Temporary Work, Labour Market Careers and First Births in Sweden. Ola Sjöberg. Working Paper 2009: 7

Transcription

1 Stockholm University Linnaeus Center on Social Policy and Family Dynamics in Europe, SPaDE Temporary Work, Labour Market Careers and First Births in Sweden Ola Sjöberg Working Paper 2009: 7

2 Temporary Work, Labour Market Careers and First Births in Sweden Ola Sjöberg Swedish Institute for Social research Abstract: This paper analyzes the effect of temporary employment contracts on labour market and demographic careers in Sweden. The incidence of this form of employment contract has increased rather dramatically on the Swedish labour market in the last few decades. A central question for evaluation of the welfare effects of temporary work is the degree to which such contracts might be considered to be a stepping stone into permanent employment. Using life-event data from the Swedish Level of Living Survey, it is shown those whose first job was temporary have a significantly higher risk of also having a temporary job 10 years after labour market entry. However, he results also indicate that for those whose first job was temporary, continuing to accumulate experience from temporary jobs (as opposed to being unemployed or out of the labour force) is helpful for getting a permanent job. There are also striking gender differences in the risk of entering end exiting temporary employment: there is a higher likelihood that a woman s first job will be temporary, and there is a lower likelihood that women will enter permanent employment if their first job was temporary. Moreover, young women especially have a higher risk of ending up in a temporary job carousel, where exits from temporary jobs are to other temporary jobs. There is no evidence to suggest that having a temporary job as such has a negative effect on the risk of transition to first birth. However, the sample analyzed here is young, and the results suggest that temporary employment have a much more pronounced effect on first birth propensities in older women. Moreover, transitions from temporary to permanent employment have a significant positive effect on first births. 1

3 The purpose of this chapter is to analyse the effect of temporary employment (i.e. employment contracts of fixed duration) on labour market and demographic careers in Sweden.The incidence of temporary employment has increased rather dramatically in the last few decades in Sweden, especially during the 1990s. In 1999, this form of employment contract accounted for over 15 per cent of total employment, as compared to around 10 per cent in This expansion coincided with the most dramatic period in the Swedish labour market to date since the depression in the 1930s. In a few years at the beginning of the 1990s, unemployment rates increased from two to over eight per cent, and employment fell by close to 500,000 people, which is equivalent to about 16 per cent of the labour force. In the public debate, discussions about temporary employment tend to be polarised between those arguing that fixed-duration contracts are necessary for employers to successfully adapt to changes in demand for their products and services, and those arguing that while such contracts may serve the needs of employers, they might also have negative implications for the welfare of individuals. When studying the welfare implications of the increased prevalence of temporary jobs in Sweden, we must not make the mistake of equating temporary jobs with bad jobs; employment contracts with a fixed duration can serve important purposes not only for the employer but also for the individual worker. Even so, there is evidence that at least some types of temporary employment in Sweden are associated with, for example, higher risks of becoming unemployed and lower levels of on-the-job training. The long-term consequences of temporary employment from a welfare point of view may therefore depend to a large extent on whether such jobs are in fact temporary and transitory. This is not least the case for fertility, where it has been shown that holding a temporary job may negatively affect the incentive to have a first child. Whereas previous studies on, above all, labour market but also demographic consequences of temporary work in Sweden have generally considered a rather short follow-up period, this study will follow individuals born in 1950 and after from their first entry into the labour market up until the year This chapter therefore fills an important gap in the literature on the long-term consequences of temporary work in Sweden. The chapter is structured as follows: the first part will provide an overview of social policies and labour market developments and regulations in Sweden, with particular emphasis on temporary employees access to welfare benefits and the regulation of fixed-term contracts. The next section will provide a summary of the development of temporary work in Sweden since the late 1980s and also an overview of the research on labour market and demographic consequences of temporary employment in Sweden. After introducing the data used, 2

4 the results from a number of regression models (both logistic regression and event-history models) will be presented that will highlight the consequences of temporary work on labour market and demographic careers in Sweden. With these models, we will analyse determinants behind entries into and exits from temporary employment as well the effect of this form of employment contract on first births. Social policies and labour market regulations in Sweden The Swedish welfare model is characterized by a broad scope of public social policy, the central ingredients of which are an income security system where universal flat-rate benefits are complemented by an earnings-related component for those with a work history, as well as publicly funded social and health-care service production to cater for all needs and the whole population (Kautto et al. 2001). Earnings-related benefits emphasises the importance of stable employment with relatively high earnings. Although there is no explicit discrimination against temporary jobs in the income security system, existing qualifying conditions imply that individuals need a more or less stable attachment to the labour force in order to be eligible for various benefits. For example, the work condition in the unemployment insurance scheme (which currently stipulates that during the year preceding unemployment, the unemployed must have worked at least 6 months and at least 80 hours each month, or alternatively, must have worked at least 480 hours during six consecutive months and at least 50 hours each month) means that individuals with short, temporary employment and recurrent gaps between periods of employment may have difficulties in qualifying for earnings-related benefits. Similarly, to qualify for income-related parental leave benefits, a certain amount of uninterrupted work (currently 240 days) is required. Family policy in Sweden is based on the dual-earner family model (Korpi 2000). Central to this model is the combination of earnings-related parental leave (currently at a rate of 90 per cent of gross earnings for 16 months), access to publicly subsidised day care and separate taxation of spouses. When an employee is absent because of parental leave, a worker is often employed on a temporary contract to replace the absent worker. The expansion of parental leave legislation in Sweden is therefore important for explaining not only the overall development of temporary work in the form of replacement for absenteeism, but also for explaining the development of temporary work in different sectors of the economy (private vs. public) and between men and women. Since 1979, all parents employed full-time for at least six months also have the right to reduce their working hours to 75 per cent of full time until the child is eight years old. The expansion of the public sector and the development of the dual-earner model have led to comparatively high levels 3

5 of female labour force participation in Sweden; from a participation rate of around 60 per cent (ages 20-65) in 1970, it increased to between 78 and 85 percent in the 1990s and early 2000s. Since many women got jobs in the growing public sector, above all in health-care and social services, the Swedish labour market is rather gender-segregated. Moreover, there is also a gender bias in part-time work: in 2002, around one-third of all employed women worked less than 35 hours per week. However, a comparatively large proportion of women in Sweden work what can be called long part time, i.e. between 30 and 35 hours a week. Traditionally, an important feature of the Swedish welfare model has been a political commitment to full employment. Central to achieving this goal was the implementation of the so-called Rehn- Meidner model in post-war Sweden (Milner and Wadensjö 2001; Erixon 2008). This economicpolicy model combined external-numerical flexibility and geographical mobility, largely to the benefit of employers, with the security for employees provided by full employment and labour market and social policies. Using extensive labour market polices such as re-location and retraining grants, occupational training programmes and actions to improve the matching on the labour market, and also well-developed income maintenance schemes, the Rehn-Meidner model aimed at combining full employment and economic growth with price stability and equity. It can therefore be argued that the three components of what has been known as the flexicurity triangle flexible rules for hiring and dismissing, a high degree of income security in case of unemployment, and active labour market policies played central roles in Sweden s economic and social policies early on. Although employment security was an essential part of the Rehn-Meidner model, it was not a promise of security in a particular job a worker had at any time. Instead, employment security was to be provided through mobility in the external labour market, by the supply of alternative jobs through demand management and information about those jobs, training needed to do them and income maintenance during the transition to them (Martin 1995). Sweden also managed to keep unemployment rates at comparatively low levels for a long time, but in the early 1990s the country moved from full employment to mass unemployment. This particularly affected groups such as young people and immigrants. Between 1990 and 1999, the age at which 75 per cent of a cohort had found employment (the establishment age ) increased from 21 years for both sexes to 26 years for men and to 30 years for women. Partly as a result of the developments in the labour market, the proportion of young people who studied at universities increased sharply in the 1990s, and this decade also saw a substantial expansion of the adult education system (Palme et al. 2002). In this context, it is also worth mentioning that the median 4

6 age at which students commence their higher education in Sweden is high from a European perspective: close to 23 years. As regards immigrants, labour market ties are generally poorer the shorter the time the person has been in Sweden and the less economically developed the part of the world he or she comes from (Palme et al. 2002). Of particular importance when explaining the development of temporary work are the legal regulations of fixed-term contracts. In 1974, employment protection legislation (LAS) for the whole labour market was instituted in Sweden. Up until then, the only regulation of employer s right to hire employees on fixed-term contracts was through collective agreements. This meant that there were no general restrictions to hiring employees for a limited period of time. According to LAS, however, every employment contract should be considered to be of unlimited duration unless otherwise stated, and fixed-term contracts were only allowed in special cases. Although this law therefore meant a significant strengthening of the employment protection of employees with standard open-ended contracts, it also introduced the possibility for employers to hire employees on a fixed-term contract in specific circumstances. These specific circumstances were, for example, in the form of hiring a replacement for an absent employee (leave replacement) and work that was inherently of fixed duration (project work). It has been argued that LAS is the first and also the most important departure from the traditional way of regulating labour market relations through collective agreements in Sweden, and that no other law has had such far-reaching consequences for the functioning of the Swedish labour market as LAS (Nycander 2008). 1 In the mid-1970s, a review of LAS was initiated. The main motive was to ease the restrictions regarding temporary employment. In the new legislation of 1982, collective agreements got a weaker position relative to legislation. Most important, this reform gave employers a legislated right to probation employment and to use fixed-term contracts at demand peaks. Fixed-term contracts were also allowed in situations when there was a natural time limit for the employment (such as before military service), and also when employing old-age pensioners. This act also stipulated that temporary employment was allowed if the nature of the work itself gave reason for this form of employment contract: for example, if the work required specials skills. Since 1982, the trend has been to allow temporary work in more situations but within the basic framework of the 1974 Employment Protection Act (i.e. that an employment contract is until further notice unless otherwise stipulated). In 1992, temporary contracts for probationary work and to accommodate temporary increases in labour demand were allowed for a maximum period of six 5

7 months. In 1997, the possibility for employers to hire for a fixed duration without having to specify a particular reason was introduced. The employer could use a maximum of five such contracts, and a particular individual could not be employed on a fixed-term contract for more than twelve months during a three-year period. In 1997, to take effect on January 1, 2000, it was also prescribed that if a leave replacement was employed for a maximum duration of three years during a five-year period, the fixed-term contract should be automatically transformed into an open-ended contract. Although this paragraph became law in 2000, it was announced in 1997 to give the employers time to adjust and thus probably had an effect on hiring polices before 2000 (Holmlund and Storrie 2002). There are also a large number of special regulations and/or exceptions to the Employment Protection Act that give employers extensive opportunities to use temporary employment contracts. These special regulations apply to areas such as education and research, and in public administration these regulations are often incorporated into the regulations for each public authority. Moreover, collective agreements are still important for regulating temporary work in Sweden. In most countries, as in EU law, the legal framework can best be described as a floor of rights, and collective agreements may add (but not remove) rights to this floor. In Sweden, however, collective agreements may provide important exemptions to statutory law. In practice, the Employment Protection Act of 1974 gave the parties on the labour market (employers and unions) a great deal of leverage to decide when and under what conditions temporary work should be allowed. Thus, collective agreements can both allow temporary contracts in circumstances other than those stipulated by the Employment Protection Act, and make the use of such contracts more restrictive than stipulated by the same Act. However, most commonly, collective agreements in fact provide important restrictions to the use of temporary contracts (SOU 1993:32; Ds 2002: 56: 228). The development and extent of temporary employment in Sweden The Swedish labour force surveys (AKU) have asked the respondent which form of attachment permanent or temporary he or she has had to the labour market since Whereas the share of employees with fixed-term contracts (as a proportion of total employment) decreased slightly from 12 per cent to around 10 per cent between 1987 and 1991, this share increased rather dramatically during the rest of the 1990s and reached 16 per cent in This development is closely related to fluctuations in the business cycle; the dramatic downturn in total employment and the resulting increase in unemployment in the early 1990s was primarily confined to those with open-ended 6

8 contracts, while the number of fixed-term contracts was relatively less affected by the recession. In real numbers, the growth in the proportion of employees with fixed-term contracts between the early 1990s and 2007 corresponds to an increase from around 400,000 to 685,000 individuals. Figure 1 about here Employment contracts with a fixed duration come in many forms in Sweden, the most important ones being replacement of an absent worker, probation employment (where a worker is employed on a trial basis), on-call jobs and project jobs (to perform a special task). These different forms of temporary employment differ with regard to several important employment characteristics (Walette 2004; Wikman 2002). Generally, those employed on a project basis (i.e. to perform a special task) are the most favoured group, with more influence over their working conditions and more opportunities to develop their skills (Aronsson et al. 2002). At the other end, we find those who are employed during a specific period (on-call jobs), a group that generally have less control over their work situation and more limited opportunities for on-the-job training. Unfortunately, we cannot distinguish between different forms of temporary employment in this study. It is nevertheless important to bear in mind that these different types of temporary jobs have evolved differently since the early 1990s (Nelander and Goding 2003). In men, the replacement, project and on-call forms of temporary employment were equally prevalent in 2003 and accounted for slightly more than 20 per cent of all temporary jobs. Of these, on-call jobs in particular increased in the 1990s, whereas both probation and (above all) replacement temporary jobs declined. In women, replacement jobs are the most prevalent form of temporary employment, but its share of all temporary jobs actually decreased in the 1990s (from 58 per cent in 1990 to 39 per cent in 2003). Instead, the proportions of project and on-call temporary jobs doubled and they accounted for 14 and 27 per cent, respectively, of all temporary jobs in The share of employees with fixed-term contracts also varies considerably between different groups on the labour market. Temporary employment is particularly prevalent among women, youths and blue-collar workers (Table 1). Perhaps most striking is the fact that while the proportion with fixed-term contracts was over 50 per cent in employees aged years in 2005, the corresponding figure for employees aged 45 years and above was only 7 per cent. An important reason for this difference between different age groups is that temporary work often involves probation employment, where the worker is employed on a trial basis, which is naturally more common in youths and other newcomers to the labour market. Note that while a higher 7

9 proportion of women had fixed-term contracts both in 1986/87 and in 2005, the increase in the proportion of employees with fixed-term contracts has actually been more pronounced in men. Table 1 about here Temporary work and labour market careers in Sweden: previous research Bearing in mind the important differences between different demographic and socioeconomic groups in the incidence of temporary employment in Sweden, is it possible to say anything about the net effect of different background variables for the likelihood of holding a temporary job? Using Labour Force Survey data for the period , Walette (2004) has found no systematic differences between men and women in the probability of holding a temporary job once worker and job characteristics are controlled for. 2 He has also found systematic differences between age groups, with young workers having a higher probability of holding a temporary job, as well as between native-born workers and foreign-born workers, with the latter having a significantly higher risk of holding a temporary job. Moreover, although temporary jobs are not a phenomenon confined to the lower part of the earnings distribution, unskilled blue-collar workers have a higher probability of holding a temporary job. Analysing cross-sectional Labour Force Survey data from 1995, Persson (1997) has also found that there are no systematic differences between men and women in the probability of holding a temporary job once worker and job characteristics are controlled for. Thus, the fact that more women than men have temporary jobs is explained by the fact that women are over-represented in sectors where these kinds of jobs are more frequent. She has also found that part-time employed, white-collar workers and those working in service sectors and in the public sector have a higher likelihood of being temporarily employed, whereas the likelihood of holding a temporary job decreases with age. Much of the research on temporary employment in Sweden has concerned the question of whether this form of employment is a stepping stone into more secure employment. Using Labour Force Survey data for the middle of the 1990s and a follow-up period of two years, Håkansson (2001) has found that those with temporary contracts have a significantly greater likelihood of being regularly employed two years later than of being unemployed. While temporary employment frequently seems to function as a stepping stone into secure employment for those employed on a probation contract, this form of transition was less prevalent among those with seasonal and oncall employment contracts. The same author has also found important differences between socioeconomic and educational groups, with white-collar workers and those with post-secondary 8

10 education having a significantly higher probability of being permanently employed two years later. Women also had a significantly lower probability of being in secure employment two years later as compared to men, as did those between 25 and 44 years of age. Walette (2004) has also analysed exits from temporary jobs using a relatively short follow-up period (two years). The results are rather discouraging for everyone except prime-age Swedish-born males with a high socioeconomic position. Thus, compared to this group Swedish-born females have a lower probability of exiting to open-ended jobs, as do younger workers, unskilled blue-collar workers and foreign-born workers (controlling for a number of worker and job characteristics). Overall, exit probabilities were also lower in than they were in the period , and regional unemployment rates also reduced exit probabilities. However, Wikman (2002: 24) found that exits from temporary employment are rather similar across age groups, genders, educational groups and socioeconomic groups. Based on this, he has drawn the conclusion that it is the characteristics of the temporary employment itself, rather than the background characteristics of the individual, that are important in explaining exit rates from temporary employment. Korpi and Levin (2001) have analysed a sample of initially unemployed individuals; they found that in comparison to remaining unemployed, taking up temporary employment is associated with a clear reduction in subsequent unemployment risks, an effect that is, however, significantly weaker than the reduction in unemployment risks associated with permanent employment. Thoursie (1997) also analyzed transitions from unemployment and found that males have a higher probability than females of leaving unemployment for open-ended jobs, and a lower probability of leaving unemployment for a temporary job. Temporary jobs and first births: previous research During the post-war period, fertility in Sweden hovered between high and low and it has therefore been aptly described as roller-coaster fertility (Hoem and Hoem 1996; Andersson 2000). The fluctuations in total fertility rates in recent years are closely related to the economic business cycle. From a peak level with a total fertility rate of over two in the early 1990s, fertility decreased sharply in tandem with the deep economic crisis from 1992 and onwards. A closer look at firstbirth fertility for different age groups reveals an interesting pattern (Figure 2): while first-birth propensity for older women was stable up until the early 1980s, for younger women it declined sharply between the late 1960s and the early 1980s. Both age groups then experienced an increase in first-birth propensity of similar magnitude in the 1980s up until the advent of the economic crisis of the 1990s. However, whereas the decrease in first-birth fertility for older women halted in 9

11 1993 and eventually turned into a marked increase by the end of the century, first-birth fertility for younger women continued to fall up until the end of the 1990s (in the early 2000s, first-birth fertility for younger women was relatively stable at around the end-of-1998 level). The relative increase in first-birth fertility for older women (as compared to younger women) means that mean age at first birth has also increased. This trend started already in the 1960s and was temporarily halted in the late 1980s and early 1990s, but then regained momentum as labour market conditions worsened. Figure 2 about here As already discussed, the rise in unemployment in the 1990s was largely concentrated to those with open-ended contracts, as well as to youths, and one response to the worsening conditions on the labour market was that young people went into higher education to a greater extent than before. Unemployment, higher education and temporary employment are all activities that may be difficult to combine with child-rearing, and indeed it appears as if fertility behaviour in the last decades may be explained to an important extent by an increased inflow into these three activities. A number of studies have also provided evidence for the pro-cyclical pattern of fertility and economic business cycle in later years. Using individual-level register data for all Swedish women born 1950 or later, Hoem (2000) has shown that first-birth rates moved in step with municipal unemployment rates during the years 1985 through This effect was particularly strong for young women, as they may postpone entry into motherhood during slumps in the business cycle. Andersson (2000; 2004) also found strong evidence for a pattern of pro-cyclical fertility in Sweden in recent decades. There is also empirical evidence to suggest that temporary employment contracts have a negative effect on fertility and first-birth propensities in Sweden. Women and men holding temporary contracts have been shown to have a more than 20 per cent lower first-birth risk than those who have permanent jobs (Ds 2001: 57: Persson 2001; SCB 2008). One explanation for this is that women attach great significance to stable employment when deciding to become mothers (Ds 2001: 57; Ungdomsstyrelsen 1998, 2003). It has also been shown that people with temporary employment often feel that their form of employment contract is a problem when, for example, they want to apply for a loan or a telephone subscription, or take out an insurance. To be able to take a loan is often a must in order to get a place to live that is suitable for raising children. The reduction in fertility associated with temporary employment is especially pronounced for older 10

12 women; in women aged 30 and above, the propensity to become a mother is about 35 per cent lower for those on temporary employment (Persson 2001). One explanation for this might be that younger women see more opportunities to get a more permanent position relatively soon, whereas older women might feel more trapped in their current employment situation (Ds 2001: 57). Data and method The data used in this analysis of the effects of temporary employment on labour market and demographic careers were obtained from the Swedish Level of Living Survey (LNU) from 1991 and The LNU contains a random sample of 1 in 1,000 of the Swedish population between 18 and 75 years of age. In the LNU, the respondents were asked questions about their living conditions in several areas, such as social relationships, material living conditions (income and wealth), health, education, working conditions etc. In addition to the interview data, register information was added, mainly in order to calculate household income (for an overview of the LNU data and its use in life-course analysis, see Jonsson and Mills 2001). LNU 2000 also contains information on the respondent s spouse, but we will not make use of this information because of data limitations. 3 The LNU surveys from 1991 and 2000 also contain life-event information on economic activities. Based on retrospective questions, these data begin with the question Have you ever had a job that has lasted at least six months? If the answer was affirmative, the respondent was asked what year and month this (first real ) job started, and then went on to complete a biographic section on economic activity. Amongst other things, these biographic data contain information on sector, size of workplace, occupational classification, working time and type of work contract until the time of the interview. Thus, for every respondent we have monthly data on economic activity. The question about type of contract was, however, only asked in the survey from 2000 and not in that from 1991, and we will therefore only analyse data from the LNU survey from The indicator about type of job contract is based on the following question, which was asked if the respondent was employed: In that job, were you permanently employed or were you there on a time-limited basis, or was the job first time-limited and then permanent? For temporary contracts that were transformed into permanent contracts, we do not know when this happened, only that it happened some time during the spell of employment. For this reason, spells of employment with a temporary contract will generally be defined as only those spells for which the respondent stated that he or she was employed on a time-limited basis during the whole time of employment. 11

13 Additional analyses have been performed where temporary spells of employment that were transformed into permanent jobs are analysed separately or treated as temporary employment; these will be commented upon in the text. Based on these data, it is not possible to distinguish between different forms of temporary employment (i.e. replacement, probation, on-call, project etc.). This is unfortunate, as earlier research has indicated that employment characteristics and also exit rates from temporary employment may differ between these different forms. However, in comparison with other studies that have analysed temporary employment in a Swedish context, the data used in this study allow rather long follow-up periods. The present study thus fills an important gap in the literature (Wikman 2002: 24). All analyses (except the analyses presented in Table 6) have been performed on individuals born in 1950 or later. In this sample, over 50 per cent of individuals experienced at least one episode of temporary employment at the time of the interview in 2000 (Table 2). The average value for months spent in temporary employment is 16, while the median value is 5. This indicates that the distribution of months spent in temporary employment is rather skewed, with relatively few individuals spending a relatively long period in temporary employment and the large majority with experience of temporary work only spending a few months in this form of employment. Table 2 about here Using these data, a number of different transitions and outcomes can be analyzed. Type of first employment is type of employment contract in the individual s first job episode (lasting at least six months). Here, we will estimate a multinomial logit model, where the dependent variable can take on three different values: (i) permanent contract, (ii) time-limited contract, and (iii) time-limited contract that was later transformed into a permanent contract. Entry into first permanent job is estimated with the help of a piecewise constant exponential model, where the risk is defined as transitions from a temporary contract to the first permanent contract. Since the event data on economic activities start with the first job that lasted six months, the analysis is confined to those whose first employment contract was time-limited. Exit from temporary employment is estimated with the help of a piecewise constant exponential model with competing risks, in which risks are defined as exits from atypical employment to (i) typical employment, (ii) repeated atypical employment (defined as at least two subsequent episodes of atypical employment), or (iii) not working (including unemployed, studying etc.). In this model, each individual can contribute with several transitions; an individual may transfer from atypical employment to some other activity, 12

14 then return to atypical employment and then again exit to some other activity. This analysis is confined to individuals up to 45 years of age. The last model in the section on labour market careers will analyse the employment situation ten years after labour market entry using binary logistic regression models, where the dependent variables are (i) risk of being employed (whatever kind of employment) as opposed to not being employed, and (ii) risk of being in temporary employment as opposed to being in permanent employment ten years after entry into the labour market. For many of the respondents in the sample, ten years have not elapsed since their first job. To get reasonably robust estimates, we will therefore analyse the whole sample in these models (i.e. not restrict it to those born in 1950 and later). In the section on demographic careers, the event being analysed is the conception of the first live birth, using piecewise constant exponential models. For the analysis of conceptions, it is important to remember that the respondents in the sample are relatively young: the average age at which women had their first baby was approximately 28 years in 2000, which means that only slightly half of the women in the sample had reached the age at which women on average have their first child, at the time of the interview in These models will contain both time-constant and time-varying independent variables. Cohorts are defined in such a way as to contain approximately equal numbers of respondents: ; and The respondent s education is based on the CASMIN educational categories, distinguishing between elementary education (1ab), basic vocational education (1c), intermediate vocational education (2a), intermediate general education (2bc), lower tertiary education (3a) and higher tertiary education (3b). 5 Social origin is based on the social class of the mother or father, whichever is highest, coded according to the so-called EGP classification (Erikson and Goldthorpe 1992). The coding of the respondent s social class is also based on the EGP scheme and varies according to each episode of employment. Also, working hours (full- or part-time) vary according to each employment episode. In some of the models, we also control for previous experience of temporary employment, unemployment and part-time work, all measured in months. Certain models will also separate between atypical work experience in a low-/mid-/high-qualification level job, where qualification is based on the EGP classification of the job. The analyses of first births will also control for whether or not the respondent s mother worked and the level of education of the parents (primary, secondary or tertiary education). Over-qualification is measured with a simple variable that takes on the value one (otherwise zero) if the respondent has a tertiary education and works in an occupation classified as belonging to EGP classes IIIa-VII. Finally, occupational mobility is defined as upward or downward moves (> 5 units) in occupational 13

15 prestige between subsequent jobs, where occupations are ranked according the Treiman (1977) prestige scale. The analyses will also take two macro-level variables into account: the youth unemployment rate (16-34 years of age) and the regulation of temporary employment contracts. The de-regulation index is a simple measure of year-to-year changes in the regulation of fixed contracts, based on the verbal account of the regulation of such contracts given earlier. 6 This index varies between 0 and 3, where higher values denote a more strict regulation of fixed-term contracts. The modelling strategy in the event history analysis is to use a piecewise constant exponential model in which the time axis is divided into periods and the transition rate can vary across periods but is assumed to be constant within them. To take the macro-level variables into account, the various episodes were split into sub-episodes every year and the macro-level variables were updated in each episode for all individuals every year. Temporary contracts and labour market careers: results When analysing the effect of temporary jobs on labour market careers, an appropriate starting point is to determine whether the probability that an individual s first job is time-limited or not differs between groups (Table 3). To determine this, multinomial logit models were estimated, where the dependent variable could take on three values according to whether the first job was permanent, temporary, or temporary but later transformed into a permanent job. Table 3 about here The first thing to note is that there is a significantly higher likelihood that a woman s first job will be temporary, whereas there is no difference between men and women in the likelihood that their first job will be a temporary one that is later transformed into an open-ended contract. As discussed earlier, previous research has not been clear on whether the risk of holding a temporary job differs between men and women. Walette (2004), for example, has shown that the likelihood of men and women holding a temporary job differs between different types of temporary jobs, where women have a higher risk of holding replacement temporary jobs but a lower risk of holding other forms of temporary jobs. However, his analysis concerned the probability of holding temporary jobs in general, not the probability of the first job being temporary. Additional analysis (not shown) indicates that the higher probability that a woman s first job is temporary is not affected by restricting the sample in terms of cohort (borne before/after 1970) or in terms of entry- 14

16 date into the labour market (before/after 1990). In the youngest cohort (born 1973 or after), the men especially have a significantly higher likelihood of being employed on a fixed-duration contract at their first job. The educational background, as well as the social origin and country of birth, of the respondent seem in principle to be unrelated to the likelihood that the respondent s first job is temporary. There is also a consistent effect of the regulation of temporary contracts: the more regulated temporary contracts are, the lower is the likelihood that the respondent s first job will be a temporary one. There is no effect of the unemployment rate on the likelihood that an individuals first job is a temporary one. However, in models where unemployment rate is included as the only macro-level variable, the effect of this variable is significant (and positive) for both men and women. Thus, gender, age, the regulation of temporary contracts and possibly also aggregate labour market conditions appear to be important influences regarding whether people s first job will be temporary or permanent. Now, for respondents whose first job was a temporary one, what will affect their transition into a permanent job? This question is intimately linked to the question of whether temporary jobs are a stepping stone into a more secure position on the labour market. For the analysis of transitions into a permanent job, we will use event-history analysis as described earlier (Table 4). Looking firstly at the time-invariant covariates, the first thing to note is that a woman will have a significantly lower probability of entering a permanent job if her first job was temporary. Thus, there is not only a higher likelihood that a woman s first job will be temporary; there is also a lower likelihood that women (as compared to men) will leave this first temporary employment for a permanent job. However, the opposite pattern seems to be the case for the youngest cohort (born 1973 or later): while there is a higher likelihood that their first employment will be temporary, there is also a higher likelihood (as compared to the oldest cohort, born ) that they will leave it for permanent employment. Education also seems to be important for explaining the transition into a secure job for those whose first job was temporary: in the combined sample, those with intermediate vocational education (2a), intermediate general education (2bc) and higher tertiary education (3a) have a higher likelihood of transferring to a permanent job (as compared to the reference category, elementary education). These effects differ somewhat between men and women: for men, the only significant effects are for those with intermediate vocational education and possibly also for those with higher tertiary education, while for women there is a significant effect for those with lower and higher tertiary education and intermediate general education (but no significant effect for those with intermediate vocational education). The only effects of social origin found to be significant are the positive effect of lower- and higher-grade 15

17 routine non-manual workers among male respondents and the negative effect for skilled manual workers among female respondents. Table 4 about here Moving to the time-varying covariates, there is a significant positive effect on the risk of transitions into a permanent job for time spent in education for men as well as in the total sample, but not for women. There is also a positive and significant effect of time spent in jobs with a temporary contract on transitions into permanent jobs for women, as well as in the sample as a whole. When we distinguish the experience of temporary jobs according to the different EGP categories, this (positive) effect seems to be present at the two extremes of the social class spectrum, i.e. in unskilled blue-collar workers as well as higher and lower service classes (I and II). There is no effect of sector (private vs. public), nor of country of birth. Finally, whereas unemployment rates does not seem to influence the transition to permanent jobs among those whose first job was temporary, the regulation of temporary contracts appears to matter for women: the more regulated fixed-duration contracts are, the higher is the likelihood that women will exit to open-ended contracts. In this model, the effect of unemployment does not become significant if we exclude the variable measuring the regulation of fixed-term contracts from the model. People may leave temporary jobs not only for permanent jobs, but also for other positions on and off the labour market. Table 5 presents the results from an event-history model with competing risks, by means of which we analyse three different forms of exits from any employment spell (i.e. no only exits from the respondent s first job) with a lime-limited contract: (i) to permanent employment, (ii) to repeated temporary employment (defined as at least two subsequent episodes with fixed-term contracts), and (iii) to non-work. Table 5 about here Once again, we can note that women have a lower likelihood than men of exiting temporary employment and entering permanent employment. This result agree with the findings of both Håkansson (2001) and Walette (2004), who also show that the lower probability of women entering a permanent job is not affected by the type of temporary job. There is an almost linear positive and significant effect of birth cohort on the likelihood of exiting temporary employment for repeated temporary employment, an effect that is however only present in the combined sample 16

18 and in women. Thus, it is above all young women who run the risk of ending up on the temporary job carousel. There is also an effect of birth cohort on exits to positions outside the labour force, an effect that differs somewhat between men and women: for women, it is the youngest birth cohort that has a significantly higher risk of exiting to an out-of-work situation, while the corresponding risk is highest in men born between 1968 and 72. Respondents born outside of Sweden seem to have a somewhat lower probability of exiting to a permanent job as compared to respondents born in Sweden. Perhaps somewhat surprisingly, sector (private vs. public) does not seem to matter for women, while men working in the public sector with a temporary contract have a significantly higher risk of repeatedly exiting and entering temporary employment, but a significantly lower risk of exiting to a position outside the labour force. There are very few significant effects of educational attainment and social origin on transitions to both permanent employment and repeated temporary employment. However, in the combined sample, all CASMIN categories have a higher probability of ending up outside the labour force compared to the reference category, elementary education. The positive effect of education on transitions to non-work is mainly accounted for by transitions to studying: additional analyses (not shown) indicate that there is a significant positive effect (in relation to the reference category, elementary education) for all CASMIN categories on transitions from temporary employment to studying. As a consequence, if we define non-work as everything except studying (as well as permanent employment and repeated temporary work), the only positive effects of educational attainment that remain are for basic vocational education and intermediate vocational education for men and a negative effect for women with lower tertiary education. For both men and women, further analysis indicates that the higher probability of these forms of transition from temporary work is accounted for by transitions to unemployment. The only significant effect of the class position of a temporary job (i.e. the time-varying covariate social class) on the probability of moving from temporary employment to jobs with open-ended contracts is the lower probability of unskilled manual workers in the total sample. Whereas Walette (2004) has shown that the effect of socio-economic class on transitions out of temporary work is rather dependent on which form of temporary employment is exited and that the effect of social class on such exits is limited if temporary work is treated as a homogeneous category, Håkansson (2001) has shown that there is a higher probability that higher-grade professionals will enter permanent employment from temporary employment. However, there are a range of significant positive effects of the covariate EGP class on transitions from temporary employment 17

19 to non-work. Also, here a more refined measure of the category non-work is illuminating for interpretation of these effects: the significant positive effect of EGP classes VII (non-skilled manual workers) for men, and IIIb (lower-grade routine non-manuals) and VII for women are accounted for by transitions to studying. Consequently, if we exclude studying from the definition of non-work, there are no significant effects of the different EGP categories on transitions from temporary work to non-work. The only significant effect of accumulated experience of temporary work is the effect on transitions to a position outside the labour force for women. A more detailed definition of nonwork indicates that this positive effect for women is accounted for by transitions into studying, while the accumulated experience of temporary work does not increase the risk of transitions into unemployment for women. In the combined sample, there is also a negative effect of accumulated part-time employment experience on the probability of entering all three destinations analysed here. It is worth noticing that the negative effect of part-time work seems to be especially pronounced for transitions into repeated temporary work for women. Finally, there is a significant negative effect of aggregate unemployment levels on the probability of entering permanent employment, an effect that is somewhat stronger for men than for women. This suggests that the role of temporary employment as a stepping stone into permanent employment is greater when labour market conditions are good. Higher unemployment rates also seem to have a negative impact on the probability of entering positions outside the labour force for women, but not for men. However, this effect is something of an artefact since unemployment is included in the destination-state non-work; if unemployment is not included in the definition of non-work, this negative effect of aggregate employment on transitions to non-work disappears. There is also a negative effect of the regulation of temporary contracts on transitions into permanent employment in the total sample, indicating that the de-regulation of fixed-term contracts will increase the likelihood of respondents entering permanent employment after being temporarily employed. The analyses presented in Table 6 are an attempt to assess the long-term consequences of temporary employment on individuals labour market careers. The first three columns show the results of a binary logistic regression model where the dependent variable measures whether the respondent was employed 10 years after his or her entry into the labour market (i.e. 10 years after he/she first had a job that lasted at least six months) as opposed to not being employed (which 18