Better Off Jobless? Scarring Effects of Contingent Employment in Japan

Abstract

Previous research fails to address whether contingent employment benefits individuals’ careers more than the alternative they often face: being without a job. Using work history data from Japan, this study shows that accepting a contingent job delays individuals’ transition to standard employment more than remaining jobless. Moreover, having a contingent job, rather than having no job, leads Japanese men to have lower occupational status after they transition back to standard employment. I argue that in a highly segmented labor market like Japan’s, the strict separation of labor pools for standard and contingent jobs makes being labeled as a contingent worker particularly detrimental. Meanwhile, the legacy of Japan’s welfare corporatism alleviates the stigma of unemployment, making individuals better off jobless than having a contingent job. This research thus demonstrates the importance of labor-market contexts in shaping the scarring effects of contingent work arrangements.

Increases in global economic competition during the past few decades have exerted great pressure on firms in advanced industrial countries. This pressure, accompanied by loosening employment-security regulations, has led employers in those countries to increase their use of contingent employment practices (DiPrete et al. 2006; Kalleberg 2000; Shire 2002). Unlike standard wage employment, which assumes continuous and potentially indefinite employment relations, contingent work arrangements involve fixed-term employment contracts and, in some cases, less than full-time or somewhat irregular working hours. By enhancing personnel flexibility, contingent-employment contracts allow firms to better endure market uncertainty and achieve higher profits (Houseman and Osawa 2003; Kalleberg 2003). Thus it is not surprising that in European countries such as France and Spain, the proportion of the labor force on fixed-term contracts has doubled since the late 1980s (DiPrete et al. 2006; Dolado et al. 2002). Similarly, in Japan, nearly all job growth since the early 1990s is attributable to increases in atypical and temporary jobs (Yu 2009).

Rapid increases in the contingent workforce across the industrialized world have enhanced scholars’ interest in atypical, fixed-term employment arrangements. Research on contingent employment generally focuses on the differences between workers in this type of employment and those with standard jobs. Workers on fixed-term contracts are concentrated in lower-status occupations, receive less pay and fewer fringe benefits, and face a greater risk of unemployment compared to standard employees with similar skills and characteristics (Gash and McGinnity 2007; Giesecke 2009; Giesecke and Groß 2003; Kalleberg et al. 2000; McGovern et al. 2004). As a result, contingent employment is also associated with relatively low job satisfaction and a sense of job insecurity (Booth et al. 2002; DiPrete et al. 2006).

Despite the disadvantages of contingent employment, one might imagine that for anyone seeking wage employment, a contingent job would be better than having no job at all. In fact, a few researchers view the use of contingent contracts as an economy’s way to prevent the worse alternative, a heightened unemployment rate (DiPrete et al. 2006; Polavieja 2006). From individuals’ point of view, not only does being jobless lead to a complete loss of income, but it may also harm job prospects over the long run by signaling their low desirability in labor markets (Gangl 2004; Gibbons and Katz 1991). Furthermore, human capital theorists would expect that by staying in the labor force, contingent workers experience less depreciation of their marketable skills than those without jobs (Becker 1964). In addition, individuals in contingent employment should be more able than the unemployed to afford the time needed to find a good-quality job, because they do not need to rush in search of an income. Therefore, those with contingent jobs might be more likely than their jobless counterparts to achieve a higher occupational status at a later point in their lives.

Although the economic benefit of contingent employment over unemployment is plausible, the existence of this benefit is likely to be conditioned by labor market structures. Specifically, in highly segmented economies where employers sharply distinguish between the labor pool for organizational insiders and that for outsiders, contingent employment may have an effect of labeling individuals as belonging to the outsider pool. Hence, beyond contingent jobs’ inferior compensations and working conditions, such jobs may also stigmatize individuals, thereby affecting their career prospects. In comparison, whether those without jobs are associated with the insider or outsider labor pool is more ambiguous. This ambiguity may make it easier for the jobless to be accepted in the standard employment sector than their counterparts in the contingent employment sector.

Because existing research generally focuses on comparing standard wage employment with contingent employment (e.g., Giesecke 2009; Kalleberg et al. 2000; McGovern et al. 2004), there is virtually no evidence on how beneficial or detrimental settling for a contingent job temporarily, as opposed to remaining jobless, may be. We also do not know how labor market institutions affect the scarring effects of contingent employment on individuals. In this study I use the case of Japan to compare the long-term job prospects of contingent workers and jobless individuals. Specifically, I examine the likelihood that these two groups will obtain standard jobs, as well as the quality of the standard jobs they obtain. Japan is particularly interesting because its standard employment sector is highly rigid, with various employment practices institutionalized to ensure job protection for regular full-time workers (Houseman and Osawa 2003; Moriguchi and Ono 2005). This job protection, on the one hand, has led management to demand long working hours and high worker devotion in return (Yu 2009). On the other hand, because job protection ultimately obstructs management’s ability to penalize those who fail to meet its demands, Japanese employers are likely to actively avoid offering standard jobs to those who might not honor their end of the bargain. The preemptive approach taken in screening standard employees contributes to a strict separation of labor pools for standard and contingent jobs, with all those having questionable traits in the latter pool. At the same time, Japan’s well-known welfare corporatism has held employers responsible for supplying jobs and minimizing unemployment (Lincoln and Kalleberg 1996; Schoppa 2006). Although economic fluctuations in recent years have made it more difficult for management to maintain the welfare corporatist model, the model’s legacy may lead many to view unemployment more as a result of macroeconomic crises than individuals’ lack of ability. The alleviated blame on the jobless, along with the little mixing of the labor supply for standard and contingent jobs, makes Japan a setting in which being labeled as a contingent worker may be more harmful in the long run than being labeled as unemployed. Japan thus provides a useful case for understanding how labor market institutions shape the effects of contingent employment on individuals’ long-term occupational attainment.

Background

Much research on advanced economies notes a departure from so-called standard or regular employment arrangements over the past two decades (Dolado et al. 2002; Houseman and Osawa 2003; Kalleberg et al. 2000; Polavieja 2006). Instead of offering jobs to be routinely performed for an indefinite term, employers increasingly incorporate short-term contract workers, contract-based part-time employees, seasonal or day laborers, and workers dispatched from temporary help agencies. Studies of contingent employment, which researchers also refer to as fixed-term, atypical and nonstandard employment (Gash and McGinnity 2007; Giesecke 2009; Houseman and Osawa 2003), are primarily concerned with its effects on workers. Research repeatedly shows that contingent employment practices greatly disadvantage incumbents. Not only do contingent employees suffer from insecurity and stigma associated with their job status (Boyce et al. 2007; Giesecke 2009; Kalleberg 2000), but they also experience wage penalties, as well as receive limited training and few fringe benefits (Kalleberg et al. 2000; McGovern et al. 2004). Moreover, contingent workers are at a higher risk of having another contingent job and becoming unemployed than those with regular full-time jobs (Booth et al. 2002; Gash and McGinnity 2007; Giesecke and Groß 2003).

Given existing research evidence, few would doubt that contingent jobs are inferior to standard or regular jobs. When assessing the impact of contingent employment, however, making comparisons with standard employment is misleading because many of those accepting a contingent job do not have a regular full-time job as an alternative. As prior studies indicate, individuals with characteristics deemed unfavorable by the standard employment sector, as well as those recently unemployed, are the ones likely to take contingent jobs (Farber 1999; Yu 2002). The real alternative to contingent employment, therefore, is unemployment. Thus, for most job seekers the most meaningful question to ask is not whether a standard job is better than a contingent one, but whether a contingent job is better than no job.

Surprisingly, labor market researchers rarely compare the consequences of contingent employment with those associated with a lack of employment. Nevertheless, a handful of studies suggest that a contingent job might be better than no job. Specifically, research on West Germany finds that individuals accepting fixed-term jobs more frequently transition to standard employment than unemployment at a later point of their lives (Gash 2008). Furthermore, individuals who obtain regular jobs by way of temporary employment in the Netherlands, as well as women who do so in Britain, are shown to have wages equivalent to or higher than those who move directly from unemployment to regular employment (Booth et al. 2002; De Graaf-Zijl et al. 2011). Taken together, these findings imply that a temporary job can be a stepping stone to regular work, hence a better option than remaining jobless. Research evidence on the comparative advantages of contingent employment vs. joblessness, however, is indirect and limited. More important, whether a contingent job is better than no job may depend on the labor market context. In the following I discuss the Japanese context and how labor market institutions may mediate the impact of contingent employment on individuals’ careers.

Contingent Employment and Its Scarring Effects in Japan

The rise of contingent employment has closely corresponded to Japan’s economic shifts over the past several decades. From the mid-1970s to late 1980s, as economic growth slowed, Japanese management rapidly increased its use of so-called part-time jobs. During this period, most workers holding these jobs were married women returning to the labor market and men approaching retirement (Yu 2002). Starting from the early 1990s, Japan experienced an unprecedentedly long and severe economic recession, resulting in a greater need to reduce business costs and enhance personnel flexibility. This need led to a drastic expansion of temporary workers, in addition to the continued growth of part-time jobs. Temporary work arrangements became even more prevalent beginning in the late 1990s, after Japan’s deregulations enabled management to use temporary help agencies to fill a wider array of jobs on a more regular basis (Shire 2002; Yu 2009). As temporary jobs increased, middle-aged married women and older men were no longer the only demographic groups involved in contingent employment in Japan. Men of prime working age, as well as single men and women recently graduated from school, all became likely candidates for temporary jobs (Genda and Kurosawa 2001; Yu 2010).

Figure 1 illustrates changes in the share of contingent workers among Japanese employees over time, along with trends in fixed-term employment in several major European countries. While Japan’s proportion of female employees with contingent contracts has been strikingly high, among men the percentage of contingent workers has more than doubled since the mid-1990s. Here contingent employment for Japan includes both part-time and temporary work. The difference between part-time and temporary jobs is particularly small in that country, because “part-time” refers to employment status, rather than the number of work hours. Hence, a part-time job in one Japanese company might be labeled as a temporary job in another. Regardless of how jobs are labeled, workers with both part-time and temporary statuses have fixed-term contracts (Yu 2002).

Figure 1.

Trends of Contingent Employment in Japan and the EU15 Countries

Source: Japan: Annual Report on the Labour Force Survey, prepared by Statistics Bureau, Ministry of Internal Affairs and Communications, Japan, various years. EU15: OECD.Stat Extracts, http://stats.oecd.org/index.aspx.

Note: Whereas part-time workers who are not on fixed-term contracts are excluded from the percentages for the 15 European countries, all part-time workers are included for Japan, because virtually all part-time employment is on a fixed-term basis in that country.

EU15 refers to the 15 countries in the European Union prior to the accession of 10 candidate countries in 2004, including Austria, Belgium, Denmark, Finland, France, Germany, Greece, Ireland, Italy, Luxembourg, Netherlands, Portugal, Spain, Sweden, United Kingdom.

Contingent jobs’ limited durations sharply differentiate them from standard jobs in Japan. Although Japan has only modest legal restrictions limiting employee dismissal, firms face strong expectations to maintain long-term employment for regular full-time employees, to the extent that this expectation has frequently shaped judges’ rulings in related lawsuits (Moriguchi and Ono 2005). The norm of long-term employment has compelled employers to put great effort into ensuring job security for standard workers, even after Japan’s severe recession in the 1990s (Yu 2010).

Part of the reason Japanese firms are expected to maintain the permanent employment practice is that it constitutes a key component of Japan’s social-protection system (Schoppa 2006). Instead of a welfare state, Japan has followed a welfare corporatist model, in which firms provide long-term job security, seniority-based wages and promotions, and extensive benefits to ensure the wellbeing of workers and their families. This system excludes contingent workers, however, as their positions render them as “organizational outsiders.” Thus, not only are contingent workers always on limited-term contracts, but they also rarely receive fringe benefits, training, promotions or raises (Houseman and Osawa 2003). In addition, their hourly wages are only half to two-thirds of regular full-time employees’ (Hanami 2004).

At the same time, as organizational outsiders, contingent employees are expected to exhibit different commitment behaviors from those of regular full-time workers. Anecdotes from previous research show that contingent workers are exempt from doing overtime work or participating in after-hour company activities, which generally are part of the working lives of those with standard jobs (Yu 2002, 2009). Although the outsider status is ultimately responsible for contingent employees’ seemingly low commitment, they may nevertheless be stigmatized as inadequate workers as a result. Research on U.S. temporary workers indicates that these workers are often stereotyped as having a weak work ethic and inferior abilities (Boyce et al. 2007). The label of contingent workers may be even more harmful in Japan because the stereotype about these workers’ low commitment, or at least their inability to prioritize job demands over personal obligations, contradicts the fundamental requirements for standard employees in that country. With employers maintaining practices like lifetime employment and seniority-based wages, the implicit employment contract in Japan has been that standard employees must exhibit an extraordinary level of work commitment in return (Lincoln and Kalleberg 1996). These same employment practices have nevertheless made it difficult for employers to sanction those unwilling or unable to honor the implicit contract. Consequently, when selecting standard employees, Japanese management is likely to focus on hints about individuals’ likelihood of honoring this contract. Graduating from a prestigious university or a high school that has an informal linkage with the company (Genda and Kurosawa 2001), for example, signals that a candidate is likely to sacrifice personal needs for the company despite management’s inability to enforce this behavior, whereas being a mother signals the opposite (Yu 2002). Similarly, a history of contingent employment suggests that a candidate prefers more confined job demands or has a competing devotion, if not a poor work ethic. Either way, this history is likely to dissuade Japanese management from offering a standard job. Thus, unlike in some European countries, contingent employment might not serve as a stepping stone to permanent work in Japan.

If contingent workers in Japan are indeed labeled as unsuitable for serious or demanding jobs, fixed-term employment should also harm the quality of jobs individuals obtain subsequently. Those with a history of contingent work may face penalties in occupational prestige and wages even if they can shift to standard employment. The fact that accepting a contingent job may scar one’s career over the long run makes it similar to another critical event in working life, an interruption of employment. Not only does being jobless generally depreciate a person’s human capital, but it may also signal that this person is relatively undesirable in the labor market (Gibbons and Katz 1991). Moreover, the absence of income during a jobless period forces one to be less choosy about jobs, leading to a suboptimal job match. For these reasons, an interruption of employment is associated with decreases in wages, occupational status and job stability both immediately and years after reemployment (DiPrete and McManus 2000; Gangl 2004, 2006).

Because both contingent employment and unemployment may scar individuals’ careers, the question of whether having a contingent job is better than no job is essentially about which option is more detrimental to subsequent job prospects. The answer to this question is likely to depend on the labor market context. Just as an overall emphasis on worker commitment may affect the stigma associated with contingent jobs, a country’s welfare policies and employment practices are likely to influence the extent of harm caused by work interruption. Researchers find, for example, that displaced workers in countries with generous unemployment benefits fare better after reemployment, because they are less likely to accept low-quality jobs out of desperation (DiPrete 2002; Gangl 2004). Being jobless is also associated with lower wage penalties in the future when the labor market enables workers to move across occupations easily (Gangl 2006).

Findings from previous research suggest that employment-interruption penalties are likely to be considerable in Japan. First, Japan’s unemployment benefits are far from generous. The wage replacement ratio of Japan’s unemployment insurance has been below the average, and the maximum duration of this insurance has been one of the shortest among members of the Organization for Economic Co-operation and Development (Nickell et al. 2005). Second, Japanese management has placed great emphasis on firm-specific training, making it difficult for workers to transfer their human capital from one firm to another (Moriguchi and Ono 2005; Yu 2009). Such low skill transferability means that individuals experiencing work interruptions are extremely unlikely to achieve their pre-interruption wage levels after reemployment.

The potentially extensive penalties associated with being jobless in Japan, however, do not necessarily indicate that having a contingent job is better. To be specific, the lack of skill transferability across Japanese firms also implies that taking a contingent job to shorten a jobless spell will not lessen human capital depreciation, as most of one’s human capital is lost at the moment one exits a firm. At the same time, the stigma of contingent employment might actually be greater than that of being jobless. Following the welfare-corporatist model, Japanese firms have been expected to take responsibility for the employment of all pre-retirement citizens except for women of childbearing age (Schoppa 2006). Until the severe economic recession of the 1990s, it was rare for working-age men, as well as single women, to be without a job. In this context, unemployment is likely to be seen as firms’ failure to maintain welfare corporatism, rather than any given individual’s fault. This view regarding unemployment, according to Gibbons and Katz (1991), helps lessen its scarring effects. In contrast, not only is the act of accepting a contingent job likely to be considered as a personal choice, but the stereotype about contingent employees – that they are persons who give priority to non-work activities – merely concerns their individual traits. For this reason, while being jobless is detrimental to long-term career prospects in Japan, contingent employment may scar individuals’ careers even more.

Because Japan’s welfare corporatism has assumed sharply different work and family roles between men and women (Yu 2009), the relative harm of contingent employment to unemployment may also vary by gender. Specifically, Japan’s welfare model has treated the male-breadwinner family as the norm, hence taking for granted men’s willingness to work and ability to prioritize job demands over family obligations. Conversely, women’s primary roles have been assumed to be as caregivers for their families, rather than as workers. Japanese management may therefore question the likelihood that women will commit to their jobs at the level expected for standard employees, regardless of their history of contingent employment. For the same reason, employers are more likely to consider women’s employment interruptions as resulting from their family or other personal conditions, rather than the decline in welfare corporatism during economic downturns. Thus, the presumed family roles for Japanese women, on the one hand, decrease the importance of contingent employment as a signal for worker commitment and, on the other hand, increase the likelihood that employers personalize the incidence of being jobless for this group. As a result, if contingent employment indeed scars individuals’ careers more than employment interruption in Japan, the tendency will be more pronounced among men than among women.

In summary, the Japanese labor market features a dualist system that strongly emphasizes worker commitment – with the provision of long-term job security in return – in the standard employment sector, while expecting a low commitment and offering minimal job protection for contingent employees. The label of contingent workers, often suggesting a lack of devotion, can be expected to be particularly damaging under this system, as management would be highly reluctant to offer standard jobs to workers regarded as less than fully committed. By contrast, although unemployment is generally thought to stigmatize workers, the legacy of Japan’s welfare corporatism is conducive to assigning more blame to macroeconomic shifts than to the jobless. Thus, having a contingent job is likely to have greater scarring effects than having no job in Japan. Because of women’s family roles, however, Japanese employers are less likely to differentiate them based on work histories. A history of contingent employment may therefore have greater significance for men’s careers than women’s.

Data and Methods

This study uses data from the Social Stratification and Social Mobility Survey (SSM) conducted in 2005, which contains a multistage probability sample of 5,765 Japanese men and women ages 20 to 69 years old. Respondents were asked to provide information about each job they had ever held, including the occupation, employment status, starting and ending years, and their reasons for exiting that job. The survey also recorded the years in which respondents completed a given level of schooling, became married and had children.

To analyze the long-term consequences of contingent employment, I first examine the effect of contingent employment on the rate of obtaining a regular full-time job. Throughout individuals’ lives, they might face the possibility of entering regular employment at multiple times. Figure 2 illustrates different pathways to a standard, rather than any other, employment status. Except for those who start their careers with and remain in regular jobs, the route to standard employment inevitably involves experiences with another employment status, be it contingent employment, self-employment or nonemployment. I pay special attention to how holding a contingent job, in comparison with being jobless, contributes to the event of switching to a regular full-time job. In the analysis, I separate the first entry from reentry into standard employment because the dynamics might be different. Having prior experience with regular full-time jobs, for example, might alleviate the stigma of contingent employment, thus accelerating the rate of obtaining a regular job. For first entry into standard employment, I also fit separate models for the entry that occurred immediately after school and the one that occurred later, given that the experience of contingent employment is relevant to only the latter scenario.

Figure 2.

Pathways to Standard Employment

Note: For simplicity of presentation, this figure does not include the scenario of returning to school after an individual’s first job. The statistical analysis, however, does consider the possibility of returning to school.

Here “other employment status” includes nonemployment, contingent employment and self- and family enterprise employment.

I should note that nonemployment refers to both voluntary and involuntary jobless states in this study. Because the SSM does not include information about whether respondents were seeking jobs within a given year, I am unable to distinguish the unemployed from those voluntarily out of the labor force. This lack of distinction is unlikely to cause a major problem because, in reality, the boundary between unemployment and nonparticipation in the labor force is often blurry. With sufficient incentives, nonparticipants can be lured into the labor market (Murphy and Topel 1997); thus many nonparticipants actually exhibit behaviors similar to those of the unemployed (Jones and Riddell 2006). Moreover, although existing research is generally about unemployment’s scarring effects (e.g., Gangl 2004), a voluntary employment interruption is likely to have the same scarring effects. One’s human capital would depreciate with the jobless duration regardless of whether one intends to be without a job. Besides, a potential employer might not be able tell whether one was actively seeking employment during a jobless period, thus treating all those without jobs in the same manner.

I adopt an event history approach to investigate how contingent employment affects the rate of obtaining standard jobs because it is sensitive to changes in individuals’ life stages and employment experiences over time (Yamaguchi 1991). Rather than individuals, the unit of analysis for event history models is the person-duration of exposure to the outcome of interest – obtaining a regular full-time job in this case. Because the SSM recorded respondents’ life events by year, person-year is used as the unit of analysis. Based on respondents’ life and work histories, I transformed the SSM data into person-year observations with time-varying individual and job characteristics. For the analysis of first entry into standard employment, I selected the year when respondents left school through the year the entry occurred, if they did not obtain regular full-time jobs immediately after leaving school. I define “leaving school” based on the occurrence of the first time gap in respondents’ educational histories. Respondents are assumed to be subject to the possibility of full-time employment as soon as they stop attending school in a continuous fashion. For the analysis of reentry into regular employment, I selected all the years following respondents’ exits from their regular full-time jobs until either the year of reentry or the survey year, whichever occurred earlier. Because a respondent might exit and reenter regular jobs more than once, as indicated in Figure 2, the sample for the reentry analysis consists of multiple nonregular employment spells from some respondents.

I further selected the nonregular employment spells that resulted from an involuntary job exit for an additional analysis. The SSM contains no information regarding when respondents were looking for standard jobs or why they accepted a given job. Hence, obtaining a regular full-time job at a slower rate might not always indicate obstacles faced in the labor market; it might result instead from an individual’s preference for other types of jobs or being jobless. Among Japanese men, it is nevertheless reasonable to assume a general preference for regular jobs over contingent jobs or nonemployment. Because the social norm prescribes them as breadwinners (Yu 2009), prime-age Japanese men are unlikely to be satisfied with a low-paying contingent job or a jobless state. Still, to be certain that individual preferences play a minimal role, in the additional analysis I examine only regular employment reentry among those were forced to leave this type of employment in the first place. Such respondents were least likely to stay in the contingent employment sector or remain jobless as a result of preference.

The rate of entering or reentering regular employment is not the only indicator of the long-term consequences of taking a contingent job instead of remaining jobless. Therefore, in the second part of the statistical analysis, I focus on how contingent employment experience shapes the quality of individuals’ jobs. Because this part of the analysis emphasizes workers’ job quality, rather than their rates of transitioning into a certain state, the individual constitutes the unit of the analysis. Specifically, I examine the socioeconomic status of the jobs obtained by those who had just had a period of nonregular employment. Similar to the analysis of the rate of transitioning to standard jobs, I separate respondents’ first standard jobs from the ones they obtained upon returning to the standard employment sector. Data from the years in which respondents entered or reentered standard employment are selected for this part of the analysis.

Finally, I analyze the effect of contingent employment experience on earnings, another important indicator of job quality. Because the SSM collected income information only for the year immediately prior to the survey time, the analysis is limited to those with earnings during that period. To maximize the sample size, I also include respondents who reported not working at the survey time, but had a job and income during the previous year. I nevertheless refer to all jobs observed in the earnings analysis as current jobs for simplicity.

Variables and Measurement

As discussed, the statistical analysis focuses on five different outcomes: (1. first entry into standard employment, (2. reentry into standard employment, (3. occupational status of standard jobs obtained for the first time in individuals’ careers, (4. occupational status upon reentering standard employment, and (5. current earnings. I use binary variables to indicate whether the entry or reentry, respectively, occurred within a given person-year. For the third and fourth outcomes, I measure occupational status according to the International Socioeconomic Index (ISEI) proposed by Ganzeboom and Treiman (1996). Finally, regarding current earnings, the SSM asked respondents to report their personal income for the past year. Although this measure includes non-wage income, previous research demonstrates that the measure of total personal income serves as a reasonable proxy for Japanese workers’ earnings (e.g., Sakamoto and Chen 1993). Using the midpoints of the 30 income categories listed in the survey, I convert respondents’ earnings into a continuous variable and then take the logarithm of earnings.

The predictors included in models for the five dependent variables are largely similar. The main predictor of interest in the analysis is individuals’ employment statuses during the period of nonregular employment. For the event history models predicting entry and reentry into standard employment, I create a series of dummies for the employment status of each person-year: nonemployment, contingent employment and self-employment. Contingent employment is defined as employment under a fixed-term contract, regardless of whether it is fulltime or parttime. I include family enterprise employment in the category of self-employment mostly because few Japanese men are family workers. For the cross-sectional models predicting occupational status, instead of using time-varying dummies, I measure the years spent on nonemployment, contingent employment and self-employment during the nonregular employment spell just before obtaining the standard jobs examined. Somewhat similarly, for the earnings analysis, I divide individuals’ work experience before the current job into years spent in standard employment, contingent employment and self-employment to compare how these different types of employment experiences, as opposed to nonemployment, contribute to current earnings.

In addition to employment status, the models contain a series of variables that may affect labor market outcomes, including gender, education, family status, previous experiences with regular full-time jobs and unemployment rate. The models on earnings further control for critical job characteristics, such as job tenure, current employment status, occupational status, firm size, industry and hours of work. Table 1 provides definitions for the control variables used in the analysis.

Table 1.

Description of Control Variables

Variables	Definition	Time-variant
Gender	Male = 1; female = 0
Education	Level of education completed by the year under observation, in 4 categories: (1. middle school or less, (2. high school, (3. junior college, and (4. university and above	V
Additional training received	Having receiving vocational training outside the formal schooling system	V
Years enrolled in school	Number of years spent in school since the beginning of the period under risk for the event of interest	V
Marital status	Married during the year under observation = 1; otherwise = 0	V
Preschool child present	Preschool child present during the year under observation	V
Close to retirement	Age 50 or older = 1; otherwise = 0	V
Birth cohort	Having turned age 18 (1. before the end of Japan’s rapid industrial expansion (1973), (2. during Japan’s steady economic growth (1974–1989), or (3. after Japan’s severe and prolonged economic crisis (1990)
Unemployment rate	Unemployment rate for the year observed	V
Duration of last standard job	Number of years spent on last regular full-time job	V
Status of last standard job	Occupational status (measured by ISEI) of last regular full-time job	V
Experience of standard job recovery	Number of times respondents have switched to standard employment from another status before entering the current nonregular employment spell: (1. none, (2. once, and (3. twice and more	V
Job tenure*	Number of years at current job
Current employment status*	Employment status of current job: (1. standard employment, (2. contingent employment, and (3. self- employment
Firm size*	Number of workers reported in respondents’ firms: (1. 1–29, (2. 30–999, (3. 1,000 and more, (4. government office, and (5. firm size unknown
Industry*	Measured with six dummies indicating: (1. primary; (2. manufacturing, mining, and construction; (3. personal service; (4. professional service; (5. finance and insurance; (6. public service, respectively
Hours of work*^	Number of hours spend on the current job every month; average working hours of the sample assigned for those who reported having irregular working hours
Irregular working hours*^	Having irregular working hours = 1; otherwise = 0

^*Used only in the cross-sectional models predicting earnings.

^{^} Information unavailable for those had a job and earnings earlier in the survey year but were not employed at the time of the interview

Analytical Strategy

I apply discrete-time hazard rate models to analyze rates of entry and reentry into regular employment, given that the SSM recorded respondents’ job histories and life events in a discrete time unit: year. Specifically, the models estimate respondents’ log odds of entering or reentering standard employment during an observed year, conditional on not having done so earlier. For the purpose of estimating the baseline hazard, all discrete-time logit models consist of a set of dummy variables for respondents’ duration of exposure to the event of interest – that is, the duration since respondents’ exits from school or a standard job. I measure duration with indicators for 1 year or less, as well as 2–3, 4–5, 6–10, 11–20 and 20 or more years from the beginning of a nonregular employment spell.

For analyzing occupational status and log earnings, I generally rely on ordinary least squares regression models. Nevertheless, because the analysis of occupational status upon standard employment reentry includes all incidents of such reentry in the sample, a single respondent could have more than one observation. Similarly, some respondents have multiple nonregular employment spells included in the analysis of reentering regular full-time employment. To correct potential bias caused by repeated observations on the same respondents, I model these two outcomes using a generalized estimating equation (GEE) approach (Zeger and Liang 1986). The GEE method adjusts for covariance among clustered observations by estimating within-subject correlations separately from the regression parameters. For models estimated with the GEE method, I also report robust standard errors using the Huber-White variance estimator.

Results

Before turning to multivariate results, Table 2 provides descriptive statistics for Japanese men and women who have had time spells away from regular full-time employment by their timing of experiencing these spells. A considerable proportion of respondents had a nonregular employment spell at some point in their lives: More than one third did after leaving school before entering first standard job, and 42.5 percent of men and 69.6 percent of women did after having some experience of standard employment. Women on average endured more time than men did before entering or reentering standard employment.

Table 2.

Characteristics of Japanese Men and Women with Spells of Nonregular Employment

	Prior to First Standard Job		Since First Standard Job
	Men	Women	Men	Women
Rate of experiencing such spells	35.8	37.2	42.5	69.6
Average spell duration (years)	9.7	16.2	9.4	16.9
% duration in:
Nonemployment	19.9	53.2	18.9	55.4
Contingent employment	11.9	19.7	15.4	23.1
Self-employment	68.2	27.1	65.7	21.6
Demographic characteristics#
Education
Up to middle school	28.9	32.8	28.7	2.6
High school	55.5	5.3	5.5	63.3
Junior college	1.2	9.7	2.2	1.1
University and above	14.4	7.3	18.7	6.0
Additional training received	5.9	13.1	5.9	12.5
Returning to school during the spell	9.0	2.1	3.2	2.0
Married during the spell	19.7	42.8	66.5	83.5
Preschool child present during the spell	19.9	41.2	27.1	67.1
Average age (years)	22.5	25.5	43.9	36.7
Birth year:
≤ 1955	54.8	6.1	7.7	58.9
1956–1971	27.4	2.9	22.5	31.8
1972 and later	17.7	19.0	6.8	9.4
Duration of last standard job^			2.3	8.8
Average status of last standard job (ISEI)^			39.9	41.7
Having previous experience of standard job recovery^			16.9	17.4
Rate of transitioning to standard employment:	77.4	59.2	34.4	29.5
% from nonemployment	7.1	76.3	42.6	37.9
% from contingent employment	16.4	18.2	37.3	22.8
% from self-employment	13.5	5.5	21.8	12.5
Average post-transition occupational status (ISEI)	4.9	4.3	38.2	4.6

Note: Except for ISEI scores, age and spell duration, all values are in percentages.

^*Because an individual can have more than one spell of nonregular employment since departing from the first standard job, most numbers in this column are calculated using person-spell as the unit of analysis. The exceptions are rate of experiencing such spells and percentage of duration in each employment status, for which the units of analysis are person and person-year, respectively. The distributions are similar when using person, instead of person-spell, as the unit.

^# All demographic indicators except for birth year are time-variant. Hence, I present the distributions of education and additional training based on those at the beginning of the spell and average age based on respondents’ age at mid-spell. For returning to school, being married, and having a preschool child, the numbers represent the percentages of respondents ever experiencing these events during the spell.

^{^}The indicators vary by spell but are constant within each person-spell.

Nearly 80 percent of men and about 60 percent of women who did not hold regular full-time jobs after departure from school transitioned to such jobs by the time of the survey. In contrast, only a third of men and less than a third of women returned to standard employment after leaving it. For both men and women, it was more common to transition to standard employment from nonemployment than contingent employment. The relatively few transitions from contingent to standard employment suggest that Japanese management rarely uses contingent jobs as means of screening long-term employees. In fact, my additional analysis indicated that only 2.9 percent of contingent jobs in the sample turned into standard ones within the same firm. In this sense, contingent jobs are unlikely to be stepping stones to regular work in Japan.

Moving to the multivariate analysis, Table 3 presents logit models predicting the attainment of standard jobs for the first time since leaving school. As indicated in Figure 2, only those without regular full-time jobs upon school departure face the question of whether accepting a contingent job will accelerate or decelerate their first entry into standard employment. To help explain the selectivity of this group of individuals, the first panel of Table 3 includes estimates for the log odds of entering standard employment in the same year of leaving school for all respondents. Based on the results, those with a higher educational level and born to cohorts in 1956 or later had a better chance of obtaining a regular full-time job after school departure. Conversely, being married, having a young child, and facing a high unemployment rate reduce one’s likelihood of transitioning directly from school to a standard job.

Table 3.

Logistic Regression Models Predicting First Standard Job Entry

	Standard Job Entry Right After Graduation	Standard Job Entry Rate, If No Entry Right After Graduation
		All		Men		Women
Duration since leaving school (reference 1 year):
2–3 years		−1.208** (.083)	−1.206** (.083)	−1.009** (.122)	−1.003** (.122)	−1.289** (.115)	−1.291** (.115)
4–5 years		−1.669** (.109)	−1.667** (.109)	−1.245** (.153)	−1.235** (.154)	−1.883** (.161)	−1.885** (.161)
6–10 years		−1.771** (.112)	−1.771** (.112)	−1.061** (.156)	−1.051** (.157)	−2.320** (.176)	−2.325** (.176)
11–20 years		−2.078** (.149)	−2.069** (.149)	−1.737** (.239)	−1.710** (.240)	−2.347** (.200)	−2.349** (.201)
21 years and more		−2.112** (.189)	−2.107** (.190)	−1.500** (.310)	−1.473** (.312)	−2.483** (.245)	−2.494** (.245)
Employment status (reference Nonemployment):
Contingent employment		−.489** (.084)		−.675** (.124)		−.564** (.120)
Self-employment		−1.620** (.103)		−2.229** (.145)		−1.410** (.177)
Contingent employment, high status			−.650** (.122)		−.837** (.196)		−.778** (.160)
Contingent employment, low status			−.377** (.103)		−.600** (.143)		−.353* (.152)
Self-employment, high status			−2.174** (.230)		−2.839** (.294)		−1.947** (.388)
Self-employment, low status			−1.464** (.111)		−2.092** (.151)		−1.220** (.196)
Male	−.011 (.062)	.629** (.070)	.608** (.070)
Education (reference: Up to middle school):
High school	1.037** (.079)	.033 (.073)	.056 (.073)	.073 (.108)	.103 (.109)	.038 (.101)	.061 (.101)
Junior college	1.373** (.135)	.244† (.148)	.296* (.149)	−.178 (.432)	−.131 (.434)	.171 (.168)	.238 (.170)
University and above	1.584** (.110)	.668** (.117)	.754** (.121)	.999** (.153)	1.115** (.161)	.208 (.197)	.293 (.200)
Additional training received	−.141 (.105)	.200† (.110)	.208† (.110)	.285 (.202)	.312 (.203)	.088 (.135)	.094 (.135)
Years enrolled in school (since 1st school departure)		.255** (.039)	.255** (.040)	.142** (.046)	.137** (.046)	.359** (.109)	.369** (.109)
Married	−1.015** (.344)	−1.028** (.145)	−1.038** (.145)	−1.014** (.262)	−1.018** (.263)	−.774** (.182)	−.781** (.182)
Preschool child present	−.746* (.308)	−.387* (.152)	−.397** (.152)	.807** (.252)	.792** (.253)	−.894** (.199)	−.902** (.199)
Close to retirement age	—	−1.255** (.335)	−1.251** (.335)	−1.219† (.634)	−1.222† (.634)	−1.150** (.396)	−1.143** (.396)
Birth years (reference ≤ 1955)
1956–1971	.524** (.082)	.283** (.092)	.294** (.092)	−.086 (.129)	−.081 (.130)	.714** (.132)	.725** (.132)
1972 and later	.422** (.141)	.500** (.155)	.510** (.155)	.096 (.224)	.099 (.224)	.922** (.216)	.925** (.216)
Unemployment rate	−.326** (.051)	−.273** (.051)	−.277** (.051)	−.227** (.075)	−.232** (.075)	−.326** (.072)	−.331** (.072)
Constant	−.005 (.109)	−.444** (.109)	−.453** (.109)	.069 (.155)	.041 (.155)	−.274† (.144)	−.286* (.144)
Log-likelihood	−3281.31	−3843.56	−3837.19	−1741.31	−1737.23	−2038.56	−2034.74
N	5192	27425	27425	8956	8956	18469	18469
Number of events	3118	1396	1396	720	720	676	676

Note: Except for the model predicting standard job entry immediately after leaving school, discrete-time event-history techniques are applied to all models presented. Numbers in parentheses are standard errors.

^†p < .10

^*p < .05

^**p < .01 (two-tailed tests).

The remainder of Table 3 contains coefficients from event history models predicting the first entry into standard employment among those who did not obtain regular full-time jobs immediately after leaving school. Starting from the second panel, contingent employment clearly has a negative effect on the rate of entering standard employment, compared to nonemployment. Specifically, holding a contingent job, rather than having no job, reduces one’s odds of obtaining a regular full-time job by nearly 40 percent (exp.[−.489] = .615). The results are similar in the models that examine only men or women. These results are consistent with the argument that accepting a contingent job, as opposed to remaining jobless, increases the difficulty for the Japanese to transition from school to standard employment. Self-employment also has a negative effect, but this effect is likely related to the appeal of self-employment. Previous research shows that self-employment in Japan largely results from personal preferences (Ishida 2004), rather than a lack of wage employment opportunities. Conversely, although some contingent workers may prefer the time flexibility or limited job demands associated with their jobs, existing research indicates that even among Japanese women, a substantial proportion of those with contingent work wanted standard jobs in the first place (Houseman and Osawa 2003; Yu 2002).¹

Because contingent jobs can be rather heterogeneous, ranging from professional contract work to minimum-skilled day labor, it is important to examine whether all contingent workers have lower rates of transitioning to first standard employment than those without jobs. Table 3 includes additional models in which contingent jobs are divided into high- and low-status ones. I consider jobs with ISEI scores lower than the average of the current Japanese labor market as low-status, and otherwise as high-status. As shown in the third panel of Table 3, having a contingent job delays one’s transition into regular full-time employment more than having no job, regardless of the contingent job’s status. Results from the models for solely men or women are virtually the same. The negative effects of low-status contingent jobs on first entry into standard employment are particularly notable, because holders of such jobs would probably prefer regular full-time jobs. In addition, because those with low-status contingent jobs tend to receive low wages, they may not be much more able to afford passing on less-than-ideal standard jobs than the jobless. Therefore, work orientations or job-search tendencies are unlikely to fully explain the negative association between low-status contingent employment and first entry into standard employment. It is likely that a stronger stigma against contingent workers than the jobless also accounts for this association.

Table 4 presents results from models predicting reentry into standard employment. The first two panels are for all individuals who have ever exited their regular full-time jobs. Similar to the results for first entry into standard employment, contingent employment reduces individuals’ odds of reentering standard employment within a given year by 31 percent (exp.[−.369] = .691), compared to nonemployment.² When separating high-status contingent jobs from low-status ones (second panel, Table 4), only low-status contingent employment has a significant negative effect on reentering standard employment. The nonsignificant effect of high-status contingent employment nevertheless suggests that despite their relatively high occupational skills, individuals with high-status contingent jobs are no more likely than those without jobs to be offered a regular full-time job.

Table 4.

GEE Estimates of Discrete-Time Event History Models for Reentering a Standard Job

	Standard Job Reentry		Standard Job Reentry Following an Involuntary Exit
			All		Men		Women
Duration (reference 1 year):
2–3 years	−.360** (.088)	−.362** (.088)	−.286 (.182)	−.287 (.182)	−.381 (.244)	−.381 (.244)	.007 (.298)	.019 (.296)
4–5 years	−.842** (.112)	−.839** (.112)	−.692** (.227)	−.684** (.227)	−.582* (.293)	−.582* (.296)	−.473 (.366)	−.491 (.368)
6–10 years	−1.123** (.107)	−1.120** (.107)	−1.484** (.246)	−1.471** (.248)	−1.373** (.316)	−1.374** (.316)	−1.235** (.422)	−1.217** (.422)
11–20 years	−1.14** (.115)	−1.138** (.115)	−1.441** (.256)	−1.425** (.257)	−1.328** (.354)	−1.332** (.356)	−1.288** (.411)	−1.256** (.408)
21 years and more	−1.632** (.176)	−1.628** (.176)	−2.385** (.547)	−2.367** (.546)	−1.889* (.754)	−1.895* (.750)	−2.651** (.857)	−2.622** (.849)
Employment status (reference Nonemployment):
Contingent employment	−.369** (.082)		−.118 (.170)		−.484* (.246)		.207 (.251)
Self-employed	−1.783** (.114)		−1.636** (.247)		−2.463** (.338)		−.567 (.373)
Contingent employment, high status		−.085 (.106)		.168 (.239)		−.551 (.478)		.629* (.311)
Contingent employment, low status		−.583** (.104)		−.293 (.203)		−.467† (.273)		−.264 (.324)
Self-employment, high status		−2.048** (.176)		−1.632** (.363)		−2.503** (.471)		−.610 (.617)
Self-employment, low status		−1.701** (.129)		−1.677** (.273)		−2.439** (.361)		−.580 (.443)
Duration of last standard job	−.019** (.006)	−.020** (.006)	−.019 (.012)	−.020† (.012)	−.028† (.015)	−.028† (.015)	−.024 (.022)	−.026 (.022)
Status of last standard job	−.004 (.003)	−.006 (.003)	.000 (.006)	−.002 (.007)	−.001 (.008)	.000 (.008)	.004 (.013)	−.001 (.014)
Experience of standard job recovery (reference None)
Once	.070 (.108)	.068 (.108)	.551** (.170)	.546** (.171)	.290 (.228)	.291 (.227)	.885** (.309)	.885** (.311)
Twice and more times	.430 (.294)	.438 (.296)	.948** (.370)	.956* (.374)	.381 (.582)	.383 (.580)	1.535** (.450)	1.611** (.450)
Male	1.212** (.101)	1.256** (.102)	.925** (.184)	.964** (.186)
Education (reference Up to middle school):
High school	.010 (.090)	.000 (.090)	.028 (.174)	.014 (.177)	.118 (.228)	.122 (.232)	−.099 (.285)	−.125 (.290)
Junior college	−.259 (.166)	−.281† (.165)	.085 (.400)	.007 (.411)	−.467 (.542)	−.443 (.552)	.329 (.579)	.272 (.579)
University and above	.153 (.149)	.145 (.151)	.093 (.290)	.080 (.290)	.217 (.342)	.221 (.341)	.030 (.582)	.001 (.573)
Additional training received	.128 (.102)	.113 (.103)	.098 (.240)	.103 (.239)	.383 (.391)	.378 (.391)	−.226 (.330)	−.240 (.330)
Years enrolled in school (since exiting standard emp.)	.335** (.062)	.333** (.063)	.446** (.097)	.440** (.097)	.196† (.118)	.196 (.119)	.878** (.127)	.878** (.127)
Married	−.543** (.087)	−.526** (.088)	−.141 (.186)	−.142 (.187)	.167 (.276)	.167 (.276)	−.401 (.276)	−.404 (.275)
Preschool child present	−.478** (.085)	−.474** (.085)	.136 (.205)	.149 (.204)	.452 (.281)	.448 (.279)	−.056 (.324)	−.055 (.323)
Close to retirement age	−1.364** (.174)	−1.375** (.175)	−1.500** (.323)	−1.522** (.326)	−1.614** (.422)	−1.609** (.427)	−1.621** (.501)	−1.677** (.504)
Birth years (reference ≤ 1955)
1956–1971	.172† (.092)	.160† (.093)	−.375† (.210)	−.387† (.212)	−.531† (.295)	−.536† (.298)	−.516 (.348)	−.624† (.365)
1972 and later	.308† (.171)	.276 (.173)	−.446 (.397)	−.466 (.397)	−.469 (.534)	−.469 (.536)	−.767 (.648)	−.848 (.655)
Unemployment rate	−.082† (.043)	−.081† (.043)	.127 (.093)	.133 (.094)	.106 (.116)	.106 (.117)	.235 (.159)	.256 (.161)
Constant	−1.610** (.164)	−1.569** (.166)	−3.000** (.343)	−2.937** (.353)	−1.685** (.423)	−1.703** (.437)	−3.697** (.652)	−3.492** (.651)
N	56030	56030	13649	13649	7471	7471	6178	6178
Number of events	1219	1219	234	234	143	143	91	91

Note: Numbers in parentheses are robust standard errors.

^†p < .10

^*p < .05

^**p < .01 (two-tailed tests).

The control variables in Table 4 reveal that the process of reentering standard employment is time-dependent. The more time elapsed, the less likely it is for Japanese people to return to regular full-time employment. Furthermore, men reenter regular employment at a higher rate than women. This is also the case for first entry into standard employment, as shown in Table 3. In addition, each year of school enrollment during the nonregular employment period accelerates individuals’ reentry into standard employment, probably because school attendance increases their human capital. Being married, having a young child, and being close to retirement age are all negatively associated with the rate of standard employment reentry, suggesting conflicts between major life-course events and standard employees’ obligations in Japan. Somewhat surprisingly, the occupational status of the last regular job has no association with the return to standard employment, and those with a longer tenure at the last regular job are actually less likely to make this return. Given Japan’s seniority-based wage system, it is possible that those with a long tenure at their last regular job have reached relatively high salaries, making their matches to available standard jobs particularly difficult.

To be sure that a preference for contingent employment does not account for contingent workers’ slow rates of re-obtaining regular full-time jobs, Table 4 also shows a series of models that estimate the rate of reentering standard employment only among those who exited this type of employment involuntarily. The results differ between men and women. Among Japanese men who had lost their regular full-time jobs, accepting a contingent job delayed the return to the standard employment sector more than having no job. The coefficients become less stable, with only low-status contingent employment reaching the .10 level of statistical significance, in the model further separating contingent employment based on occupational status. I should nevertheless note that after limiting the analysis to those who involuntarily exited their regular full-time jobs, the number of observations with high-status contingent jobs is rather small, which affects the result’s stability. Unlike their male counterparts, Japanese women with high-status contingent jobs were able to switch back to regular full-time jobs at a faster rate than those without jobs. This result is consistent with the argument that the stereotype for contingent workers affects men more than women, as Japanese employers tend to view women as less devoted workers anyway. Because of the reduced effect of contingent employment on women, their likelihood of returning to standard employment corresponds more closely to occupational skills.

The results presented so far are consistent with the argument that being labeled as contingent workers hampers individuals’ transitions to standard employment. Nevertheless, it is possible that the relatively slow transitions shown for contingent workers merely reflect that they can afford to pass on less rewarding standard jobs, unlike the jobless. If this is the case, then the quality of the regular full-time jobs to which contingent workers move will be better than those the jobless transition into. To test this possibility, regression models shown in Tables 5 and 6 estimate the occupational status of the regular full-time jobs Japanese people entered and reentered, respectively, after a period of nonregular employment. According to Table 5, the durations of both contingent employment and nonemployment have nonsignificant effects. Thus, for those who are unable to land a standard job upon school departure, taking a contingent job in the meantime is no better than remaining jobless as far as their future occupational status is concerned.

Table 5.

OLS Regressions Predicting Job Quality, Conditional on Obtaining First Standard Job

	Model 1	Model 2
Duration since leaving school	−.103	(.067)
Duration, nonemployment		−.043 (.088)
Duration, contingent employment		.028 (.138)
Duration, self-employment		−.195* (.080)
Male	−2.363** (.579)	−2.188** (.590)
Education (reference Up to middle school):
High school	8.191** (.672)	8.263** (.673)
Junior college	14.193** (1.328)	14.175** (1.329)
University and above	22.599** (1.005)	22.627** (1.004)
Additional training received	.778 (.901)	.740 (.901)
Years enrolled in school (since 1st school departure)	1.190** (.300)	1.103** (.310)
Married	.910 (1.178)	.842 (1.182)
Preschool child present	−1.743 (1.291)	−1.425 (1.304)
Close to retirement age	−6.103* (3.043)	−5.622† (3.068)
Birth year (reference ≤ 1955):
1956–1971	1.375† (.818)	1.304 (.819)
1972 and later	−1.687 (1.515)	−1.820 (1.515)
Unemployment rate	−.028 (.491)	−.024 (.492)
Constant	33.584** (.930)	33.362** (.944)
Adjusted R-squared	.395	.396
N	1382	1382

Note: Numbers in parentheses are standard errors.

The models are also conditional on not entering a standard job immediately after leaving school.

^†p < .10

^*p < .05

^**p < .01 (two-tailed tests).

Table 6.

GEE Coefficients Predicting Job Quality, Conditional on Reentry into Standard Employment

	All	Men	Women
Duration, nonemployment	−.102 (.063)	−.333 (.339)	−.055 (.066)
Duration, contingent employment	−.295** (.110)	−.824** (.218)	−.139 (.120)
Duration, self-employment	−.102 (.084)	−.215 (.140)	−.079 (.103)
Duration of last standard job	.014 (.061)	−.106 (.101)	.140† (.08)
Status of last standard job	.436** (.038)	.441** (.064)	.417** (.043)
Experience of standard job recovery (reference None):
Once	.882 (.840)	1.258 (1.328)	.843 (1.12)
Twice and more times	.639 (1.812)	2.915 (2.787)	−.581 (2.215)
Male	−2.552** (.685)
Education (reference Up to middle school):
High school	3.681** (.707)	2.067† (1.070)	5.184** (.961)
Junior college	5.401** (1.272)	−.572 (2.924)	7.931** (1.436)
University and above	11.543** (1.382)	8.322** (1.848)	15.521** (2.105)
Additional training received	1.770* (.79)	3.801* (1.832)	1.036 (.863)
Years enrolled in school (since exiting standard emp.)	1.263* (.513)	1.575* (.737)	1.263* (.582)
Married	.103 (.706)	.681 (1.435)	−.517 (.789)
Preschool child present	−1.244† (.669)	−.645 (1.407)	−1.138 (.737)
Close to retirement age	−2.487† (1.473)	−2.003 (2.415)	−2.091 (1.83)
Birth year (ref. ≤ 1955):
1956–1971	.662 (.759)	−.590 (1.371)	1.309 (.893)
1972 and later	.524 (1.382)	−.800 (2.328)	.778 (1.786)
Unemployment rate	−.363 (.364)	.286 (.654)	−.855* (.431)
Constant	2.685** (1.627)	19.647** (2.523)	2.457** (1.754)
N	1208	451	757

Note: Numbers in parentheses are robust standard errors.

^†p < .10

^*p < .05

^**p < .01 (two-tailed tests).

For Japanese men and women who returned to regular full-time jobs after a period of nonregular employment, Table 6 indicates that the number of previous years spent on contingent employment is negatively associated with the status of the regular full-time jobs they were able to obtain. Conversely, the duration of nonemployment prior to standard employment reentry does not significantly affect Japanese workers’ post-reentry occupational status. This result opposes the hypothesis derived from the argument that contingent workers are more able than the jobless to afford the wait for a good job match. When fitting separate models by gender, the coefficients reveal that a history of contingent employment is detrimental only for men. Experience with contingent employment clearly worsens men’s career prospects, but nonemployment experience does not. Moreover, the more time spent on contingent work, the lower men’s occupational status is after reentering standard employment. Taken together, these results are consistent with the argument that contingent employment is highly scarring for men in Japan.

Most results for the control variables shown in Table 6 are consistent with what conventional wisdom would expect. For example, formal education, additional vocational training, being enrolled in school during the nonregular employment spell, and having held a high-status regular job all contribute to higher occupational status upon standard employment reentry, whereas being close to retirement age has the opposite effect. One notable finding, however, is that men tend to have lower occupational status upon reentering standard employment than women. The pattern is similar with regards to first entry into regular full-time employment (Table 5). Based on the average ISEI scores of last standard jobs (Table 2), Japanese men facing the possibility of returning to the standard employment sector had lower status than their female counterparts to begin with. This difference, however, is unlikely to fully explain women’s higher post-transition status because the statistical model controls for human capital and status of the last standard job. Rather, the gender difference in post-transition status likely has to do with men’s higher transition rates to standard employment than women’s (Tables 2-4). It is possible that because of employers’ overall preference for men, only particularly outstanding women are able to transition to regular full-time jobs. The greater selectivity of women accounts for their higher post-transition status. It is also possible that their breadwinner roles make Japanese men opt for a stable, albeit low-status, standard job more frequently than women. The gender difference in willingness to sacrifice occupational prestige for regular work would also lead to higher female occupational status among those who have shifted to standard jobs.

The final table, Table 7, provides results from OLS regressions predicting current earnings. In general, past experience of contingent employment has no significant effect on men’s or women’s earnings in Japan, whereas the numbers of years spent on regular full-time jobs and self-employment prior to the current job have positive influences. That is to say, as far as the economic return to work experience is concerned, having contingent work experience is exactly the same as having no work experience at all. Once again, the finding confirms that having a contingent job in Japan is no better than not having a job. Table 7 also shows that men and women currently holding contingent jobs receive lower wages than both standard wage employees and self-employed workers, even after controlling for various individual characteristics, job tenure, occupational status, firm size and hours of work. This finding corroborates previous research that contingent workers face great financial disadvantages in Japan.

Table 7.

OLS Regression Models Predicting Current Earnings

	All		Men		Women
Current job tenure (in years)	.015** (.002)	.016** (.002)	.012** (.002)	.016** (.002)	.014** (.002)	.013** (.002)
Work experience before current job (in years):
Standard employment	.015** (.002)	.016** (.001)	.011** (.002)	.014** (.002)	.016** (.003)	.015** (.002)
Contingent employment	.005 (.003)	.008* (.003)	.006 (.006)	.009 (.006)	.003 (.004)	.006 (.004)
Self-employment	.011** (.002)	.010** (.002)	.008** (.003)	.009** (.003)	.007† (.004)	.005 (.004)
Employment status (reference Standard employment):
Contingent employment	−.636** (.030)	−.532** (.030)	−.429** (.044)	−.419** (.044)	−.756** (.042)	−.566** (.044)
Self-employment	−.197** (.038)	−.208** (.036)	−.026 (.042)	−.054 (.041)	−.505** (.064)	−.454** (.062)
Hours of work		.002** (.000)		.001** (.000)		.004** (.000)
Irregular working hours		−.047 (.04)		−.014 (.046)		−.092 (.066)
Occupational status	.011** (.001)	.011** (.001)	.011** (.001)	.012** (.001)	.011** (.002)	.010** (.002)
Firm size (reference 1–29 employees):
30–999 employees	.171** (.028)	.162** (.027)	.157** (.035)	.155** (.034)	.184** (.043)	.160** (.041)
1,000 and more employees	.298** (.039)	.327** (.038)	.317** (.044)	.333** (.043)	.213** (.067)	.260** (.064)
Government office	.281** (.058)	.253** (.055)	.187* (.075)	.195** (.073)	.337** (.083)	.260** (.078)
Firm size unknown	.107† (.057)	.132* (.056)	.088 (.091)	.121 (.092)	.102 (.073)	.112 (.071)
Male	.644** (.027)	.578** (.027)
Education (reference Up to middle school):
High school	.174** (.033)	.177** (.032)	.220** (.037)	.190** (.036)	.123* (.054)	.147** (.053)
Junior college	.254** (.052)	.252** (.050)	.359** (.086)	.326** (.083)	.187* (.074)	.191** (.071)
University and above	.275** (.045)	.280** (.044)	.306** (.049)	.291** (.048)	.253** (.087)	.265** (.083)
Additional training received	.034 (.034)	.018 (.033)	−.034 (.047)	−.045 (.045)	.068 (.047)	.046 (.045)
Married	−.036 (.027)	−.036 (.026)	.294** (.036)	.240** (.035)	−.296** (.04)	−.229** (.038)
Preschool child present	.039 (.035)	.041 (.034)	−.009 (.042)	.008 (.040)	−.008 (.057)	.009 (.054)
Close to retirement age	−.133** (.034)	−.075* (.033)	−.186** (.047)	−.156** (.045)	−.041 (.049)	.027 (.046)
Constant	13.619** (.062)	13.225** (.069)	14.048** (.070)	13.858** (.082)	13.932** (.104)	13.237** (.116)
Adjusted R-squared	.517	.558	.315	.343	.386	.446
N	3586	3385	1914	1811	1672	1574

Note: Numbers in parentheses are standard errors. The models also control for industry, for which the coefficients are omitted to conserve space.

^†p < .10

^*p < .05

^**p < .01 (two-tailed tests).

Conclusions

With rapid rises in contingent employment arrangements in advanced industrial economies, researchers are paying increasing attention to the long-term impact of these arrangements on workers. By comparing the consequences of taking a fixed-term job with those of nonemployment, this analysis provides new evidence on how scarring contingent employment can be on individuals’ careers. Taking a contingent job, rather than continuing to be jobless, generally decelerates individuals’ rates of obtaining standard jobs in Japan. Contingent workers’ slower transition to, or longer search for, standard jobs by no means indicates that these workers will achieve a higher post-transition occupational status than the jobless. On the contrary, for Japanese men attempting to recover from their loss of regular full-time employment, the experience of contingent employment is more detrimental to their post-recovery occupational status. Results from this study are therefore consistent with the argument that contingent employment experience signals to employers that job candidates belong to the secondary labor market. Without being labeled the same way, jobless adults in Japan, especially males, are likely to have more and better job opportunities in the standard employment sector than those with fixed-term jobs. Thus, to the question of whether a contingent job is still better than no job, the answer is clearly negative for Japanese men (and not so positive for Japanese women, either).

Using the case of Japan, this study demonstrates the importance of labor market contexts in shaping the influences of contingent employment on individuals’ careers. Whereas fixed-term jobs are thought to be a stepping stone to permanent work in some European countries (Booth et al. 2002; De Graaf-Zijl et al. 2011; Gash 2008), results from my analysis indicate that such jobs only postpone Japanese workers’ transitions to, as well as diminish their future prospects for, standard employment. I argue that Japan’s welfare corporatist practices, including lifetime job protection and abundant fringe benefits for standard employees, have led employers to be extremely risk averse in selecting workers for regular full-time jobs. As labor-market theorists (Sørensen and Kalleberg 1981) suggest, this tendency of Japanese employers is conducive to an exceedingly segmented labor force, with the labor pool for standard jobs being highly exclusive. In this context, upon becoming an organizational outsider – by taking a contingent job – an individual is likely to be permanently blocked from the insider labor pool. Thus, transitioning from contingent to standard employment appears to be more difficult in Japan than in certain European countries.

The strict separation of labor pools for standard and contingent jobs also explains the finding that contingent employment obstructs men’s career prospects more than women’s in Japan. Women are often considered as members of the secondary labor pool in a highly segmented labor market (Sørensen and Kalleberg 1981). This is especially the case in Japan, as its welfare model assumes that women shoulder caring responsibilities, making all women less committed workers. A history of contingent employment therefore makes relatively little difference in Japanese women’s likelihood of entering the standard employment sector. Conversely, whether Japanese men possess contingent job experience serves as a key indicator of their labor pool membership, thus affecting their opportunities for regular full-time jobs.

In addition to the particularly negative implications of contingent employment, I also argue that the Japanese context helps mitigate the scarring effects of joblessness. Specifically, the welfare-corporatist model Japan has followed for the past half-century has made employers responsible for providing jobs for all able workers, especially male ones, in the society. This social norm alleviates the blame on individuals for their unemployment. Therefore, a jobless spell in Japanese men’s work histories scars their careers less than a spell of contingent employment. In this sense, this study also demonstrates that the extent to which unemployment scars workers’ careers is context-dependent. More important, while previous research on how national contexts mediate the scarring effects of unemployment generally focuses on the role of welfare states (Gangl 2004, 2006), this study suggests that the prevalence of welfare corporatism constitutes another important contextual factor.

By linking Japan’s labor market institutions to the long-term consequences of contingent employment, this study further generates insights for research on fixed-term work arrangements in advanced industrial economies in general. Specifically, the results from Japan suggest a close connection between labor market segmentation and the scarring effects of contingent work. Contingent jobs are likely to be worse options for workers in highly dualist economies, in which a sharp divide in both rewards and requirements exists between standard and contingent jobs. Paradoxically, this divide is also likely to create incentives for management to adopt contingent employment practices, leading to rapid increases in contingent workers in such economies.

The second general insight derived from this study has to do with stratification researchers’ increasing attention to life events that have long-term implications for individuals’ economic wellbeing, such as family dissolution, employer change and job displacement (DiPrete 2002; DiPrete and McManus 2000; Gangl 2004). This study demonstrates that in a highly segmented labor market like Japan’s, the long-term impact of taking a contingent job can be even greater than that of unemployment. Future stratification research therefore must consider entering contingent employment as a major event that triggers social mobility and focus more on how this event impacts individuals differently across the industrialized world.