You are on page 1of 22

Social Science Japan Journal Vol. 15, No.

1, pp 3152 2012 Published online January 5, 2012

doi:10.1093/ssjj/jyr040

............................................................................................................................................................................................................

Not by Education Alone: How Young Adults Employment Status Is Determined by Employment Environments and Family Backgrounds
Takehisa SHINOZAKI*
This paper quantitatively investigates the connections between the growing number of young adults in Japan who are out of the workforcenot in school, working, or seeking workand their employment environments and family backgrounds by drawing upon six years of data from sampling studies that were conducted from 2000 to 2006. A multinomial probit model is used to generate estimates of the possible impacts of changes in local job markets and varying family characteristics on the likelihood of young adults being unemployed (not working but seeking work) or out of the workforce altogether. The paper presents the most detailed testing to date of hypotheses concerning the factors that lead young adults to leave the labor force. I nd that apart from the respondents educational backgrounds, their parents employment status when the respondents were adolescents and past household income are signicant determinants of young adults employment outcomes. Leaving the workforce is particularly correlated with previously having a high standard of living and not attaining the same level of education as ones same sex parent.
Keywords: youth labor market; employment status; family background; NEET.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

1. Introduction
This paper uses the results of large-scale surveys conducted from 2000 to 2006 to examine how young adults employment status is affected by their local employment environments and family backgrounds. Japans economy has been lagging since the mid-1990s, which has not only reduced the demand for new graduates in the labor market but also forced many middle-aged and older workers into early retirement. Compared to every other age group, young people have seen the sharpest decline in employment rates. For example, from the 1990s to the 2000s, the unemployment rate among 15- to

............................................................................................................................................................................................................

Takehisa SHINOZAKI is an Associate Professor of labor economics at the Faculty of Science and Engineering, Waseda University. He has authored several articles on youth labor, the economics of education, and wage inequality, including Wage Inequality in Japan, 19792005 [Japan Labor Review 3(4): 422] and Why Part-time Workers Do Not Accept a Wage Gap with Regular Workers [with Mamiko Ishihara, Japan Labor Review 2(2): 5577]. He can be contacted at the Faculty of Science and Engineering, Waseda University, 3-4-1 Okubo, Shinjuku-ku, Tokyo 169-8555, Japan, or by e-mail at shinozaki@waseda.jp. *Translated from the Japanese by Margaret Gibbons. This research project was made possible through the cooperation of the Joint Usage/Research Center for Japanese General Social Surveys (accredited by the Minister of Education, Culture, Sports, Science and Technology) at the Osaka University of Commerce and the Institute of Social Science at the University of Tokyo. I would like to express my gratitude for the very valuable comments given to me by Genda Yuji (University of Tokyo), Ohta Souichi (Keio University), Iwai Noriko (Osaka University of Commerce) and Takahashi Yoko (University of Tokyo). In revising this article, I beneted greatly from the insightful comments of the four anonymous referees. I also wish to convey my appreciation to the JGSS Research Center at the Osaka University of Commerce which designed the surveys I drew upon and the Social Science Japan Data Archive (SSJDA) at the University of Tokyos Institute of Social Science which provided me with the data. Any errors are entirely my responsibility.
The Author 2012. Published by Oxford University Press in conjunction with the University of Tokyo. All rights reserved.

32

Takehisa SHINOZAKI

Age 15-19 (right axis)

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

Figure 1. The Growing Age Gap in Non-Employment. 24-year olds soared, reaching a peak of 10.1% in 2003. For a few years after that, economic recovery pushed up the employment rate of young people, but those gains were reversed by the 2008 Lehman shock recession. The unemployment rate of young workers rose as high as 9.1% in 2009. Beyond the worsening unemployment rate, a growing proportion of 15- to 24-year olds have been exiting the labor force altogether since the early 1990s. Figure 1 illustrates this trend using results from the Labor Force Survey, conducted monthly by the Ministry of Internal Affairs and Communications. Over the same time period, however, the labor force participation rate of people age 2534 continued to rise as it has been for the past 30 years. This rise has been credited to women spending fewer years as full-time housewives due to marrying later and having greater access to maternity leave, due to improvements in labor laws and company practices, which have made it somewhat easier for mothers to remain in the workforce. Returning to people aged 1524 years, their labor force participation rate has been trending downward since 1992. It is reasonable to assume that some of the young adults who expected to enter the job market in the early 1990s, after the collapse of the economic bubble had brought hiring to a standstill, opted instead for some type of post-secondary education. In all likelihood, this rst group of post-bubble graduates saw their return to school as a temporary departure from the workforce and resumed their job searches as graduation approached. However, in the early 2000s, researchers began to suspect that more new graduates were neither looking for work nor returning to school, a category referred to below as non-employed. Because they were not actively job hunting, these young people were not counted as unemployed in Japans , a term applied to young, often part-time employment statistics. Nor were they getting by as furiita workers who go from one insecure, low-wage job to another. According to the Ministry of Health, Labour and Welfares (MHLW) (2010) White Paper on the Labour Economy, 630,000 people ages 1534 (referred to collectively as young adults herein) were non-employed in 2009.

Young Adults Employment Status

33

In order to truly understand these young adults and what led them to abandon the search for employment, Japanese scholars began to study this group in the early 2000s, looking to prior research from the UK on that nations NEET (not in education, employment, or training) youth as a guide. In 2003, Kosugi and Hori, scholars in the eld of the sociology of education, published the rst Japan-based study focusing on changes in the ways Japanese students transitioned from school to work. Seeking ways to help young people who had left school but not joined the labor force for whatever reason, Kosugi and Hori interviewed members of various organizations dedicated to helping young people nd work and presented their ndings. Taking a more quantitative approach, Genda (2004) analyzed published government statistics from numerous reports and found that the number of young adults not in the labor force had grown rapidly. Shortly after these initial articles were published, many other studies contributed to our understanding of non-employed young adults in Japan. Several quantitative studies mined data from the Employment Status Survey, conducted by the Ministry of Internal Affairs and Communications, to identify the nature, extent, and evolution of the non-employment problem. Their ndings clearly showed that many young adults who had left the labor force had relatively little education, lived with their parents, and were not necessarily from afuent families (Kosugi and Hori 2005; Genda 2005b). Among articles on region-specic factors and employment, Kosugi (2004) found a positive correlation between regional unemployment rates and local youth non-employment rates and Ohta (2005) reported that young adults attitudes toward work were affected by regional variations in economic conditions. Yugami (2005) found that the worse local job markets became, the less enthused young adults were about working and therefore concluded that this reactive loss of enthusiasm is a major cause of leaving the labor force. In surveys by the MHLW (2003) investigating the links between non-employed individuals personal circumstances, career aspirations, and their decisions to leave the labor force, many respondents said they were not looking for work because they doubted they were capable of doing well on the job while many others said that health problems were keeping them sidelined. Genda (2005b) reported that many young people either said they had searched to no avail or that they could not nd work that suited them or their career goals. In Japans labor market, people with limited education or long histories of unemployment are highly unlikely to nd work that is not at the bottom of the pay scale. Genda (2007a,b) also found that people with low expected earnings are more likely to quit searching and leave the labor force.1 Much of the research on people out of the labor market has been based on aggregate statistics or micro-level data from government surveys. Given that the number of young adults who have withdrawn from the workforce is small relative to the overall population, using microdata from large national surveys has the benet of ensuring a sufcient sample size. On the other hand, these surveys do not tell us much about respondents personal attributes and circumstances, such as their family backgrounds, that may be closely linked to their employment status. Due to these omissions, it is quite possible that conclusions based solely on government survey data will be biased. An alternate source of data could be of considerable value in augmenting our analysis.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

1. In addition to these studies, Gaston and Kishi (2005) examine the connections between Japans labor policies and the decline in labor force participation rates among young adults. For more on studies of young adults in Japan falling into the non-employed category and efforts to reverse this trend up to the mid-2000s, see bibliographies compiled by Kameyama (2006) and Ito (2006).

34

Takehisa SHINOZAKI

Fortunately, we have such a source in the Japan General Social Surveys (JGSS), described in more detail below. This paper uses detailed data from the JGSS to show the relationships between the personal characteristics, family backgrounds, and the local economic environments of young adults in Japan and their participation in the workforce. More specically, it examines how family characteristics and the employment environments affect whether one is working, unemployed or out of the workforce altogether. To summarize my ndings, whether young people are employed, unemployed or non-employed is determined not only by whether they dropped out of school or their level of education but also by their families living standards and their parents employment status when the respondents were adolescents. Moreover, across these variables the unemployed have considerably more in common with people who are working than with those who out of the labor force. In other words, educational and family backgrounds help to predict who will stop looking for a job, but they do not help us differentiate between the employed and unemployed. The remainder of this paper is divided into four sections. Section 2 presents an overview of the issue of young adults leaving the workforce in Japan and its signicance. This overview is integrated with a review of prior research. Section 3 describes the data used in the analysis and Section 4 describes my ndings on how the economic and home environments of young adults are linked to their employment status. Section 5 concludes and discusses the implications of my results and directions for further research.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

2. Research in Japan on Non-employed Young Adults


Before delving into the details of my analysis, this section looks at prior studies of young people who have opted out of the workforce in Japan to assess the signicance of these investigations as well as how they relate to research in other nations, especially work done in the UK on youth not in education, employment or training. This section also summarizes the analytical implications of placing people without jobs into separate categoriesunemployed and non-employed excluding the unemployedas done in most studies on young Japanese adults, including this one (Genda 2007a is one of the rare exceptions). Researchers outside of Japan may nd it odd to distinguish between people who are jobless based on whether they are looking for a job or not. Studies in the UK and reports by the Organisation for Economic Co-operation and Development (OECD) tend to treat all of those out of work as one population of jobless people. For example, OECD reports in 2000 and 2006 compared changes in the numbers of young adults out of school and out of work across OECD member nations. The OECD found that the rate of joblessness among young people in Western nations fell in response to the economic expansion of the 1990s but climbed upward again around 2000 before leveling off at roughly 10%.2 Although some UK and OECD reports do not place all young people out of work in one category (e.g. OECD 2009), they are relatively few in number. What explains the disparity between how Japanese and British or Western European researchers approach the issue of joblessness? At least part of the answer may lie in differences in how joblessness is framed as a policy issue in Japan and Western nations.3 The UK and other Western European nations have faced the problem of high youth unemployment since the 1970s, and over the decades, their governments have implemented numerous

.........................................................................................................................................................................................................................

2. The OECD (2000, 2006) reports did not cover Japan, precluding a JapanEuropean Union (EU) comparison. 3. The following section is based on Furlong (2006b) and Kosugi (2010).

Young Adults Employment Status

35

countermeasures to reduce joblessness.4 Among these efforts was a program in the UK that made participation in job training programs a condition for receiving unemployment benets. Despite the nancial inducement, the training programs failed to attract many young people who then lost their unemployment benets. Beginning in the late 1980s, people who were cut from the benet rolls for dropping out of training programs were no longer counted as unemployed in UK government statistics. Despite this ofcial reclassication, the actual employment status of these young adults remained the same and they still needed support to make their way into the workforce. As a result, labor scholars in the UK began to use the NEET category to encompass all young people who were out of school and out of work regardless of what government benets they were receiving (Furlong 2006b: 7071).5 British researchers continuing inclusion of young job seekers in their studies of teenagers out of the workforce reects these prior events. After the UK government tightened the qualications for unemployment insurance, it offered new job training programs to youths 16- to 18-years old who had lost their unemployment benets. These training programs were prompted by fears of the various social costs that would arise from doing nothing to help young people extricate themselves from joblessness. Labor scholars turned their attention to evaluating how effective these programs were and how costly it would be to society if they failed. For example, the Social Exclusion Unit (1999) found that people who were out of school and work for extended periods were thereafter more likely to receive public assistance, be unemployed, commit crimes, and suffer health problems. They concluded that an increase in the NEET population would therefore be likely to lead to higher social security expenditures. Bynner and Parsons (2002) studied longitudinal data from the 1970 British Birth Cohort Study and found that teens in the NEET category continued to have greater difculty in the job market after they entered their 20s. Godfrey et al. (2002) and Coles et al. (2002) also estimated the costs of being out of the labor force to young people themselves, their families, and society. The ndings of these studies and others prompted the UK government to redouble its efforts to train jobless teens through new programs such as Connexion career services (Social Exclusion Unit 1999; Okita 2003; Kosugi 2010). Unlike in the UK and other European nations, youth unemployment in Japan garnered little attention in the 1990s and was not considered a signicant problem. Unemployment countermeasures were primarily targeted toward people in their 40s and 50s. At the end of the decade, the employment outlook for young people in Japan sharply deteriorated for a variety of reasons including economic recession, rapid developments in information technology, and removal of government restrictions on hiring temporary workers. Although young adults job prospects had grown much dimmer, unemployment programs continued to focus on helping older adults nd work due to the governments scal problems and other factors. At that time, youth unemployment was regarded by many as a lifestyle choice and not taken seriously. There was a strong sense that young people losing their jobs or oating from one low-level job to the next are problems they need to sort out for themselves (Kosugi 2008: 8).

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

4. For an assessment of the effectiveness of European and North American government programs targeting joblessness among young adults, see, for example, Mitani (2001a). 5. For a brief period before the NEET term was settled on, non-employed young adults in the UK were said to be in status zero, but this label was dropped due to its negative connotations (Furlong 2006b: 72). Some British studies of non-employment use various other terms such as generation X and off-register (Bynner and Parsons 2002).

36

Takehisa SHINOZAKI

Early in the 2000s, such attitudes began to change. Researchers such as Genda (2001, 2005a) demonstrated that unemployment and insecure employment among young adults resulted not from their reluctance to work but from objective conditions such as insufcient demand for workers and the sclerotic nature of the labor market. Kosugi and Hori (2003) found that changes in the process by which students moved from the classroom to the workplace had created other obstacles so that fewer graduates were able to smoothly transition into regular employment. However, the current weak demand for labor in Japan will not last, which makes developing the abilities of young adults important for the economy as a whole. Since the 1970s, it has been apparent that Japans falling birth rates would eventually lead to a drop in the working age population while the number of elderly would continue to grow. By 2000, that eventuality was clearly approaching more quickly than in Europe, and guring out how to deal with the impending labor shortage jumped toward the top of the policy agenda (MHLW 2005). Given the failure of government efforts to increase the birth rate and the resistance to allowing large-scale immigration, Japans labor demand will have to be met by bringing a higher percentage of adults into the workforce. Policymakers began paying more attention to raising labor force participation rates, especially among retirees, women, and young people. Research in Japan on young adults who had exited the labor market was shaped by the earlier ndings mentioned above and scholars awareness of the social and economic implications of substantial numbers of people being largely cut off from society. In terms of research objectives, while UK scholars focused on evaluating the effectiveness of programs to promote youth employment and treated jobless young people alike regardless of whether they were seeking work or not, in Japan, researchers factored in the coming labor shortage along with young peoples dwindling chances of nding regular employment. Japans policies to reduce youth unemployment and isolation also differed greatly from the UK where large-scale job training programs for young people were established. As Japans employment programs did not specically target young people, little was done to evaluate how those programs impacted people in their late teens and early 20s. Japans concern with ensuring a sufciently large supply of labor led to researchers and policy analysts to broaden the target youth population, in some cases including people from age 15 to 34. The analysis in this paper was shaped by Japans policy debates and therefore its approach differs from UK studies in how they conceive of people not in education, employment or training. In Japan whether someone is receiving unemployment benets or not is irrelevant, but whether they are looking for work or not does determine their status. Given Japans concerns about avoiding the looming shortage of labor, examining why people remain outside of the workforce is especially relevant today.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

3. Data Overview
The data used in this paper come from the Japan General Social Surveys which are conducted jointly by the Institute of Regional Studies at the Osaka University of Commerce and the Institute of Social Science at the University of Tokyo.6 The JGSS is a large-scale sampling study which covers a range of topics including attitudes, behavior, household structure, and employment. More specically, it includes questions on what type of industry respondents work or worked in, their occupations, political opinions, religious beliefs, lifestyles, and legal consciousness. The JGSS also collects individual and household

.........................................................................................................................................................................................................................

6. A detailed English language description of the JGSS can be found at http://jgss.daishodai.ac.jp/english/index.html.

Young Adults Employment Status

37

information such as sex, age, educational background, and household income. The JGSS can be used for cross-national comparisons as it is one of the international surveys that include questions modeled on the General Social Surveys conducted by the National Opinion Research Center at the University of Chicago. The JGSS targets men and women across Japan aged 2089 years. A two-stage, stratied sampling method is used to select subjects. The rst stage of the stratication is comprised of six geographic regions, and the second stage divides each region into three categoriesmetropolises, medium-sized cities, and towns and rural areasresulting in 18 strata.7 The surveys are generally conducted from October to November and include both interviews and self-administered questionnaires.8 In general, interviews are used to collect information on work and family matters while the self-administered questionnaires cover daily activities in addition to respondents opinions and expectations on a variety of personal, social, and political topics. After completing the questionnaires, respondents return them by post. The approximate number of valid survey responses was 3,000 from 2000 to 2002, 3,500 in 2003, 2,000 in 2005, and 4,000 in 2006. The JGSS sample size is small relative to the Employment Status Survey and other government sources used by Genda (2007a,b) and others, but the use of strict random sampling allows us to assume that the sample is representative. Moreover, the JGSS collects more information on respondents family backgrounds and parents than the Employment Status Survey.9 3.1. Data Set Characteristics Linked data sets from six years are used in this paper including the four years covered by JGSS Cumulative Data 20002003, as well as the 2005 and 2006 JGSS surveys (a survey was not conducted in 2004). Although the percentage of young people out of work in Japan grew from the late 1990s to the 2000s, it remained smaller than in the EU. The six years of data ensure a sample size large enough for quantitative analysis. From this linked set, responses from people aged 2034 who were not working and had never been married were selected.10 Married people were left out in order to exclude wives who had opted out of paid employment to look after their families, as their motives for leaving the labor force presumably differ from other non-employed people.11

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

 /To hoku, Kanto  , Chu bu, Kinki, Chu goku/Shikoku, and Kyu  shu  . The largest cities (13 from 2000 to 7. The six blocs are Hokkaido 2002, 14 from 2003 on) were placed in one category. The remaining cities form the second category, and towns and rural areas comprise the third. As respondents were sampled from local voter registration rolls they were all at least 20-years old, the minimum voting age. 8. The 2005 JGSS was conducted from August to November 2005. 9. The other great advantage of the JGSS over government surveys is its availability to researchers. The use of microdata from the Employment Status Survey and other government sources is strictly controlled within Japan; its use is simply denied to researchers outside of Japan. JGSS microdata have been made available to researchers everywhere including graduate and undergraduate students through the Inter-university Consortium for Political and Social Research and the SSJDA at the Institute of Social Science, University of Tokyo. 10. The JGSS does not include questions on vocational training so it could not be used as a selection criterion. 11. Removing married people from the sample enables us to exclude women in male-breadwinner households who have opted not out of employment but out of paid employment. However, if there is some trait that is associated with being unmarried that is also related to employment outcomes, the estimation results may be biased. More specically, if there is an unobserved variable that is characteristic of single people (such as a commitment to remaining unmarried) that is correlated with both an independent variable and its error term, the coefcient of the independent variable will be overestimated. In this article, the only independent variable that might be at all likely to correlate with wanting to stay single is the highest level of school attended, and so its interpretation comes with this caveat.

38

Takehisa SHINOZAKI

Finally, as described in the following section, out of the 1,584 respondents who were age 2024 and unmarried, 61 (3.85%) were non-employed, 98 (6.19%) were unemployed, and 1,425 (89.96%) were working. Adding more variables to the basic model leads to sample attrition but not to the point where robustness is lost.

4. Methodology and Findings


4.1. Statistical Models and Variable Selection Multinomial probit models are used to evaluate possible determinants of young peoples employment status.12 Multinomial logit models are generally used when there are several independent variables to choose from, but such models must satisfy the independence of irrelevant alternatives (IIA) criterion by passing the Hausman and SmallHsiao tests.13 In this study, a logit model that passed the tests would still face a problem because the similarities between jobless young people and their unemployed peers, as well as the similarities between employed and unemployed young adults, make it unlikely that the error terms will be independent.14 Consequently, this analysis takes a conservative route by using multinomial probit models which do not need to meet the IIA standard.15 The dependent variable is employment statusemployed, unemployed or non-employed. Young, unmarried people who are non-employed are identied by their selection of three survey answers: I am not working, I am not attending school, and I am not looking for work. Their unemployed counterparts answered, I am not working, I am not attending school, and I am looking for work. Finally, the employed group selected I am working and I am not in school. The base category in the multinomial probit model is employed.16

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

12. Employment outcomes can be regarded as nested choices. For example, one can rst choose between working and not working, and then choose between looking for work and not looking for work. Generally speaking, nested probit models are the preferred method of analyzing nested choices. However, given the lack of objective criteria for specifying the appropriate nested choice structure, I chose multinomial probit models for this article. 13. For more details on multinomial probit models, see, for example, Long (1997). 14. To see if the IIA test could be met, instead of the basic model specied below, I used a multinomial logit model. In this article, it is assumed that there are correlations among error terms within clusters and so instead of the standard HausmanMcFadden tests, I applied the generalized HausmanMcFadden tests and found that the IIA assumption could not be rejected. However, the IIA condition was not met using the SmallHsiao tests. As Fry and Harris (1998) recommend using more than one test of the IIA assumption, and the multinomial logit model failed the SmallHsiao tests, it was rejected in favor of the multinomial probit model. Furthermore, according to Cheng and Long (2007), commonly used tests of the IIA assumptionsuch as HausmanMcFadden and SmallHsiaofrequently yield misleading results. Cheng and Long generated a data set that satised the IIA condition and then applied the most widely used IIA tests to that data. They found that these tests frequently resulted in false negatives, i.e. indicated that the IIA condition was not met. Cheng and Long conclude that multinomial logit models should only be used when the possible values of the dependent variable are clearly distinguishable from each other. 15. In the multinomial probit models used in this paper, robust standard errors were used to evaluate correlations between error terms within clusters. Each strata of the JGSS would be preferable as clusters but as JGSS strata-specic data are not available, prefectures with similar employment environments were chosen as the cluster units. 16. Although this paper uses broad categories in its assessment of employment status, earlier studies such as Genda (2007b) and Honda and Hotta (2006) treat unemployment and non-employment more as part of a continuum, and look more closely at peoples desire to work and how they look for work. It is not the case that one method is superior to another. Rather, both types of methods need to be used. For a discussion of how youth employment is conceptualized and the consequences of those conceptualizations, see Furlong (2006a).

Young Adults Employment Status

39

The independent variables evaluated here are related to individual traits, local employment environments, and family backgrounds.17 The following list describes the variables and their expected values. a) Gender dummy: Female respondents are given the dummy value of 1. b) Age: Respondents age at the time of the survey and the age squared term. c) Highest level of school attended dummy: Any post-secondary institutionjunior college, technical college, four-year college or graduate schoolis coded as 1. Formal schooling that ended at the middle school or high school level is a dummy variable with a value of 0. d) School leaver dummy: If the respondent did not graduate from his or her last school, that response is the dummy variable with a value of 1. In Japan, a large fraction of new hires are recruited directly out of school and relatively few people are hired mid-career. In other words, there are few second chances for people who quit school and then fail to nd work. The expected value is positive, i.e. school leavers are more likely to leave the workforce or be unemployed. e) Regional unemployment rate: unemployment rate (%) of the prefecture or major metropolis where the respondent resides. Percentages are taken from the Labor Force Survey. The presumption is that the rate of prefectural unemployment will be positively correlated with respondents being unemployed or out of the workforce. f) Regional minimum wage: Minimum wage (in yen) of the prefecture or major metropolis where the respondent resides. There is a comparatively wide range of minimum wage levels across Japan. The assumption is that higher minimum wages will depress young adults employment rates.18 g) Generation dummy, birth year-based dummy: The reference group was born before 1971. In Japan, the employment environment a person faces at graduation generally has a long-term impact on his or her wages and employment status. This phenomenon is known as the generational or cohort effect.19 The objective is to evaluate whether a similar generational effect is determining the employment status of young people in the 2000s. The expectation is that respondents in their teens and early 20s experienced a tougher employment environment than respondents in their late 20s and early 30s. The expected value is positive. h) Fathers occupation dummy20: Dummy is coded as 1 if the respondent reports his or her father was a rank-and-le worker when the respondent was age 15.21 Willis (1981) is representative of the classic treatment of the question of whether and how childrens educations and career paths

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

17. Ryan (2001) found that lack of motivation to work among young people is related to general economic conditions, education levels, and household characteristics. 18. In the United States, there has been a long running debate over whether raising the minimum wage suppresses employment among young people and service sector workers. The most noted antagonists in the debate are the group of David Card and Alan B. Krueger on one side and the group of David Neumark and William Wascher on the other (Card and Krueger 1994; Card and Kreuger 2000; Neumark and Wascher 2000). Neumark and Wascher (2006) reviewed prior studies of the effects of minimum wage and found that two out of three reports concluded that minimum wage increases negatively affected employment, which they used to bolster their argument that raising the minimum wage raises unemployment. 19. A number of studies on cohort effects in Japans labor market have been published including Ohtake and Inoki (1997), Genda (1997), Kondo (2007), and Genda et al. (2010). There is a path dependency problem with other factors related to entering work. For more information, see Mitani (2001a). 20. Genda and Okada (2004) refer to the signicant potential impact of programs that introduce students around 14 years of age to work settings. These programs, which are made possible through the cooperation of schools, local businesses and families, are credited with helping to develop students interpersonal and communication skills. Such positive outcomes led Hyogo Prefecture to create a ve-day work experience program for students across the entire prefecture. 21. Occupations that are labeled rank-and-le here correspond to JGSS occupation code numbers 620689, 702 and 704.

40

Takehisa SHINOZAKI

are inuenced by those of their parents. My focus is slightly different from that of Willis and other earlier works in how I evaluate the effects of parents occupations. i) Mothers employment dummy: If a respondents mother was employed when he or she was 15, the dummy value is one. Using JGSS data, Tanaka (2008) found that women were more likely to work full time if their mothers had worked full time and more likely to be self-employed if their mothers had their own businesses. Tanaka concluded that mothers employment status inuenced their daughters career paths. The dependent variable values in Tanakas study were full time, part time, self-employed, and not employed. We will see if a similar relationship exists between mothers and childrens work experiences when the dependent variable values are employed, unemployed, and out of the workforce. j) Past household income dummy: If a respondent believes that his or her households annual income was above average when the respondent was 15, the dummy value is 1. This subjective assessment indicates the living standard of the respondents household in the past. If consumption has a habit-forming aspect, we can expect that young people who had a high standard of living while growing up would aspire to achieve a similar standard in the present and therefore hope to earn enough to sustain that level of consumption.22 The living standard variable is related to the discussion below on the reservation wages of young people. k) Fathers highest level of school attended dummy: If a respondents father attended a postsecondary school, the dummy value is 1. l) Mothers highest level of school attended dummy: If a respondents mother attended a post-secondary school, the dummy value is 1. Tanaka (2008) reported that women whose mothers had high levels of education were more likely to work full time. m) Highest level of school attended dummy and parents highest level of school attended dummy interaction term: To clarify what is meant by interaction effects in this paper, I use the dummy variable for respondents highest level of school attended and the dummy variable for the highest school level of her or his same sex parent to create three dummy variables to express the combinations of parent and child educational backgrounds. In other words, there is a dummy variable for respondents who attended a post-secondary school although their same sex parents did not, another dummy for the reverse caseparents with more education than their same sex children, and a third dummy for respondents who, like their same sex parents, attended post-secondary schools. The base category is respondents whose education, like their same sex parents, did not go beyond high school. Comparing the respondents education levels with their same sex parents in this way differs subtly from using a simple interaction term. Survey year dummies are introduced as control variables. 4.2. Descriptive Statistics Table 1 below presents the descriptive statistics on the variables used in the multinomial probit estimates.23 Women make up a high percentage of the non-employed and a low percentage of the unemployed. There is virtually no difference in the average ages of the three groups nor in the standard deviations of their ages. At rst glance, this homogeneity seems odd given that, according

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

22. For a literature review of studies of habit-forming consumer behavior in the eld of economics, see Ikeda (2003). 23. In addition to the variables listed in Table 1, household income over the preceding year would be a useful variable to include but is omitted from this analysis because the response rate on income questions is low in the JGSS and including it would reduce the sample size and make estimates unstable. Genda (2007b), relying on data from the Employment Status Survey, found that respondents who reported that other members of their households had higher incomes were more likely to be non-employed or unemployed and less likely to be employed.

Young Adults Employment Status

41

to the Labor Force Survey, unemployment is more prevalent among the youngest adults than their somewhat older peers. However, this unexpected result is the product of excluding married people from this analysis and not due to any sampling bias in the JGSS data. Including all young people in the JGSS data set produced the same pattern of higher unemployment among the youngest adults as found in the Labor Force Survey.24 School leavers comprise a high proportion of the non-employed. Several previous studies have found a link between dropping out of school and leaving the labor force (Genda 2004; Kosugi 2004) and the JGSS data enable us to test this link. A primary goal of this paper is to assess whether leaving school is still closely linked to non-employment after controlling for other factors. The unemployment rates are slightly higher and the minimum wages slightly lower in the areas where the non-employed and unemployed live compared to where the employed respondents live, as one would expect. A birth cohort effect appears among those born from 1977 to 1981, who have a higher unemployment rate than those born from 1972 to 1976 and 1982 to 1986. The job market was at an all-time low when this group nished high school and college and many were unable to nd work. It is also believed that many members of the 19771981 cohort who found work were often a poor match for their jobs and likely to quit or be red fairly quickly.25 Compared to the employed and unemployed respondents, non-employed respondents were less likely to report that their fathers had rank-and-le jobs when the respondents were 15. They also reported that their mothers were less likely to have been employed when the respondents were 15 than the mothers of the employed and unemployed. This result probably reects that women who are married to rank-and-le workers, whose average wages are lower than those of managers, are more likely to work to keep their households incomes above a certain level. In addition, we can also see that a greater fraction of non-employed people reported having higher than average household incomes than the other two respondent groups. In short, we can infer that on average the nonemployed respondents had a higher standard of living when growing up than respondents who remain in the job market or employed. The non-employed group further differs from the other two groups in that a greater share of its members report having parents with at least some higher education, although the non-employed themselves have less education than the other groups. Kikkawa (2006) and others have found that children of highly educated parents are predisposed toward pursuing higher education themselves, but this pattern does not apply to the non-employed respondents. It may be the case that failing to match the educational achievements of ones parents affects ones employment status. The interaction term between the highest level of school attended and same sex parents highest level of school will verify this relationship. The non-employed are distinct from the other two groups in that only a small

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

24. The pattern of employment rates rising with age holds true for both men and women in the sample when assessed separately. The Labor Force Survey also reports higher rates of unemployment for men than women. 25. The merging of survey data which span seven years is based on the implicit assumption that there were no major changes in the employment environment during that time. To test this assumption, I divided the sample by age cohort to see if the percentages of people unemployed and non-employed remained the same in each year using a chi-squared test (P-values were calculated using Fishers exact test). The null hypothesis of there being no variation in rates could not be rejected at the 5% condence level. Based on this result, we can conclude that there is no difference in the ratio of unemployed to non-employed people from 2000 to 2006.

Table 1.

Descriptive Statistics.
Non-Employed Sample Unemployed Sample Employed Sample

42 Takehisa SHINOZAKI

Cases Mean

SD

Minimum Maximum Cases Mean Value Value 1 34 11.56 1 1 7.8 7.17 1 1 1 1 1 1 1 1 98 98 98 98 98 98 98 98 98 98 81 94 89 78 82 0.459 26.235 7.043 0.439

SD

Minimum Maximum Cases Mean Value Value 1 34 11.56 1 1 8.3 7.19 1 1 1 1 1 1 1 1

SD

Minimum Maximum Value Value 1 34 11.56 1 1 8.4 7.19 1 1 1 1 1 1 1 1

Female dummy (female 5 1) Age Age squared/100 Highest level of school attended dummy (post-secondary 5 1) Dropout dummy (dropout 5 1) Prefectural unemployment rate (%) Prefectural minimum wage (U100) Generation dummy (base: born before 1972) 19721976 Birth year dummy 19771981 Birth year dummy 19821986 Birth year dummy Father blue-collar worker when respondent was age 15 dummy (yes 5 1) Mother employed when respondent was age 15 dummy (yes 5 1) Household income at age 15 dummy (above average 5 1) Fathers highest level of school attended dummy (higher education 5 1) Mothers highest level of school attended dummy (higher education 5 1) Interaction term between respondents and parents education levels (base: both respondent and parent high school or less) Respondent post-secondary; parent high school or less dummy Respondent high school or less; parent post-secondary dummy Both respondent and parent post-secondary dummy

61 61 61 61 61 61 61 61 61 61 54 61 52 51 54

0.541 26.410 7.125 0.262

0.502 0 3.909 20 2.090 4 0.444 0 0 2.7 6 0 0 0 0 0 0 0 0

0.501 0 4.023 20 2.187 4 0.499 0 0 2.7 6 0 0 0 0 0 0 0 0

1,425 0.484 0.500 0 1,425 26.284 3.841 20 1,425 7.056 2.063 4 1,425 0.545 0.498 0 1,425 1,425 1,425 1,425 1,425 1,425 1,313 1,379 1,273 1,247 1,277 0.037 0.189 4.720 1.156 6.616 0.356 0.361 0.360 0.122 0.337 0.481 0.480 0.328 0.473 0 2.3 6 0 0 0 0 0 0 0 0

0.115 0.321 4.921 1.351 6.666 0.337 0.361 0.361 0.148 0.259 0.484 0.484 0.358 0.442

0.041 0.199 4.872 1.267 6.624 0.333 0.296 0.429 0.122 0.444 0.459 0.497 0.329 0.500

0.623 0.489 0.288 0.457 0.373 0.488 0.259 0.442

0.691 0.464 0.124 0.331 0.231 0.424 0.171 0.379

0.720 0.449 0.164 0.371 0.281 0.450 0.185 0.388

54 54 54

0.111 0.317 0.167 0.376 0.148 0.359

0 0 0

1 1 1

77 77 77

0.351 0.480 0.078 0.270 0.156 0.365

0 0 0

1 1 1

1,264 1,264 1,264

0.394 0.489 0.051 0.219 0.176 0.381

0 0 0

1 1 1

Note: SD 5 standard deviation.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

Young Adults Employment Status

43

share of the non-employed have more education than their same sex parents. Correspondingly, the proportion of non-employed who are less educated than their same sex parents is high. In light of the fact that more and more people have gone on to post-secondary schools since the end of World War II, the general expectation is that children will exceed their parents academic achievements. A large share of young adults who have left the labor force deviate from this expectation and fail to match, let alone surpass, their same sex parents education levels. 4.3. Estimation Results Estimation results are presented in Table 2. The base model is shown in Estimation 1 and contains only respondents age, gender, and education. The highest education level dummy and the school leaver dummy are estimated to be signicant in the base model. If a person has attended a junior, technical, or four-year college, he or she is unlikely to be either non-employed or unemployed. People who dropped out of the last school they attended have a high probability of becoming non-employed. It is important to point out that these results call for careful consideration and can easily be misinterpreted. For example, the nding that the school leaver dummy is positive and signicant for the non-employed group means only that it is relatively more likely that someone who quit school will become non-employed rather than employed. To nd out which employment status school leavers are most likely to have, the size and signicance of the marginal effects must be calculated.26 Table 2 also presents the marginal effects of such calculations. The highest level of school dummy is estimated to be signicant for the non-employed and employed groups. Attending a post-secondary school reduces the probability of being non-employed by 4.2 percentage points and raises the likelihood of being employed by 6.4 percentage points. In contrast, the school leaver dummy is not signicant for any of the three groups. Interpreting the coefcients and the marginal effects results as a whole shows that whether respondents left school early accounts for some of the difference between the non-employed or employed groups, but it does not demonstrate that leaving school early increases the probability of being non-employed. Genda (2004) and Kosugi (2004) observed that many nonemployed people were early school leavers, but controlling for other individual attributes and educational backgrounds makes it clear that leaving school does not increase ones chances of falling into any particular employment status.27 Estimation 2 adds variables on local unemployment levels, minimum wages, and respondents age cohorts to the base model. In the non-employment group, the school leaver variable continues to be signicant and the minimum wage is signicant as well. Local unemployment levels are also signicant for the non-employed, albeit at a lower level. In terms of marginal effects, for every 1% increase in the local unemployment level, a person has a 0.4 percentage point greater risk of being non-employed, and a 1.2 percentage point lower probability of being employed. It seems obvious that there are more

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

26. For more on how to estimate coefcients and standard errors in multinomial choice models, see Greene (2003). 27. Honda and Hotta (2006) categorize young, non-employed adults into several types and use binomial logit models to assess the factors determining employment status. In only a very small number of estimations was school leaving a signicant variable, which indicates that failing to nish school is not a strong predictor of non-employment. The authors use a different sampling study for their analysis, but their ndings are similar to the results presented in this paper.

Table 2.

Multinomial Probit Models: Employment Status.


Estimation 1 Estimation 2

44 Takehisa SHINOZAKI

Employment Status

Non-Employed

Unemployed

Employed

Non-Employed

Unemployed

Employed

Coefcient (SE) Female dummy (female 5 1) Age Age squared/100 Highest level of school attended dummy (post-secondary 5 1) Dropout dummy (dropout 5 1) Prefectural unemployment rate (%) Prefectural minimum wage (U100) 19721976 Birth year dummy 19771981 Birth year dummy 19821986 Birth year dummy Father blue-collar worker when respondent was age 15 dummy (yes 5 1) Mother employed when respondent was age 15 dummy (yes 5 1) Household income at age 15 dummy (above average51) Fathers highest level of school attended dummy (higher education 5 1) Mothers highest level of school attended dummy (higher education 5 1) Constant term Survey year dummy Sample size Pseudo log-likelihood 0.263# (0.144) 0.176 (0.282) 0.297 (0.517) 0.816*** (0.218) 0.781* (0.359)

Marginal Effect (SE) 0.014 (0.007) 0.010 (0.015) 0.017 (0.027) 0.042*** (0.011) 0.064 (0.042)

Coefcient (SE) 0.011 (0.115) 0.097 (0.228) 0.182 (0.420) 0.322# (0.166) 0.100 (0.321)

Marginal Effect (SE) 0.003 (0.010) 0.010 (0.020) 0.018 (0.037) 0.022 (0.015) 0.000 (0.029)

Marginal Effect (SE) 0.011 (0.013) 0.000 (0.022) 0.002 (0.041) 0.064*** (0.016) 0.064 (0.046)

Coefcient (SE) 0.259# (0.145) 0.210 (0.248) 0.260 (0.438) 0.870*** (0.224) 0.740* (0.364) 0.099# (0.054) 0.343* (0.172) 0.390 (0.373) 0.623 (0.562) 0.877 (0.853)

Marginal Effect (SE) 0.013# (0.007) 0.012 (0.013) 0.015 (0.023) 0.044*** (0.012) 0.058 (0.040) 0.004# (0.002) 0.017# (0.009) 0.021 (0.022) 0.033 (0.037) 0.068 (0.099)

Coefcient (SE) 0.011 (0.117) 0.110 (0.252) 0.245 (0.464) 0.345* (0.169) 0.069 (0.329) 0.098 (0.080) 0.071 (0.213) 0.026 (0.373) 0.298 (0.615) 0.136 (0.799)

Marginal Effect (SE) 0.003 (0.010) 0.011 (0.023) 0.023 (0.042) 0.023 (0.015) 0.002 (0.029) 0.008 (0.007) 0.004 (0.019) 0.001 (0.033) 0.022 (0.059) 0.002 (0.076)

Marginal Effect (SE) 0.010 (0.013) 0.000 (0.022) 0.008 (0.040) 0.068*** (0.016) 0.056 (0.044) 0.012# (0.007) 0.020 (0.017) 0.020 (0.036) 0.055 (0.059) 0.070 (0.095)

5.140 (3.882)

0.827 (3.046) Yes 1,584 606.10

9.290* (4.053)

1.924 (3.947) Yes 1,584 597.07

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

Table 2.

Continued.
Estimation 3 Estimation 4

Employment Status

Non-Employed

Unemployed

Employed

Non-Employed

Unemployed

Employed

Coefcient (SE) Female dummy (female 5 1) Age Age squared/100 Highest level of school attended dummy (post-secondary 5 1) Dropout dummy (dropout 5 1) Prefectural unemployment rate (%) Prefectural minimum wage (U100) 19721976 Birth year dummy 19771981 Birth year dummy 19821986 Birth year dummy Father blue-collar worker when respondent was age 15 dummy (yes 5 1) Mother employed when respondent was age 15 dummy (yes 5 1) Household income at age 15 dummy (above average51) Fathers highest level of school attended dummy (higher education 5 1) Mothers highest level of school attended dummy (higher education 5 1) Constant term Survey year dummy Sample size Pseudo log-likelihood 0.103 (0.190) 0.120 (0.323) 0.127 (0.608) 1.178*** (0.216) 0.368 (0.420) 0.009 (0.070) 0.207 (0.266) 0.536 (0.401) 0.667 (0.637) 0.116 (0.975) 0.367 (0.238) 0.395* (0.184) 0.559** (0.206)

Marginal Effect (SE) 0.005 (0.008) 0.007 (0.013) 0.010 (0.025) 0.058*** (0.012) 0.019 (0.028) 0.001 (0.003) 0.008 (0.011) 0.021 (0.021) 0.023 (0.034) 0.005 (0.036) 0.016# (0.009) 0.016# (0.009) 0.031* (0.014)

Coefcient (SE) 0.146 (0.144) 0.374 (0.258) 0.860# (0.480) 0.026 (0.182) 0.050 (0.376) 0.025 (0.088) 0.119 (0.232) 0.498 (0.419) 1.034 (0.660) 1.222 (0.871) 0.340* (0.167) 0.279 (0.198) 0.153 (0.229)

Marginal Effect (SE) 0.012 (0.011) 0.030 (0.021) 0.067# (0.039) 0.010 (0.014) 0.001 (0.030) 0.002 (0.007) 0.008 (0.018) 0.038 (0.039) 0.090 (0.074) 0.156 (0.164) 0.030# (0.015) 0.020 (0.018) 0.015 (0.015)

Marginal Effect (SE) 0.007 (0.014) 0.022 (0.023) 0.057** (0.043) 0.048 (0.015) 0.020 (0.037) 0.001 (0.007) 0.016 (0.022) 0.059 (0.042) 0.113 (0.072) 0.150 (0.157) 0.014 (0.014) 0.037# (0.019) 0.017 (0.018)

Coefcient (SE) 0.164 (0.155) 0.176 (0.266) 0.079 (0.481) 1.209*** (0.237) 0.682# (0.401) 0.065 (0.058) 0.105 (0.207) 0.486 (0.456) 1.026 (0.723) 1.432 (0.989)

Marginal Effect (SE) 0.007 (0.006) 0.010 (0.011) 0.008 (0.021) 0.060*** (0.014) 0.044 (0.038) 0.003 (0.002) 0.004 (0.009) 0.021 (0.024) 0.050 (0.047) 0.125 (0.147)

Coefcient (SE) 0.050 (0.133) 0.411 (0.277) 0.838 (0.524) 0.184 (0.207) 0.057 (0.359) 0.036 (0.086) 0.126 (0.225) 0.223 (0.404) 0.544 (0.667) 0.294 (0.894)

Marginal Effect (SE) 0.005 (0.010) 0.033 (0.022) 0.066 (0.042) 0.006 (0.016) 0.002 (0.028) 0.002 (0.007) 0.009 (0.018) 0.015 (0.033) 0.038 (0.059) 0.007 (0.076)

Marginal Effect (SE) 0.002 (0.013) 0.023 (0.022) 0.057 (0.042) 0.066** (0.020) 0.043 (0.042) 0.005 (0.006) 0.013 (0.018) 0.036 (0.039) 0.088 (0.077) 0.133 (0.154)

Young Adults Employment Status

6.009 (4.689)

0.095 (4.159) Yes 1,254 434.71

0.501* (0.232) 0.422* (0.210) 8.074# (4.345)

0.026# (0.014) 0.021# (0.012)

0.110 (0.233) 0.053 (0.308) 1.321 (4.067) Yes 1,345 464.38

0.012 (0.018) 0.001 (0.024)

0.014 (0.018) 0.022 (0.030)

45

Note: SE 5 standard error. The base group for estimates was employed people age 1534. Robust SEs are given in parentheses. Signicance notation: #P , 0.1, *P , 0.05, **P , 0.01, ***P , 0.001.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

46

Takehisa SHINOZAKI

unemployed people in areas with higher unemployment and Table 2 shows a marginal effect of an 0.8 percentage point greater likelihood of being unemployed, but the effect is not signicant. For every U100 increase in the minimum wage, the probability of being non-employed rises by 1.7 percentage points. Employers might hire fewer unskilled workers when minimum wages are high, making it more difcult for young people to nd work. Alternatively, if few people are applying for jobs, it may reect that more people are opting out of the labor market instead of being unemployed. Especially when the cost of job hunting is high, giving up the search is a rational reaction. Estimation 3 adds variables related to family backgrounds to Estimation 2. Respondents who stated that their fathers had rank-and-le jobs when the respondents were aged 15 are 1.6 percentage points less likely to be non-employed. Similarly, those reporting that their mothers were employed had a 1.6 percentage point lower probability of being non-employed. Respondents were 3.6 percentage points more likely to be non-employed if their families had higher than average incomes. Genda (2007b) has already demonstrated that current household income levels affect employment outcomes but the results in this paper tell us that past living standards also help to determine peoples employment status. Let us consider the relationship between non-employment and past standards of living from the perspective of market wages and individuals reservation wages (the lowest wage a person will accept to work). Non-employed people do not accumulate human capital as employed people are expected to do and are at high risk of becoming less productive than working people in their age group. Furthermore, because so many of the non-employed have no more than a high school education or left school early, they will not receive the wage premium that is usually given to the more educated. As a result, the non-employed are likely to be offered a wage that is below average. On the other hand, a large share of the non-employed is from families with above-average incomes and may be reluctant to accept a sharp decline in their standard of living. It is quite possible that a more afuent background keeps peoples reservation wages high because they want to maintain their living standard. These high reservation wages are unlikely to be offered given that employers are inclined to pay below average wages to the non-employed for the reasons mentioned above. This wage gap between what is available and what is desired may be keeping the non-employed from seeking employment. Estimation 4 adds variables on the educational backgrounds of respondents parents to Estimation 2. Parents education is positively signicant for the non-employed. How might parents education levels affect their childrens employment outcomes? It could be the case that parents education levels are related to the reservation wages of the non-employed. If a child receives the same amount of education as his/her parent, he/she could expect his/her wages to be augmented by an education premium in the same way as his/her parents were. Conversely, failing to get an education equal to ones parents would put the education premium out of reach. However, people are reluctant to accept a lower standard of living. This reluctance inates the reservation wages of people with limited education. The fact of the matter is that no rm is going to pay high wages to someone with little education. Unable to accept what the job market will offer them, these young people give up the search. The results of using interaction terms to evaluate the mechanism linking parents education with childrens employment status are shown in Table 3. People with more education than their parent of the same sex have a 4.8 percentage point higher probability of being employed, those with less education than their same sex parents are 4.6 percentage points more likely to be non-employed. Looking at the estimation results as a group, we can see that there is almost no overlap in the variables that are signicant to the non-employed and the unemployed which tells us these groups have divergent individual and family attributes. Comparing results of the non-employed with the unemployed, fewer variables are signicant for the latter. The scarcity of signicant variables can be interpreted as evidence that the unemployed are not so different from the employed in their personal and family characteristics.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

Young Adults Employment Status

47

Table 3. Multinomial Probit Models: Effects of Parental and Childs Educational Backgrounds on Childs Employment Status.
Employment Status (5)

Non-Employed

Unemployed

Employed

Coefcient (SE) Female dummy (female 5 1) Age Age squared/100 Dropout dummy (dropout 5 1) Prefectural unemployment rate (%) Prefectural minimum wage (unit 5 U100) 19721976 Birth year dummy 19771981 Birth year dummy 19821986 Birth year dummy Respondent post-secondary; parent high school or less dummy Respondent high school or less; parent post-secondary dummy Both respondent and parent, post-secondary dummy Constant Survey year dummy Sample size Pseudo log-likelihood 0.212 (0.152) 0.108 (0.254) 0.045 (0.457) 0.674# (0.399) 0.063 (0.057) 0.154 (0.204) 0.527 (0.448) 1.079 (0.715) 1.432 (0.990) 1.158*** (0.224) 0.736** (0.254) 0.376 (0.322) 7.510 (4.315)

Marginal Effect (SE) 0.009 (0.006) 0.007 (0.011) 0.003 (0.020) 0.044 (0.038) 0.002 (0.002) 0.006 (0.009) 0.024 (0.024) 0.055 (0.048) 0.127 (0.148) 0.044*** (0.008) 0.046# (0.023) 0.013 (0.009)

Coefcient (SE) 0.053 (0.134) 0.387 (0.279) 0.794 (0.528) 0.033 (0.366) 0.041 (0.084) 0.089 (0.220) 0.203 (0.405) 0.534 (0.667) 0.265 (0.906) 0.142 (0.189) 0.328 (0.416) 0.198 (0.274) 1.182 (4.124) Yes 1,345 464.91

Marginal Effect (SE) 0.005 (0.010) 0.031 (0.022) 0.061 (0.042) 0.003 (0.028) 0.003 (0.007) 0.006 (0.017) 0.013 (0.033) 0.036 (0.059) 0.005 (0.075) 0.005 (0.014) 0.022 (0.041) 0.013 (0.018)

Marginal Effect (SE) 0.004 (0.013) 0.024 (0.022) 0.059 (0.041) 0.041 (0.041) 0.005 (0.006) 0.012 (0.018) 0.036 (0.039) 0.091 (0.077) 0.132 (0.155) 0.048** (0.014) 0.068 (0.048) 0.025 (0.021)

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

Note: SE 5 standard error. The base group for estimates was employed people age 1534. Robust SEs are given in parentheses. Signicance notation: #P , 0.1, *P , 0.05, **P , 0.01, ***P , 0.001.

48

Takehisa SHINOZAKI

If this interpretation is correct, policies designed to help unemployed young people to nd work will not be of much benet to non-employed. This distinct group requires a different approach that addresses their reasons for leaving the workforce.28 Because the aforementioned gap between market and reservation wages grows wider as people grow older, we need to create an environment in which young people can get support right after they graduate or leave school.29

5. Conclusion
This paper has drawn on data from six years of Japan General Social Surveys to evaluate the links between the employment status of young people in Japan and their local employment environments and family backgrounds. Quantitative analysis using multinomial probit models and descriptive statistics revealed that young people out of work do not belong in a single category. Young people who are not employed, in school, or looking for work have different personal traits and family backgrounds than their unemployed peers who persevere in their job searches. Especially noteworthy is how the chances of being non-employed are increased by having an above-average standard of living in the past and by not matching the educational achievements of ones same sex parent. The analysis also conrmed that there were few differences between unemployed and employed people in terms of their individual and household attributes. These ndings suggest that one of the reasons jobless people stop actively seeking work may be their dissatisfaction with the wages being offered. Non-employed people who enjoyed a high standard of living in the past may be trying to hold on to that standard and refuse to take low-wage jobs that would not support a middle-class lifestyle. However, the supply of higher salary jobs in Japan has shrunk since the 1990s and regular full-time employment is harder to nd. A growing percentage of workers have had to accept temporary employment with wages well below that of regular full-time employees. If rms regard their increased reliance on irregular workers as more than a short-term response to economic recession and continue to rely on them to minimize costs after the economy recovers, the types of jobs that the non-employed are holding out for will only grow scarcer, and bringing them back into the workforce will become more difcult. Substantially increasing the number of regular full-time jobs is not a pipe dream, presumably, but raising the private sectors demand for labor seems far-fetched. What the government could do is try to change the supply of labor by addressing the factors that lead young people to leave the workforce and thus push them toward re-engaging with the labor market. For example, if the non-employed can be convinced to moderate their expectations (e.g. by lowering their reservation wages), more of them will move into employment. However, creating a government program that would somehow persuade individuals to change their preferences on wages and other work conditions seems unfeasible as well.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

28. Although it is important to offer programs that t the specic needs of the non-employed and the unemployed, at the same time these programs should not operate in isolation. The policy responses to joblessness need to be coordinated and to provide continuity. Miyamoto (2004) pointed out the limitations of using the motivation to work to distinguish between the unemployed and the non-employed and the need for policies that are more effective at moving people back into the labor force. 29. According to Miyamoto (2004), while Japans young adult employment policies are targeted at people in their mid- to late 20s, European nations center their employment promotion efforts on people from their mid-teens to mid-20s.

Young Adults Employment Status

49

Should the government create some sort of counseling program to bring the expectations of nonemployed people into line with what the market has on offer, just reaching these people would present a real obstacle. After nishing or leaving school, non-employed people tend to become quite isolated and have few people outside of their families to talk to about working. Many of them occupy a narrow social space beyond the governments reach.30 However, whatever the obstacles to changing young peoples preferences may be, we need employment support programs that provide continuity in the process of guiding people into the labor force, whether those programs are based in their schools or their communities. Another approach to the problem of the widening gap between the reservation wages of the young non-employed and the market wages being offered to them would be to supplement the market wages with government subsidies. The US and other nations have introduced such programs where the subsidies take the form of individual tax credits for low-income workers. In Japan, a wage subsidy program does not seem achievable in the short term for both technical and scal reasons. On the technical side, Japan has yet to institute a taxpayer identication system that would make it possible to deliver social benet payments through the tax system. As to the scal side, Japans extensive public debt would make securing the necessary funds a very steep uphill battle. The government could also narrow the gap between market wages and the non-employeds reservation wages by instituting measures that would raise their human capital. In 2008, the government launched the Job Skills Acquisition System, also known as the Job Card system for the certicates that trainees receive listing the types of training they have completed. The Job Card system is called a dual system because its programs combine classroom instruction with work site training. There are several programs in the dual system with varying degrees of private sector participation and divisions of time between on- and off-site instructions. In the Practical Human Resource Development Syssei Shisutemu), companies apply to place new employees in off-the-job traintem (Jissengata Jinzai Yo ing programs. Another program gives government subsidies to rms to cover most of the wages and  Jisshugata  educational costs of xed-term trainees (Yuki Kunren). In addition to young adults, Job Card vocational programs are also open to single mothers and others who have had limited opportunities to work. In 2008, roughly 33,000 people participated in the Job Card system. For young people thought not to be ready to participate in the Job Card programs, other options also offer employment assistance and advice. The Job Cafe such as Hello Work and the Job Cafe program began in 2004 with the goal of providing one-stop service for job seekers. In 2006, local Support Stations were created as outreach centers to deliver necessary employment-related services to young people. It is important to evaluate the effectiveness of these initiatives, but as they were launched fairly recently, we will need to give them more time before beginning our assessments. The ndings of this paper are offered with certain reservations stemming from the following points requiring further study. First, the robustness of this analysis needs to be conrmed. The data used in this analysis from six years of JGSS surveys ensured a sufciently large sample size, but it would be useful to see if applying the same factors of analysis to a different large data set would result in similar

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

.........................................................................................................................................................................................................................

30. The Japan Institute for Labour Policy and Training (2004) presents details on the job search environments faced by young unemployed and non-employed people. For further information on the widely discussed one-stop service approach to facilitating employment, see, for example, Takahashi (2005).

50

Takehisa SHINOZAKI

outcomes. Second, although this paper found that the characteristics of the unemployed were closer to the employed than the non-employed, it would be useful to examine in more detailagain using a large data sourcewhat distinguishes the unemployed from the non-employed. Third, whether or not someone left school early was used above as an independent variable, but school leaving merits further investigation as a dependent variable to clarify what factors push students to truncate their educations.31 Finally, this paper discussed how high reservation wages might partly explain why people abandon the job market, but this possibility needs further exploration using direct interviews to uncover rst of all how high the reservation wages of the non-employed are and also to see what aspects of work matter the most to them.

References
Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013
Bynner, John and Samantha Parsons. 2002. Social Exclusion and the Transition from School to Work: The Case of Young People Not in Education, Employment, or Training (NEET). Journal of Vocational Behavior 60(2): 289309. Card, David and Alan B. Krueger. 1994. Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania. American Economic Review 84(4): 772793. . 2000. Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Reply. American Economic Review 90(5): 13971420. Cheng, Simon and J. Scott Long. 2007. Testing for IIA in the Multinomial Logit Model. Sociological Methods & Research 35(4): 583600. Coles, Bob, et al. 2002. Literature Review of the Costs of Being Not in Education, Employment or Training at Age 1618. Department for Education and Skills Research Report 347. London: Queens Printer. Available at: https://www.education. gov.uk/publications/eOrderingDownload/RR347.pdf (accessed 10 December 2011). Fry, Tim R. L. and Mark N. Harris. 1998. Testing for Independence of Irrelevant Alternatives: Some Empirical Results. Sociological Methods & Research 26(3): 401423. Furlong, Andy. 2006a. Not a Very NEET Solution: Representing Problematic Labour Market Transitions among Early School-Leavers Work, Employment and Society 20(3): 553569. . 2006b. NEET. In Fuantei o Ikiru Wakamono Tachi (Youths Living in Vulnerability), ed. Akio Inui. Tokyo: Otsuki Shoten: 69115. Gaston, Noel and Tomoko Kishi. 2005. Labour Market Policy Developments in Japan: Following an Australian Lead? Australian Economic Review 38(4): 389404. do  Kenkyu  Zasshi (Japanese Journal of Labour Genda, Yuji. 1997. Chansu wa Ichido (You Only Get One Chance). Nihon Ro Studies) 449: 212. . 2001. Shigoto no Naka no Aimai na Fuan (A Nagging Sense of Job Insecurity). Tokyo: Chuokoron Shinsha. . 2004. NEET to Iu Wakamono (Young People Who Are So-Called NEET). In NEET, eds. Yuji Genda and Mie Maganuma. Tokyo: Gentosha. . 2005a. A Nagging Sense of Job Insecurity: The New Reality Facing Japanese Youth. Tokyo: International House of Japan. sha no Jitsujo  (Current Status of Jobless Youths). In Seisho nen no Shur  o  ni Kansuru Kenkyu  . 2005b. Jakunen Mugyo sa (Research of Youth Labor), ed. Cabinet Ofce. Tokyo: Cabinet Ofce. Cho . 2007a. Jobless Youths and the NEET Problem in Japan Social Science Japan Journal 10(1): 2340.

.........................................................................................................................................................................................................................

31. Mitani (2001b) and Takahashi and Genda (2004) are among the small number of quantitative studies of early school leaving in Japan.

Young Adults Employment Status

51

 no Keizaigakuteki Saikento  (Economic Backgrounds of Non-Employed Youths in Japan). . 2007b. Jakunen Mugyo do  Kenkyu  Zasshi (Japanese Journal of Labour Studies) 567: 97112. Nihon Ro  Taisaku to shite no 14 Sai no Shu  Taiken Shien (A Policy Genda, Yuji and Daisaku Okada. 2004. Jakunen Shu  gyo  gyo Instrument for Motivating Young People Who Are Not in Education, Employment or Training). ESRI Discussion Paper Series 100. Tokyo: Economic and Social Research Institute. Genda, Yuji, et al. 2010. Long-Term Effects of a Recession at Labor Market Entry in Japan and the United States. Journal of Human Resources 45(1): 157196. Godfrey, Christine, et al. 2002. Estimating the Cost of Being Not in Education, Employment or Training at Age 1618. Department for Education and Skills Research Report 346. London: Queens Printer. Available at: https://www.education. gov.uk/publications/eOrderingDownload/RR347.pdf (accessed 10 December 2011). Greene, William H. 2003. Econometric Analysis, 5th ed. Upper Saddle River, NJ: Prentice Hall. sha no Jitsuzo  (The Reality of Jobless Youth in Japan). Nihon Ro do  Honda, Yuki and Satoko Hotta. 2006. Jakunen Mugyo  Zasshi (Japanese Journal of Labour Studies) 556: 92105. Kenkyu

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

 riteki Shu kan Keisei no Riron (Theory of Rational Habit Formation). In Gendai Keizaigaku no Ikeda, Shinsuke. 2003. Go ryu  2003 (Trends in Modern Economics 2003), eds. Yoshiyasu Ono. et al. Tokyo: Toyo Keizai Inc: 59107. Cho  to Sono Taisaku (Current Status of NEET and Measures for Them). Kokuritsu Kokkai Ito, Masayuki. 2006. NEET no Genjo kan Cho sa to Jo ho  (National Diet Library Issue Brief): 520. Tokyo: National Diet Library: 536. Tosho  no Kiki ni aru Wakamono no Jitsuzo  (Young People in Japan Institute for Labour Policy and Training (JILPT). 2004. Iko Transitional Crisis), JILPT Research Report 6. Tokyo: Japan Institute for Labour Policy and Training. Available at: http:// www.jil.go.jp/institute/reports/2004/006.html (accessed 10 December 2011).  ko  (Research Trends on Part-time Workers and  Do Kameyama, Toshiro. 2006. Furita- NEET o Meguru Kenkyu  Kenkyu  (Buraku Liberation Research) 170: 8388. NEETs). Buraku Kaiho do  (Educational Background, Disparity and Inequality). Tokyo: University of Kikkawa, Toru. 2006. Gakureki to Kakusa, Fubyo Tokyo Press. Kondo, Ayako. 2007. Does the First Job Really Matter? State Dependency in Employment Status in Japan. Journal of the Japanese and International Economies 21(3): 379402.  sha Zo ka no Jittai to Haikei (The Increase and Background of Jobless Youth). Nihon Kosugi, Reiko. 2004. Jakunen Mugyo Rodo Kenkyu Zasshi (Japanese Journal of Labour Studies) 533: 116. Available at: http://www.jil.go.jp/institute/zassi/ backnumber/2004/12/pdf/004-016.pdf (accessed 10 December 2011).  no Genjo  to Shien no Kadai (Current Status of Youth Labor and Challenges for gyo . 2008. Wakamono no Shu Support). Business Labor Trend April: 212. . 2010. Wakamono to Shoki Career (Young People and Early Career). Tokyo: Keiso Shobo.  kara Shokugyo  e no Iko  wo Shien suru Shokikan e no Hiaringu Cho  sa Kekka Kosugi, Reiko and Yukie Hori. 2003. Gakko (Hearing Investigation on School-to-Work Transition). JIL Discussion Paper Series 03001. Tokyo: Japan Institute of Labor.  , Fushu  to Shu  Keitai (Employment, Unemployment and Work Status of Youths).  gyo  gyo gyo 2005. Wakamono no Shu  o  Shien no Genjo  to Kadai (Current Status of Youth Labor and Challenge for Support), ed. Japan In Wakamono Shugy Institute for Labour Policy and Training. Tokyo: Japan Institute for Labour Policy and Training: 79135. Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables. Thousand Oaks, CA: Sage.  kokusho (Research Report Ministry of Health, Labour and Welfare (MHLW). 2003. Jakunensha Kyaria- Shien Kenkyukai Ho of Career Support for Young People). Available at http://www.mhlw.go.jp/houdou/2003/09/h0919-5e.html (accessed 10 June 2011). . 2005. White Paper on the Labour Economy. Tokyo: Nikkei Printing Inc. . 2010. White Paper on the Labour Economy. Tokyo: Nikkei Printing Inc.  do  Shijo  no Ko  zo  Henka to Ko  yo  Seisaku (Structural Changes in the Youth Labour Mitani, Naoki. 2001a. Jakunen Ro do  Kenkyu  Zasshi (Japanese Journal of Labour Studies) 490: 1932. Market and Employment Policies). Nihon Ro ki Fukyo  to Jakunen Shitsugyo  (Long-Term Recession and Youth Unemployment). Kokumin Keizai . 2001b. Cho Zasshi (Journal of Economics and Business Administration) 183(5): 4562.  (The Socially Excluded and Jobless Youth). Nihon Ro do  Miyamoto, Michiko. 2004. Shakaiteki Haijo to Jakunen Mugyo  Zasshi (Japanese Journal of Labour Studies) 533: 1726. Kenkyu

52

Takehisa SHINOZAKI

Neumark, David and William Wascher. 2000. Minimum Wages and Employment: A Case Study of the Fast-Food Industry in New Jersey and Pennsylvania: Comment American Economic Review 90(5): 13621396. . 2006. Minimum Wages and Employment: A Review of Evidence from the New Minimum Wage Research. NBER Working Paper 12663. Cambridge, MA: National Bureau of Economic Research. Available at: http://www.economics. uci.edu/les/economics/docs/workingpapers/2006-07/Neumark-08.pdf (accessed 10 December 2011). Ryan, Paul. 2001. The School-to-Work Transition: A Cross-National Perspective. Journal of Economic Literature 39(1): 3492. Tanaka, Ryuichi. 2008. The Gender-Asymmetric Effect of Working Mothers on Childrens Education: Evidence from Japan. Journal of the Japanese and International Economies 22(4): 586604. OECD. 2000. Education at a Glance. Paris: OECD Publishing. . 2006. Employment Outlook. Paris: OECD Publishing. . 2009. Education at a Glance. Paris: OECD Publishing.  Mondai (Regional Characteristics of the Japanese Youth Labour Ohta, Souichi. 2005. Chiiki no Naka no Jakunen Koyo do  Kenkyu  Zasshi (Japanese Journal of Labour Studies) 539: 1733. Market). Nihon Ro  do  Shijo  ni Okeru Sedai Ko  ka (Cohort Effects in the Labor Market). In Gendai Ohtake, Fumio and Takenori Inoki. 1997. Ro kyo  Daigaku Shuppankai. Makuro Keizai Bunseki (Modern Macro-economic Analysis), ed. Kazumi Asako, et al. Tokyo: To Okita, Toshie. 2003. Shakaiteki Haijo e no Ninshiki to Atarashii Torikumi (Recognition of Social Exclusion and New  o  Shien Seisaku no Tenkai (Development of Policy Supports for Youths in Efforts). In Shogaikoku no Wakamono Shugy Foreign Countries), ed. Japan Institute of Labour. Tokyo: Japan Institute of Labour. Social Exclusion Unit. 1999. Bridging the Gap: New Opportunities for 1618 Year Olds Not in Education, Employment or Training. London: The Stationery Ofce.  Shien to shite no Job Cafe no Genjo  (Investigation on Job Cafe as Career Takahashi, Yoko. 2005. Jichitai ni yoru Shu  gyo do  Kenkyu  Zasshi (Japanese Journal of Labour Studies) 539: 5667. Support by Local Government). Nihon Ro  ko  Chu do  Shijo  (Labor Market for Junior High  gakusotsu, Ko tai to Ro Takahashi, Yoko and Yuji Genda. 2004. Chu  ( Journal of Social Science) 55(2): 2949. Graduates and High School Dropouts). Shakai Kagaku Kenkyu Willis, Paul. 1981. Learning to Labor: How Working Class Kids Get Working Class Jobs. New York: Columbia University Press.  fuken de-ta- o Mochiita Chiiki Ro  do  Shijo  no Bunseki (Analysis of Static Differentials in Yugami, Kazufumi. 2005. Todo do  Kenkyu  Zasshi (Japanese Journal of Labour Unemployment and Non-Employment Rates by Prefecture). Nihon Ro Studies) 539: 416.

Downloaded from http://ssjj.oxfordjournals.org/ by Min Thu on September 4, 2013

You might also like