting gender differences, particularly with respect to the wage effects of temporary employment. Both in Germany and in the UK women's wages seem to be less affected by temporary employment than men's wages. As Booth et al. suggest, this may reflect women's wish ‘to retain career flexibility through a significant portion of their work lives. In this case, it can be optimal to invest in a high level of general, rather than specific, human capital, and to hold a succession of temporary posts’ (Booth et al. 2002: 192). Furthermore, women's self selection into employment may be partly responsible for the found gender differences. Further research is needed in this area.

Financial support from the Deutsche Forschungsgemeinschaft (DFG) (Project ‘Die Auswirkung atypischer Beschäftigung auf das System sozialer Ungleichheit’) is gratefully acknowledged. We thank the UK Data Archive and the Deutsche Institut für Wirtschaftsforschung for providing The British Household Panel Survey and the German Socioeconomic Panel, respectively.

Dr. Martin Groß is senior researcher at the Institute of Social Science at Humboldt University Berlin. He has carried out a number of empirical studies on educational systems, social inequality and labour markets. Recent publications include (with J. Giesecke) ‘Temporary Employment: Chance or Risk?’ European Sociological Review 19 (2): 161–77.

Johannes Giesecke is junior researcher at the Institute of Social Science at Humboldt University Berlin. His current interests include research on social inequality, labour market flexibility, and quantitative methods.