- This topic has 0 replies, 1 voice, and was last updated 2 years, 8 months ago by
jbadmin.
- Post
Self-Employment in Japan: A Microanalysis of
Personal Profiles
Jess DIAMOND and Ulrike SCHAEDE*
We use unique data from the private Keio Household Panel Survey to explore profiles of the self-employed
in Japan, separately for men and women. We analyze labor market conditions at the time of work force entry
as well as personal markers such as age, education, work experience, assets, and family situation. The most
persistent finding is that Japan’s labor market continues to be ‘sticky’. For the period 1963–2004, those in
self-employment tended to be less educated, older, and less likely to have gathered prior experience through
job-hopping. Having a self-employed father also loomed large. In contrast, for 2004–2007, younger age together
with assets and possibly young children are associated with self-employment. While the profiles for women with
only a high school degree and all men are similar , results for women with a college degree differ consistently,
suggesting that educated females face a different choice set in entering self-employment. Finally, we support
the determinative nature of Japan’s job market: entering the labor force at a time of weak local labor market
conditions significantly increases the odds of becoming self-employed and remaining so in the long run.
1. Introduction
Little is known about the self-employed in Japan of today. Most of our existing knowledge is
based on government surveys, and the prevailing image of the stereotypical self-employed, if
there is one, is based on research conducted in the 20th century. In a seminal study of familyrun businesses (with fewer than 30 workers) in the 1970s, Patrick and Rohlen (1987) found
great heterogeneity, with a multitude of backgrounds and motivations among the self-employed.
Insofar as a modal picture emerged, it was that of an elderly couple running a home-front store.
This was confirmed by Brinton (1993) who conducted a targeted survey of urban households
in 1984 and found that the self-employed were older, and while a quarter of self-employed men
worked in retail, women often did home piecework in manufacturing. In a study using a 1975
Social Stratification and Mobility Survey of 2,724 men and looking at socio-structural factors
such as family background and individual attributes, Cheng (1997) analyzed the determinants
of switching into self-employment in Japan between 1930 and 1975, a tumultuous period that
includes periods of active entrepreneurship just before and after World War II. She found that
in 20th century Japan, the father’s status as self-employed was a strong predictor, as were prior
Jess DIAMOND is an economist at the Bank of Japan, Institute for Monetary and Economic Studies. He can be reached by
e-mail at jess.diamond@boj.or.jp.
Ulrike SCHAEDE is Professor of Japanese Business at the University of California, San Diego. Her research interests include
Japanese business strategy and management practices, regulation, governance and entrepreneurship. She is the author of
Choose and Focus: Japanese Business Practices for the 21st Century (Cornell UP 2008) as well as books and papers on business policy, small firms, government-business relations, and antitrust. She can be reached by e-mail at uschaede@ucsd.edu.
*We thank the Global COE Program at Keio and Kyoto Universities for the provision of data. We are grateful for helpful
comments from Takeo Hoshi, Hugh Patrick, and Tom Roehl; participants at the IR/PS Japan Seminar at UC San Diego and
the 2010 conference of the Stanford Program on Japanese Entrepreneurship; and three anonymous reviewers.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
2 Jess DIAMOND and Ulrike SCHAEDE
self-employment or employment in a small firm, and a longer duration at the previous job (perhaps correlated with age).
More recently, government data suggest that the percentage of self-employed in Japan’s workforce has declined from roughly 16% to 9% between 1983 and 2009.1 One reason may be a phasing out of the mom-and-pop stores that were so prevalent in the 1970s. To explore this trend,
Genda (2004) and Masuda (2006) looked at labor market conditions and macroeconomic factors
to identify factors that determine self-employment. Their work confirms that Japan’s labor market
remains very sticky (with limited job switching), but also that a deterioration in job opportunities has pushed more people into self-employment, including the young. Yet, using government
surveys on labor and income, Genda and Kambayashi (2002) found a relative decline in estimated
income growth from self-employment compared to other work in the 1990s, especially for the
middle-aged, which might have contributed to lower rates of self-employment. Most of this work
is macro in nature. We are unaware of any recent study that looks at the personal profiles of the
self-employed in Japan.
Learning more about the characteristics of the self-employed in Japan has gained new relevance
with the growing pluralism in business and society since the 1990s. However, studying this subject
is fraught with many difficulties, beginning with the definition. ‘Self-employment’ covers a wide
range of activities, from small shopkeepers or low-level suppliers, to doctors, attorneys, and aspiring
entrepreneurs such as in the information technology sector. Statistics may also differ depending on
whether a founder of a very small company is considered self-employed or an employee of his or her
own company. Moreover, government surveys face the challenge of potential underreporting, either
of the activity itself or of the income. And finally, while we may have some intuition about Japan’s
self-employed men, our image of self-employed women may be fuzzier still, given that their background and motivation to enter self-employment are very different.
In this paper, we use a new, private survey conducted by the Global Centers of Excellence (COE)
Program at Keio and Kyoto Universities, referred to as the ‘Keio Household Panel Survey’ (KHPS).
This rich survey contains a retroactive panel of 4,005 households with basic entries on employment
history, as well as a more detailed panel including personal data that affords us an updated look at
personal profiles. Analytically, we look at those respondents that have switched into self-employment. From the retrospective responses we can draw a picture of general conditions that may have
determined this choice between the 1960s and the mid-2000s, covering previous employment,
education, and the economic environment. For the years 2004–2007, we have more detailed data
that include markers such as household composition, employment, income and expenditures, assets
and home ownership, marital status, and education. This allows us to analyze the profiles of new
entrants into self-employment in recent years. Our data are unique in that they offer these variables
separately for men and women.
We begin in Section 2 with an introduction to the data and summary statistics that offer a first
impression of differences between regular employees and self-employed as of 2004. In Section 3 we
formulate seven hypotheses based on a selective literature review, anchored mostly in labor economics. Section 4 contains the data analyses and discusses the findings. Educated women reveal a different
pattern from both men and less educated women. Moreover, overall we find persistent stickiness in
the labor market and that poor local labor market conditions at the time of labor force entry increase
the likelihood of entering as a self-employed. In Section 5 we explore this stickiness and calculate the
1. Japan’s labor force data can be found at http://www.stat.go.jp/english/data/roudou/lngindex.htm, item 8(4).
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 3
likelihood of being long-term self-employed if one enters self-employment at a young age. Section
6 concludes our analysis.
2. Data and Overview Statistics
2.1 The Dataset
The KHPS is an annual non-government micro level survey, and so far results are available for surveys
conducted between 2004 and 2007. This Japanese-language survey was first translated and coded for
data analysis by Diamond (2009, 2011), and extended for this study to include additional variables.
The first year of the survey, conducted in 2004, covered 4,005 individuals and included a retrospective panel about respondents’ educational and occupational histories from age 15 until 2004. The
response rate for the first year was 29.8%. Over the following three years, new questions were added
(e.g. about hobbies, time spent commuting, etc.). Attrition has reduced the number of respondents
to 3,314 (in 2005), 2,887 (in 2006) and 2,634 (in 2007).2 We have tested the data and find that
attrition has been random and does not seem to affect the balance of representation for the questions
we pursue. Because we use both the retrospective panel and the detailed data 2004–2007, missing
data for certain variables explain why we have a changing number of observations for each analysis.
Our analysis cuts off at age 60; i.e. we exclude cases where a person over 60 switched into selfemployed. While perhaps unfortunate for the overall picture, this truncation is necessary because of
the specifics of Japan’s labor market, where until recently the retirement age was set at the relatively
young age of between 55 and 60, and subsequent employment options were mostly limited to nonstandard or self-employment. Thus, the motivations to switch at that age must be analyzed separately
from switching into self-employment earlier in life, and such analysis is left for future research.
In terms of geographical distribution, of the initial 4,005 respondents, 945 lived in the largest 14
cities (of whom 498 in the Kanto area), 2,280 lived in other cities, and 780 in villages. In terms of
profession, Table 1 shows a breakdown into type of work of the 2,555 respondents that were in the
labor force as of 2004. The remaining 1,450 respondents are not included in this study, because they
were either not in the labor force (575), above the age of 60 (812), or their employment status could
not be determined (63). For this paper, we define the self-employed narrowly to comprise only the
338 jieigyō-nushi and jiyūgyō-sha (lit.: ‘business owner’ and ‘self-business person’, i.e. professionals).
Thus, we exclude the categories ‘working from home on a short-term contract basis’, ‘contractors’,
and ‘working in a family business’. Arguably these could also be considered ‘self-employed’. We
conducted our analyses with exclusion and inclusion of these three categories, and we also repeated
our analyses with only the ‘business owners’, i.e. excluding professionals such as doctors, attorneys
or lawyers. With a few minor changes for the male sample regarding age, we found no material differences in results.
Using our narrow definition, the percentage of self-employed in our sample in 2004 is 14% (338
of 2,462). This number is substantially higher than the 9% reported in the government’s Labor Force
Survey but similar to estimates by the Organisation for Economic Co-operation and Development
(OECD) of 14%.3 The difference may be due to self-definition, because the government survey categories are narrower than how people may define themselves, and the Keio survey allows multiple
2. For further detail on the KHPS project, see http://www.pdrc.keio.ac.jp/open/.
3. See, e.g. http://www.photius.com/rankings/self_employment_by_oecd_country_2008.html.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
4 Jess DIAMOND and Ulrike SCHAEDE
entries for employment. Given the well-chronicled challenges of tax and pension payment collection
from the self-employed in Japan (e.g. Patrick and Rohlen 1987, Soos 1990), it is also possible that
some respondents withheld more information in the government survey than in the private one.4
Importantly, these definitions do not change within our survey data, and we believe that the discrepancy does not affect our overall results.
In our analyses we will focus on people that switch into self-employment. Table 2 introduces
calculations of switching across employment categories between 2004 and 2007. Each column represents the employment category at time t+1, whereas each row shows the employment category
in the previous year t. For example, of 3,006 total observations of regular employees in those three
years, 2,735 (91%) remained in regular employment the following year, while 123 switched to a nonstandard job, 45 became unemployed, 72 retired, and 28 opened their own shop. Likewise, of the
966 observations of self-employed, 793 (82%) remained in self-employment the following year, while
24 (or about 2.5%) moved into regular employment, 49 (5.1%) entered non-standard employment,
30 (3.1%) left the labor force altogether, and 5 (0.5%) reported being unemployed. This Table highlights two important aspects: It is consistent with general notions of high stickiness in Japan’s labor
market, yet it also shows that switching does occur, including into regular employment.
2.2 Overview Statistics: Differences between Employed and Self-Employed
Table 3 introduces summary statistics by way of a comparison of regular workers with the selfemployed in our sample as of 2004.5 The self-employed are, on average, somewhat older than the
Table 1. Distribution of Survey Respondents into Work Categories, as of 2004.
Men Women Total
Self-employed (jieigyō-nushi) 215 82 297
Self-employed (jiyūgyō-sha) 28 13 41
Subtotal 243 95 338
Home job on contract basis (kaisha
to kōyō kankei no nai zaitaku rōdō)
9 31 40
Contractor (itaku rōdō) 33 34 67
Family business (kazokujū-gyōsha) 32 88 120
Subtotal 74 153 227
Regular workers 1,008 317 1,325
Non-standard workers 120 452 572
Subtotal 1,128 769 1,897
Unemployed 48 45 93
Total 1,493 1,062 2,555
4. Patrick and Rohlen (1987) report that in the 1980s only 40% of the heads of non-farm unincorporated enterprises paid
income tax; they argue this was a tacit government subsidy toward small firms. Soos (1990) relays the popular phrase of
ku-ro-yon (‘9-6-4’), referring to an estimation that tax authorities capture 90% of wage and salary income, 60% of selfemployed income, and 40% of farm income. Tax collection measures have been revised since, but the potential challenge
of uneven reporting across these groups persists.
5. Due to typesetting challenges, we do not report summary statistics for non-standard employees here. Readers interested
in these data are encouraged to contact the authors.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 5
employed. Our data confirm the gendered structure of Japan’s labor market. While we do not report
non-standard employment data here, of all non-standard workers in our sample, 79% are women,
whereas the proportion is 28% in self-employment and 24% in regular employment. We see no
marked difference in location (city vs. rural), but even though more self-employed own their homes
and report higher fixed assets (as evaluated for tax purposes), they have, on average, lower savings
and higher debts. This higher debt number may be explained by the fact that small business owners
in Japan often take out business loans in their personal name, sometimes using their own house as
collateral. This is confirmed by the next three entries, which show that while there appear to be no
marked differences in using debt for purchasing a house or consumer durables, for the self-employed
the numbers for ‘debt for investment in non-housing real assets’ are considerably higher. While the
survey did not ask about business debt specifically, investment in non-housing real assets can be considered a reasonable proxy for such debt.
The self-employed report slightly lower education levels than regular employees (though these are
similar to non-standard workers): 52% indicated high school as their highest level of education, as
opposed to 46% for regular employees. An additional 34% of self-employed reported college as the
highest degree, compared with 48% for regular employees (this includes junior and specialized colleges). Note that we break these degrees out for preciseness; they should be summed for the overall
impression. That is, 94% of regular employees have graduated from high school, compared with 86%
for the self-employed. This confirms Patrick and Rohlen’s (1987) observation of comparatively high
education of Japan’s self-employed. While the Table does not show this, our sample includes 583
women that completed only high school (of whom 50 in self-employment), and 424 that graduated
from junior college or college (of whom 35 in self-employment). As our analysis will reveal, this last
category has special characteristics in several variables.
A larger proportion of the self-employed is married. They have more years of work experience
and have been in their current positions longer than the salaried workers (expressed in ‘tenure’).
Table 2. Switching Patterns between Type of Employment, 2004–2007.
T
T+1
Regular
Employee
NonStandard
Employee Unemployed
Out of
Labor
Force
SelfEmployed
Family
Business Total
Regular
employee
2,735 123 45 72 28 3 3,006
Non-standard
employee
131 1,443 47 174 51 10 1,856
Unemployed 25 74 48 45 9 2 203
Out of labor
force
48 190 35 1,778 36 34 2,121
Self-employed 24 49 5 30 793 65 966
Family
business
6 18 2 28 60 256 370
Total 2,969 1,897 182 2,127 977 370 8,522
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
6 Jess DIAMOND and Ulrike SCHAEDE Table 3. Summary Statistics for Regular Employees and Self-Employed, as of 2004. Regular Employees Self-employed Obs Mean Std.Dev Min Max Obs Mean Std.Dev Min Max Age 1,325 41.13 10.60 20.05 59.99 338 45.78 9.75 22.28 59.94 Lives in a large city 1,325 0.23 0.42 0 1 338 0.23 0.42 0 1 Lives in Kanto region 1,325 0.13 0.34 0 1 338 0.10 0.31 0 1
Owns their home 1,325 0.71 0.45 0 1 338 0.80 0.40 0 1
Female 1,325 0.24 0.43 0 1 338 0.28 0.45 0 1
Household savings
(in ¥10,000)
1,118 526.48 849.83 0 9,000 277 471.91 974.93 0 8,000
Market value of household
stockholdings (in¥10,000)
1,205 71.68 371.41 0 5,650 298 52.79 265.16 0 3,800
Household debt
(in ¥10,000)
1,272 634.85 1,880.55 0 50,000 313 851.57 1,654.86 0 12,000
Fixed asset evaluation
(in ¥10,000)
843 342.99 716.98 0 10,000 183 378.90 729.75 0 4,251
Debt to purchase house 1,140 0.37 0.48 0 1 261 0.38 0.49 0 1
Debt to purchase durables 868 0.17 0.38 0 1 198 0.18 0.39 0 1
Debt for investment in nonhousing real assets 742 0.03 0.17 0 1 176 0.08 0.27 0 1
Highest education is high
school
1,316 0.46 0.50 0 1 337 0.52 0.50 0 1
Highest education is college 1,316 0.48 0.50 0 1 336 0.34 0.47 0 1
Married 1,325 0.70 0.46 0 1 338 0.76 0.43 0 1
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles
7
Table 3. Continued.
Regular Employees Self-employed
Obs Mean Std.Dev Min Max Obs Mean Std.Dev Min Max
Tenure 1,256 13.36 10.78 0 43.83 290 16.29 11.53 0 44.00
Total years of work
experience
1,325 20.44 11.02 0 43 338 24.82 10.87 0 43
Days worked in an average
month
1,276 21.79 3.84 4 30 301 22.93 5.90 4 31
Hours worked in an average
week
1,256 49.55 13.89 2 150 291 46.13 25.99 3 144
Monthly wage(in ¥1,000) 1,102 338.06 189.00 16 3,800 155 378.75 694.43 25 8,000
Daily wage(in ¥) 41 9,453.18 2,652.83 4,835 16,000 22 14,775.00 6,434.43 750 30,000
Hourly wage(in ¥) 10 833.00 415.14 600 2,000 14 1,287.86 1,790.94 600 7,500
Annual wage(in ¥10,000) 44 800.66 363.59 180 1,600 49 471.92 589.79 40 4,000
Mining, fishing and forestry 1,325 0.01 0.09 0 1 338 0.11 0.31 0 1
Construction 1,325 0.11 0.31 0 1 338 0.12 0.33 0 1
Manufacturing 1,325 0.22 0.41 0 1 338 0.08 0.28 0 1
Transport and
communications
1,325 0.11 0.31 0 1 338 0.03 0.17 0 1
Retail, wholesale, food and
lodging
1,325 0.11 0.32 0 1 338 0.29 0.45 0 1
Finance, insurance and real
estate
1,325 0.06 0.23 0 1 338 0.02 0.15 0 1
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
8 Jess DIAMOND and Ulrike SCHAEDE
Obviously, all of these aspects may be related to a higher average age, and we will explore this below.
Reported working days per month are about the same for both groups (and both are higher, by five
days, than the average for non-standard workers). Interestingly, regular employees report longer
working hours per week than self-employed.
We report four entries for separate wages—monthly, daily, hourly, annually—in line with how
people reported their remuneration (e.g. the summary statistics for ‘monthly wages’ are presented
only for those who are paid monthly). Interestingly, the average earnings of the self-employed
are higher than salaried workers across all categories except annual wages. We suspect that this is
partially due to a small number of extremely high-earners among the self-employed (lawyers, doctors), as is also suggested by the high variance in earnings among the self-employed. The median
incomes (not reported here) reveal similar patterns, but the small sample size makes it difficult to
draw any concrete conclusions. At face value the data do not suggest that regular employees on
average earn significantly more than the self-employed. However, we caution that only the annual
wage numbers may include bonuses (and those are higher for regular employees), and in addition
to the aforementioned difficulties in evaluating self-employed income there may also be variation
in how a self-employed reports business earnings as opposed to private income (after all deductions
for running the business).
Finally, we find differences by industry. Because of our small sample size, we group certain sectors
together (similar to Patrick and Rohlen 1987). Following Cheng (1997) and others, we exclude
farming from our analysis. This yields the six industry categories of (1) primary sector (mining,
fishing and forestry), (2) construction, (3) manufacturing, (4) transport and communications, (5)
retail, wholesale, food and lodging, and (6) finance, insurance and real estate. Confirming Brinton
(1993), 29% of all self-employed are in retail, wholesale, food and lodging, followed by construction
(12%); of the non-standard workers 31% are in the retail, wholesale, food and logding sector, and
14% in manufacturing. However, contrary to Brinton’s impression of self-employed women working
in manufacturing piecework, in our sample 42% of the self-employed women are in retail, wholesale,
food and lodging, and only 2% in manufacturing (not reported here).
Our findings are in line with existing research and general impressions regarding employment in
Japan, indicating that our sample is representative in important ways. In order to tease out more
information from these data, in the next section we formulate testable hypothesis for regression
analysis.
3. Literature Review and Hypotheses
In addition to existing research on Japan, labor economics offers a huge body of literature on selfemployment, especially for the US. A complete review is beyond the scope of this paper. For the
purposes of our study, we focus on economic factors and personal factors that may determine a move
into self-employment.
3.1 Retrospective Data
At the most basic level, a switch into self-employment may be affected by macroeconomic conditions,
policies, and historical events such as a war. For any given year, high economic growth could have
two contrasting effects on career choice: (1) it opens up the job market, and employment is easy to
find, so that fewer enter self-employment; or alternatively, (2) it increases the chances of survival for
an aspiring entrepreneur, so that more new businesses open up. Masuda (2006) provided evidence
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 9
for the former, arguing that Japan’s recession of the 1990s brought a new push into self-employment
in regions with higher unemployment. In addition to the overall economy, we can also look at this
‘push’ proposition by measuring local job market opportunities. We explore by stipulating the following hypotheses:
H1a: People are more likely to switch into self-employment during periods of recession.6
H1b: People are more likely to switch into self-employment when local labor market conditions
are poor.
It is often said that a successful switch into self-employment is predicated on experience or wisdom
(Cheng 1997). One way to accumulate experience is simply over time; i.e. older employees can be
assumed to be more experienced than younger ones. However, Genda (2004) has shown that the
effect of age on self-employment declined during the 1990s in Japan. As we consider age as a determinant, we have to make two refinements. First, rigid rules of lifetime employment until recently
pushed most employees into retirement at around age 55 (now extended to 60 or 65) and thus may
have created a systemic bias for elderly to open their own shop. To reduce the ambiguity caused by
mandatory retirement, we cut our sample off at age 60.
Second, the age logic may not apply to women in the same way. For the US, Edwards and FieldHendrey (2002) show that young women often opt into self-employment to increase work flexibility
and juggle multiple family tasks. In Japan, women were (and are) often forced out of the workforce
upon marriage and during child-rearing years, be that through outright company policies or due to
systemic barriers such as a lack of childcare or the impossibility to match family duties with notoriously late working hours (Holthus 2010, Brinton 1989, 1993). Therefore, it is possible that women
switch into self-employment at a relatively young age. We can shed light on these matters because we
test our hypotheses separately for men and women. We stipulate,
H2: Older people are more likely to switch into self-employment.
In addition to wisdom, education (knowledge) can be a determinant for career switches. Research
on the relationship between self-employment and education has produced ambiguous results and may
also differ for men and women. On the one hand, less education may limit job choices in employment, whereas more education increases options as well as salary levels in employment (Cheng 1997,
Blanchflower 2000). On the other hand, for higher levels of education we have two contrasting scenarios. More education may be associated with greater success in self-employment, especially in hightechnology entrepreneurial sectors or professions such as law, medicine, or consulting (Robinson and
Sexton 1994, Lazear 2004). In contrast, high-aspiring entrepreneurs who aim to exploit a window of
opportunity may truncate their education for immediate business launch, such as witnessed in the US
for founders such as Bill Gates (Microsoft), Steve Jobs (Apple) or Mark Zuckerberg (Facebook). We
can explore these conflicting scenarios for Japan by testing, separately for men and women:
H3: A higher level of education increases the likelihood of switching into self-employment.
Even if education is cut short, an important key to success in self-employment can be workrelated experience. The literature on this is vast. For example, Lazear (2004) proposes that a
6. Because we will be testing the null hypothesis, the technically precise formulation of this hypothesis should read: ‘People
are no more likely to switch into self-employment during periods of recession than during periods of boom.’ To increase
readability, even at the loss of econometric precision, in this paper we follow the established social science custom and
express our hypotheses with an implicit direction of being ‘more likely’/ ‘less likely’. To further facilitate the reading of
our findings, we express all hypotheses in the positive direction.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
10 Jess DIAMOND and Ulrike SCHAEDE
‘jack-of-all-trades’ makes for a more successful entrepreneur, and generalized human capital formation is the recommended strategy. Such generalized knowledge can be acquired not just through
education but also through previous experience, either in self-employment or through rotations
while in regular employment. Using US census data, Robinson and Sexton (1994) show a positive
relationship between years of work experience and success in self-employment. Borjas (1986) provides evidence of a positive correlation between labor market experience and self-employment across
racial groups in the US. Similarly, Bates (1995) finds that work experience is positively correlated with
self-employment, in particular for women, even though this may be overpowered by industry effects.
How do these factors play out in Japan? In the US, high labor market mobility also allows for a strategy of learning through ‘job-hopping’. In Japan, large companies often train their regular employees
into generalists through ‘on-the-job training’ and ‘rotation-on-the-job’, i.e. promotion through a
variety of tasks. While this may make for excellent preparation, the income security and rigid promotion rules of lifetime employment translate into very low mid-career labor mobility that may make jobhopping less common in Japan. Yet, Koike (1983) proposed that it occurs in small firms where some
switch employers frequently before starting their own business. Japan even has a word for people who
switch from regular into self-employment, the so-called ‘datsu-sara’ (lit.: ‘extricating oneself from the
existence of a salary-man’). Moreover, for non-standard employees, switching into self-employment
may be an attractive option, even though accumulated learning is less structured (Diamond 2011).
For women, regular employment may be less of an ‘educational springboard’ than for men, both
for career chances and family choices. Wisdom through work experience cannot be gained if regular
careers are truncated. This may mean that for women education is a more important factor for selfemployment than experience.
Finally, experience as a self-employed (successful or not) is sometimes said to increase the likelihood of future self-employment, in a career path referred to as ‘serial entrepreneurship’ (Westhead
and Wright 1998, Amaral et al. 2011). Accumulated human capital in self-employment or reduced job
satisfaction in employment may lead some to switch out of self-employment into regular employment
and then back (Cheng 1997).7 We can explore these various scenarios by stipulating the following:
H4a: Past experience in regular employment increases the likelihood of switching into
self-employment.
H4b: Past experience in non-standard employment increases the likelihood of switching into
self-employment.
H4c: Past experience in self-employment increases the likelihood of switching into self-employment.
3.2 Personal Situation
For the years 2004–2007 we can explore additional personal markers that are not included in the retrospective data, and we can also offer a more precise snapshot of the situation in the early 21st century.
Perhaps the most basic need of any entrepreneur is money, and personal assets are often a way to overcome liquidity constraints.8 Evans and Jovanovic (1989) and Evans and Leighton (1989) found that
7. Other determinants of entering self-employment suggested in the literature include the number of previous employers,
their type, size and industry, and the level of position in previous employment. Unfortunately, our data do not allow an
analysis of these factors.
8. Perhaps the most obvious economic determinant factoring into the choice of self-employment is expected earnings.
Unfortunately, we cannot look into this issue with our data. As discussed above, without a specialized survey it is impossible to differentiate among different types of self-employed, and averages are not informative. Moreover, even if such a
differentiation were possible, wealth creation for innovative entrepreneurs often occurs post-hoc, when the company is
sold or taken public, at which moment these entrepreneurs are often no longer included in the self-employed dataset. We
leave the analysis of income projections for further research.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 11
greater assets raised the probability of switching into self-employment in the US. Even if financial assets
are limited, a house can be used as collateral when applying for a bank loan or the home itself may be
used for self-employed activity (Patrick and Rohlen 1987). In Japan, financial liberalization since the
1980s is said to have lowered previous high barriers to credit access for self-employed; for example,
Masuda (2006) finds little evidence of liquidity constraints. To explore this, we stipulate the following:
H5a: The larger an individual’s assets, the higher the likelihood of switching into self-employment.
H5b: Home ownership increases the likelihood of switching into self-employment.
Moreover, one expects that the family situation, in particular marriage and child-rearing, has an
impact on career choices. In Japan, it is often said that being in regular employment increases a man’s
chance to ‘marry well’, for employment affords status and signals income security. Yet, being married
also offers an important safety net for a self-employed man, as the wife may be able to contribute to
household income. For women, in contrast, the notion is that in order to ‘marry well’ they should be
prepared to give up career aspirations; as mentioned above, many companies to this day urge women
to resign upon marriage. In addition to child-rearing duties, this may push some married women into
self-employment. To explore societal factors associated with career choice, we posit the following:
H6a: Marriage increases the likelihood of switching into self-employment.
H6b: Young children increase the likelihood of switching into self-employment.
Finally, Cheng (1997) found that having a self-employed father was the overriding predictor for
entering self-employment in 20th century Japan. This is supported by Sorensen (2007), who shows
in a study of Denmark that parental role modeling is an important driver into self-employment (as
opposed to inheritance of capital or social assets, which did not loom large). Yet, in Japan, fast economic growth between the 1960s and 1990s, the high status associated with working in a lifetime
employment setting at a large company and growing societal pluralism may have reduced the attractiveness of following the father’s career path. Our data allow us to explore whether an association
continues to exist in Japan by stipulating the following:
H7: Growing up with a father in self-employment increases the likelihood of entering self-employment later in life.
4. Analyses and Results
Our queries are ‘What are the determinants of self-employment?’, and ‘What personal profile is associated with self-employment in Japan?’. Our unit of analysis is the act of switching into self-employment. We assume simple rationality: in each period, an individual bases the decision of whether to be
self-employed or pursue some other form of employment in the following period on the expected
utility of each choice.
Factors other than preferences and characteristics of the individual may determine employment
status. To capture this, we assume that the likelihood that individual i will be self-employed in year t
is given by a latent variable, yit
*, which depends linearly on characteristics of the individual as well as
other relevant factors that may not be related to the individual:
y x it
*
it
self employment
it
not self employment = − EU EU + =it it + − − δ β αi t + + it γ ε
where, xit is a vector of individual characteristics, αi
represents individual-specific characteristics
that do not change over time (such as an entrepreneurial predisposition), gt captures factors that
change over time but not over individuals (such as the state of the macroeconomy), and εit captures
all other relevant factors that may influence the likelihood of being self-employed.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
12 Jess DIAMOND and Ulrike SCHAEDE
Thus, the analytical equivalent of our search for determinants of self-employment is ‘what is the
value of β?’. To address issues of endogeneity, we test Hypotheses 1 through 4 (i.e. the retrospective questions) by estimating a linear probability model on the panel data of employment histories
of individuals between 1963 and 2003. Table 2 suggested that an individual’s current employment
depends greatly on the previous period’s employment. To account for this persistence in employment
we estimate the model using the Blundell and Bond (1998) General Method of Moments (GMM)
estimator. We perform a similar exercise—with different variables, as explained below—on the data
between 2004 and 2007 in order to test Hypotheses 5 through 7.
4.1 Entry into Self-Employment in the Past
Tables 4 (men) and Table 5 (women) report results for the entire retrospective panel, dating from
1963 to 2003. The dependent variable, yit, is a dummy variable that takes a value of ‘1’ if individual
i was in self-employment at any time during year t, and ‘0’ otherwise. Note that this includes longterm self-employed, recent switchers (e.g. young people without other employment options) and
those who might enter self-employment only temporarily while searching for other opportunities.
To control for the effect of one’s previous job on switching choices, we construct four dummy
variables, lagged by one year. Each of these dummies takes the value ‘1’ if the person was in regular
employment, non-standard employment, other employment, or looking for employment during a
given year but did not work, and ‘0’ otherwise. Since an individual could hold multiple jobs within a
given year, these dummy variables are not mutually exclusive.
To test for macroeconomic conditions (H1a, ‘contraction’) at the time of switching into selfemployment, we created a dummy variable that takes ‘1’ if Japan’s Cabinet Office defined the given
period as one of economic contraction, and ‘0’ otherwise. To test for local labor market conditions
(H1b), we used data from the Ministry of Health, Labor and Welfare to construct a job-openings-tojob-seekers ratio (henceforth: ‘job openings ratio’) at the regional level. A larger number means better local job market conditions, as more jobs are available per applicant in that area at that time.9 Age
is measured in years. The variables ‘regular’ and ‘non-standard’ employment measure the number
of years spent in each type of employment.10 We estimate the model separately for men and women,
and instead of testing for education we separate the sample by high school versus college degree and
compare the results.
Table 4 presents the results for men. We find a high level of persistence in self-employment. The
magnitude of this effect dwarfs the impact of all other types of previous employment, and we will
explore this stickiness further below. In looking at the other employment categories, we find that a
transfer into self-employment is most likely to follow dibs in ‘other’ types of employment (including
family businesses and side jobs), rather than regular or non-standard employment. For high school
graduates, these are followed by regular employment and finally unemployment, while for college
graduates this order is reversed.
9. The Cabinet Office data are available at http://www.esri.cao.go.jp/en/stat/di/100607rdates.html. Local labor data
were drawn from http://www.mhlw.go.jp/english/database/db-l/general_workers.html. This MHLW survey is conducted
annually in January, and we use the average of the previous year. When a single region in the KHPS survey corresponds
to more than one region in the Ministry of Health, Labour and Welfare (MHLW) survey, we take the average of the ratio
across those regions.
10. In addition to regular or non-standard employment, one could presumably also work in self-employment, a family business
or a side job. Thus, experience, regular employment and non-standard employment are not collinear.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 13
Table 4. Retrospective Panel Data (H1–H4): Results for Men.
High School College
(1) (2) (3) (4) (5) (6)
Lagged self-employment dummy 0.969***
(0.008)
0.962***
(0.008)
0.410***
(0.006)
0.952***
(0.009)
0.944***
(0.009)
0.492***
(0.007)
Lagged regular employment dummy 0.145***
(0.005)
0.139***
(0.005)
−0.031***
(0.004)
0.096***
(0.005)
0.094***
(0.005)
−0.029***
(0.005)
Lagged non-standard employment
dummy
0.183***
(0.010)
0.180***
(0.009)
0.008
(0.008)
0.114***
(0.012)
0.110***
(0.012)
0.003
(0.011)
Lagged unemployed dummy 0.136***
(0.014)
0.134***
(0.014)
−0.031**
(0.011)
0.100***
(0.013)
0.099***
(0.013)
−0.001
(0.011)
Lagged other employment dummy 0.199***
(0.019)
0.196***
(0.019)
0.039*
(0.015)
0.172***
(0.025)
0.166***
(0.024)
0.039
(0.021)
Age −0.003***
(0.0002)
−0.002
(0.001)
0.047***
(0.002)
−0.002***
(0.0003)
−0.003*
(0.001)
0.025***
(0.003)
Local job openings ratio −0.003
(0.003)
0.010
(0.007)
−0.014**
(0.005)
0.002
(0.004)
−0.002
(0.009)
−0.025***
(0.008)
Contraction 0.003
(0.002)
0.002
(0.002)
−0.034***
(0.004)
−0.007
(0.006)
Non-standard employment squared 0.0001
(0.0002)
0.0001
(0.0004)
Regular employment −0.041***
(0.002)
−0.024***
(0.002)
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
14 Jess DIAMOND and Ulrike SCHAEDE High School College (1) (2) (3) (4) (5) (6)
Regular employment squared −0.00004**
(0.00001)
−0.00005*
(0.00002)
Self-employment −0.073***
(0.002)
−0.067***
(0.003)
Self-employment squared 0.001***
0.00002
0.001***
(0.00004)
Constant −0.017*
(0.007)
−0.003
(0.018)
−0.834***
(0.04)
0.003
(0.010)
0.127
(0.070)
−0.347***
(0.080)
Year dummy variables No Yes Yes No Yes Yes
N 22,764 22,764 22,764 14,896 14,897 14,899
Chi-square 15,640 16,513 17,533 11,461 11,981 12,325
Table 4. Continued.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 15
Table 5. Retrospective Panel Data (H1–H4): Results for Women.
High School College
(1) (2) (3) (4) (5) (6)
Lagged self-employment dummy
0.923***
(0.008)
0.919***
(0.008)
0.463***
(0.005)
0.888***
(0.011)
0.886***
(0.011)
0.517***
(0.008)
Lagged regular
employment
dummy
0.037***
(0.004)
0.037***
(0.004)
−0.011***
(0.003)
0.025***
(0.004)
0.026***
(0.004)
−0.007
(0.004)
Lagged non-standard employment
dummy
0.028***
(0.005)
0.028***
(0.005)
−0.019***
(0.004)
0.008
(0.006)
0.008
(0.006)
−0.039***
(0.006)
Lagged unemployed
dummy
0.042***
(0.011)
0.042***
(0.011)
−0.006
(0.009)
0.033**
(0.013)
0.034**
(0.013)
0.004
(0.011)
Lagged other
employment
dummy
0.043***
(0.006)
0.042***
(0.006)
−0.037***
(0.005)
0.016
(0.010)
0.017
(0.010)
−0.037***
(0.009)
Age −0.001***
(0.000)
0.002*
(0.001)
0.027***
(0.001)
−0.001
(0.000)
−0.001
(0.001)
0.012***
(0.002)
Local job-opening
ratio
−0.004
(0.003)
−0.005
(0.007)
−0.008
(0.005)
−0.008
(0.005)
0.016
(0.011)
0.047***
(0.009)
Contraction −0.002
(0.002)
0.0001
(0.003)
Non-standard
employment
−0.006***
(0.002)
−0.002
(0.003)
Non-standard
employment
squared
−0.00004
(0.0001)
−0.0004**
(0.0001)
Regular employment −0.006***
(0.001)
−0.010***
(0.002)
Regular employment
squared
−0.0001***
(0.00003)
−0.00001
(0.00004)
Self-employment −0.053***
(0.001)
−0.034***
(0.002)
Self-employment
squared
0.001***
(0.00002)
0.0004***
(0.00004)
Constant 0.028***
(0.008)
−0.161**
(0.062)
−1.039***
(0.053)
0.031*
(0.013)
0.020
(0.028)
−0.301***
(0.034)
Year dummy
variables
No Yes Yes No Yes Yes
N 19,611 19,612 19,614 8,456 8,457 8,459
Chi-square 15,405 15,619 16,506 6,939 7,058 6,431
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
16 Jess DIAMOND and Ulrike SCHAEDE
As mentioned, we do not test the effect of education directly, but rather report results separately. It
is interesting to note that the estimated coefficients for the previous employment variables are almost
always larger for high school graduates than for college graduates, suggesting that less education is associated with more switching into self-employment. Recall that we have structured this analysis so as to
eliminate unobserved factors that are constant over time but may vary across individuals, and also factors
that may vary over time but not individuals. Thus, our results run counter to Hypothesis 3 (education).
While macroeconomic contraction is not significantly correlated with switching into self-employment,
the sign of the coefficient on the local job openings ratio suggests that entry into self-employment is
associated with deteriorating local labor market conditions. Age has a positive effect, after controlling
for employment history: the probability of switching into self-employment rises over the time of one’s
career. Thus, we find mild support for Hypothesis 1 (labor market), and we support Hypothesis 2 (age).
Moreover, we find that past experience in self-employment, regular employment and non-standard
employment all decrease the likelihood of a switch into self-employment. Over time the magnitude
of this effect grows with regular employment, yet it wanes with non-standard and self-employment.
We reject Hypothesis 4 (job-hopping).
Table 5 shows the results for women, which are largely similar. As with men, there is a high degree
of persistence in self-employment. The effect of ‘other’ employment relative to the alternatives varies from model to model, but there is a clear ranking of the effects of the remaining sectors: the
largest impact comes from unemployment, followed by regular employment and then non-standard
employment. However, the point estimates of these coefficients are tiny in comparison with the coefficient on previous-period self-employment. Also, as for men, the coefficients are consistently larger
for high school than for college graduates, meaning that women with a high school degree are more
likely to become self-employed than more educated women. Again, age has a positive effect after
controlling for employment history.
One remarkable result is that, contrary to men, entry into self-employment for educated women is
associated with improving local labor market conditions. This result has been consistently strong in
several of our analyses, including some not reported here. It may suggest that educated women either
enter self-employment as high-aspiring entrepreneurs or professionals so that the local job market
is irrelevant for their choices, or that they can be more opportunistic and open shop only in times
of a good local economy. This latter interpretation might suggest that men and women face different career options: whereas men—as presumed breadwinners—may be driven into self-employment
during hard times due to lack of other opportunities, women might be more at liberty to wait for an
economic uptick. Of course this may apply mostly to married or independently wealthy women, and
we will look more into lifestyle matters below.
In terms of past employment experience, we find that both regular and non-standard employment
experience have a negative impact on entry into self-employment, and this effect grows over time.
In other words, the longer a woman is in some form of employment the less likely she is to open her
own shop. The effect of previous self-employment experience is also negative but diminishes over
time. As with men, these results are inconsistent with the idea that workers use experience gathered while working at a company as a springboard to launch their own businesses. Thus, we reject
Hypothesis 4 (prior work experience) for all of our categories.
The previous analysis pertained to all switchers, including those who are self-employed only temporarily. These may be different in important ways from long-term self-employed. To explore this
potential difference, we conducted the same analysis as above but limited our sample to those individuals with at least 10 years of potential work experience. That is, we truncated the age bracket
by omitting the young, and we defined ‘long-term self-employed’ as currently in self-employment
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 17
with at least 10 years of self-employment experience. The results are similar, and we do not report
them here. One noteworthy difference is that neither the overall economy nor the local job market
conditions are significant for long-term self-employment for either category. The previous finding
of a positive correlation with age holds, for both men and women, after controlling for past work
experience.
Also consistent with the earlier finding, the likelihood of becoming long-term self-employed
decreases with tenure in non-standard employment for men with high school degrees, and for
women in general. Thus, workers in non-standard employment are increasingly less likely to become
long-term self-employed as time goes on, regardless of education. The length of this negative impact
differs for men and women. For men with a college degree, being in non-standard employment initially increases the likelihood of becoming long-term self-employed, but after four years the impact
becomes negative. For women with a High School degree, chances of switching from non-standard
to self-employment turn negative after 10 years.
Regular employment continues to have a negative effect on long-term self-employment for men,
perhaps given the security associated with being a ‘salary-man’. Thus, the ‘datsu-sara’ phenomenon
may be smaller than sometimes claimed. For women, the picture is different: after 15 years in regular employment, the effect turns positive. This matches the general impression of educated women
joining the workforce for a while before getting married (or being discouraged by reduced career
options) and then looking for other options.11
4.2 Personal Attributes
To explore Hypotheses 5 through 7, we switch to the panel data from 2004 to 2007. These data
were collected for the same individuals, but they are richer and allow us to observe additional variables. However, certain data items, such as the employment status dummy variables, are now defined
differently. In the pre-2004 data, individuals were coded as being in a given employment sector, such
as self-employment, if they were in that sector at any point during a given year. From 2004 onwards,
we observe an individual’s employment status only for the month prior to the survey. Because the
survey was conducted annually in January, and to adjust for this difference, we now code the employment category dummies in year t as ‘1’ if the individual’s primary source of income in December of
year t−1 was in the same employment category, and ‘0’ otherwise.
We report the results in Tables 6 (men) and Table 7 (women). The first column provides a base
case scenario similar to that in Table 4. Although the results are not directly comparable, we now
find that for men (though not for women) the effect of self-employment in the previous year is not
overwhelmingly greater than the impact of being in other employment categories for men. In fact,
unemployment in the previous year has a larger effect for men, suggesting that switching into selfemployment may occur due to a lack of employed options.
Remarkably, for the years 2004–2007 the age variable turns negative (with the exception of female
high school graduates). This is consistent with Genda’s (2004) findings for the 1990s. Thus, contrary
to the long-term period pre-2004, younger people are now more likely to enter self-employment
than older. We take this to suggest a shift toward higher plurality as younger males, especially those
with only a high school degree, appear to be more likely to enter self-employment. This is in spite
of the fact that we find no significant impact of overall economic conditions or the local job market.
11. The effect of a previous stint in self-employment for entering into long-term self-employment generally starts off positive
but turns negative after a few years. The turning points are after 10 years for male college graduates, 17 years for female
high school graduates, and nine years for female college graduates.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
18 Jess DIAMOND and Ulrike SCHAEDE Table 6. Panel Data 2004–2007 (H5–H6): Results for Men. High School College (1) (2) (3) (4) (5) (6) (7) (8) Lagged self-employment dummy −0.134** (0.042) 0.135 (0.074) 0.147** (0.053) 0.131* (0.051) 0.417*** (0.081) 0.408*** (0.090) 0.434*** (0.079) 0.467*** (0.080) Lagged regular employment Dummy 0.102* (0.050) 0.323*** (0.095) 0.220*** (0.060) 0.214*** (0.056) 0.165* (0.070) 0.134* (0.068) 0.370*** (0.062) 0.388*** (0.063)
Lagged non-standard
employment Dummy
0.069
(0.051)
0.324***
(0.093)
0.200***
(0.060)
0.191***
(0.057)
0.104
(0.075)
0.117
(0.078)
0.317***
(0.069)
0.330***
(0.070)
Lagged unemployment
dummy
0.085
(0.058)
0.352**
(0.115)
0.228**
(0.070)
0.211**
(0.067)
0.249*
(0.104)
0.147
(0.104)
0.433***
(0.097)
0.448***
(0.098)
Age 0.060***
(0.009)
−0.141**
(0.050)
−0.098**
(0.032)
−0.089**
(0.031)
−0.021
(0.013)
−0.005
(0.074)
−0.028
(0.037)
−0.027
(0.038)
Local job-openings ratio 0.009
(0.134)
0.263
(0.219)
0.107
(0.154)
0.089
(0.150)
−0.030
(0.159)
−0.099
(0.186)
0.028
(0.159)
0.039
(0.159)
Non-standard employment 0.082
(0.069)
0.087*
(0.040)
0.080*
(0.038)
0.036
(0.051)
−0.052
(0.042)
−0.055
(0.043)
Regular employment 0.095
(0.056)
0.120***
(0.031)
0.110***
(0.029)
0.079
(0.043)
0.008
(0.032)
0.006
(0.033)
Self-employment 0.189***
(0.057)
0.184***
(0.034)
0.170***
(0.032)
0.115*
(0.045)
0.055
(0.032)
0.054
(0.033)
Lagged financial assets (¥100
million)
0.306
(0.196)
−0.071
(0.133)
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 19
High School College
(1) (2) (3) (4) (5) (6) (7) (8)
Lagged fixed assets (¥100
million)
0.225
(0.139)
0.002
(0.096)
Lagged household debt
(¥100 million)
0.096
(0.153)
−0.126
(0.138)
Homeowner 0.044
(0.042)
0.027
(0.043)
Married 0.007
(0.060)
−0.066
(0.065)
Children younger than
6 years
−0.053*
(0.024)
−0.003
(0.027)
Constant −2.527***
(0.454)
3.012*
(1.512)
0.837
(1.098)
0.768
(1.070)
0.938
(0.593)
−0.825
(2.475)
0.635
(1.066)
0.705
(1.103)
Year dummy variables Yes Yes Yes Yes Yes Yes Yes Yes
N 1,452 764 1,436 1,452 1,348 799 1,332 1,348
Chi-square 75 117 116 122 59 226 201 227
Table 6. Continued.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
20 Jess DIAMOND and Ulrike SCHAEDE Table 7. Panel Data 2004–2007 (H5–H6): Results for Women. High School College (1) (2) (3) (4) (5) (6) (7) (8) Lagged self-employ- ment dummy 0.207*** (0.041) 0.060 (0.049) 0.253*** (0.049) 0.210*** (0.047) 0.090* (0.044) 0.171 (0.101) 0.190** (0.059) 0.134* (0.058) Lagged regular employ- ment dummy 0.051 (0.035) 0.033 (0.047) 0.058 (0.037) 0.047 (0.037) 0.017 (0.038) 0.065 (0.058) 0.026 (0.039) 0.015 (0.039)
Lagged non-standard
employment dummy
0.059*
(0.025)
0.051 (0.029) 0.061*
(0.026)
0.058*
(0.026)
0.031
(0.024)
0.120**
(0.045)
0.051
(0.026)
0.042
(0.026)
Lagged unemployment
dummy
0.046
(0.042)
0.038 (0.057) 0.041
(0.044)
0.034
(0.043)
−0.035
(0.043)
0.121
(0.076)
−0.018
(0.046)
−0.032
(0.046)
Age 0.037**
(0.013)
0.008 (0.024) 0.012
(0.048)
0.052
(0.047)
−0.003
(0.008)
−0.022
(0.023)
−0.012
(0.023)
−0.004
(0.023)
Local job-openings
ratio
0.207
(0.155)
0.269 (0.179) 0.125
(0.161)
0.210
(0.159)
0.197
(0.137)
0.001
(0.220)
0.091
(0.142)
0.152
(0.142)
Non-standard
employment
−0.012 (0.026) −0.016
(0.021)
−0.017
(0.021)
0.030
(0.022)
−0.008
(0.017)
−0.012
(0.017)
Regular employment 0.006 (0.027) −0.023
(0.021)
−0.022
(0.021)
0.025
(0.022)
0.003
(0.016)
0.001
(0.016)
Self-employment −0.006 (0.022) −0.003
(0.020)
−0.016
(0.020)
0.086***
(0.022)
0.010
(0.013)
0.005
(0.012)
Lagged financial assets
(¥100 million)
0.016 (0.103) 0.164
(0.163)
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 21
High School College
(1) (2) (3) (4) (5) (6) (7) (8)
Lagged fixed assets
(¥100 million)
0.181(0.106) 0.049
(0.119)
Lagged household debt
(¥100 million)
−0.583***
(0.169)
−0.050
(0.213)
Homeowner −0.053
(0.047)
0.024
(0.044)
Married −0.004
(0.057)
−0.047
(0.049)
Children younger than
6
−0.044
(0.025)
0.002
(0.027)
Constant −1.867**
(0.663)
−0.597 (0.724) −0.326
(2.018)
−2.144
(1.988)
−0.028
(0.361)
−0.532
(0.686)
0.387
(0.758)
0.115
(0.756)
Year dummy variables Yes Yes Yes Yes Yes Yes Yes Yes
N 1,133 584 1,122 1,133 950 495 941 950
Chi-square 40 42 48 47 13 107 42 37
Table 7. Continued.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
22 Jess DIAMOND and Ulrike SCHAEDE
While the results are not significant, directionally we find that higher financial assets in the previous
year are associated with entry into self-employment (except for male college graduates), as predicted
in Hypothesis 5a. For fixed assets (not including home ownership), there is also a consistently positive correlation with self-employment, and it is interesting that the size of that correlation is larger
for high school graduates than college graduates. Our data do not offer any further insight here.
One possible scenario is that the father is in self-employment, and assets are measured for the entire
household. We explore the role of the father’s profession below. High household debt is negatively
associated with self-employment, except for male high school graduates; this effect is significant for
female high school graduates. We cannot be sure whether male high school graduates carry a higher
debt because they took out a loan to build a business, or because they had no other job options or
income in the previous year. Either way, in the years 2004–2007 we find strong suggestions that
younger male high school graduates with higher debt (for whatever reason) are more likely to enter
self-employment. Homeownership is insignificant (perhaps pointing at higher plurality) and the sign
is positive, offering mild support for Hypothesis 5b, except for high school women. Unfortunately
our data do not allow for further exploration; we can only note that high school graduates seem to
enter self-employment with different financial backgrounds than college graduates.
The impact of marriage is negative for educated men and for all women. Again, these results are not
statistically significant, and we cannot support Hypothesis 6a. In contrast, the number of young children
(below the age of six) predicts a switch into self-employment for female college graduates (though not for
men). Once again, the impression emerges that educated women stand out in comparison to the other
groups. In this instance, it appears they are pushed more into self-employment if they raise children.
In sum, our most consistent finding for the period of 2004 to 2007 is a strong persistence in selfemployment. That said, we also find increasing plurality in the profiles of the self-employed in the
early 2000s, who are also more likely to be younger. Moreover, the results for educated women differ
consistently from those for the other groups. She may perhaps be a professional with the freedom to
time her entry into self-employment to overlap with improvement in the local economy. Or she may
be married with young children, forced to resign from regular employment, and decide to enter selfemployment when the local economy is good to supplement other household income. Whatever the
case may be, the profile of the educated woman is rather different from that of men.
Overall, we find large variation in the estimated coefficients on marriage and young children, and
attribute these to a growing pluralism among the self-employed, ranging from young women with
children to elderly male ex-employees to lawyers and doctors. Even though the estimated coefficients
show consistent patterns, most of our results are not statistically significant. It is possible that the
small sample size prevents us from obtaining more precise estimates. As the Keio surveys are continued in the future, we hope that additional data will help reduce the remaining ambiguity.
4.3 Fathers in Self-Employment
Finally, as formulated in Hypothesis 7, growing up in a family where the father runs his own business
may be a strong predictor of being self-employed. Several reasons have been suggested for this effect,
in particular the parents as role models (Sorensen 2007). Cheng (1997) in her historical study of
Japan found that the father being self-employed overrode all others determinants of becoming selfemployment. We wonder whether this is still true in Japan.
Our retrospective panel includes a question about the father’s occupation when the respondent was 16 years old. Thus, we switch back to the longitudinal data and estimate a probit model
for the years 1963 through 2003, to measure the effect of the father being self-employed on the
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 23
probability that an individual has entered self-employment at some point in his life. The dependent
variable is a dummy with the value ‘1’ if an individual has ever been in self-employment between
1963 and 2003 (regardless of how long), and ‘0’ otherwise. The results are reported in Table 8.
The father’s experience in self-employment significantly increases the likelihood that a child enters
self-employment. Specifically, the average marginal effect says that having a self-employed father
at age 16 increases the likelihood that one enters self-employment by 14% for male high school
graduates, 15% for male college graduates, 6.5% for female high school graduates, and 3.7% for
female college graduates. We note that the result for female high school graduates is statistically
significant only at the 10% level, while the result for female college graduates is not significant.
At first sight, our results may be indicative of a certain social stratification, as self-employed
parents may not have the desire or means to send children to schools that result in lifetime employment careers. However, the differences in degree for men and women suggests that other mechanisms may also be at work, such as inheritance (an obligation to take over the father’s business,
which may be stronger for a son) or business-specific skill formation. Unfortunately, our data do
not allow us to control for family characteristics, and we can only conclude that a correlation exists
for men regardless of degree, and women without college education. Educated women seem to be
least affected by this influence and enter self-employment for idiosyncratic reasons.
5. Labor Market ‘Stickiness’ and Career Paths of the Self-Employed
Table 2 reported only limited switching between job categories, and job market stickiness is the most
consistent finding in our regression analysis. Perhaps the biggest factor is lifetime employment, where
Table 8. Retrospective Panel Data: The Effect of Fathers in Self-Employment.
Variable Men Women
High School College High School College
(1) (2) (3) (4)
Father
self-employed
0.505***
(0.153)
0.724***
(0.182)
0.344*
(0.200)
0.200
(0.238)
Constant −1.023***
(0.100)
−1.357***
(0.118)
−1.345***
(0.132)
−1.296***
(0.139)
N 364 324 280 220
Pseudo R2 0.030 0.061 0.015 0.005
Average arginal effect
Father
self-employed
0.140***
(0.041)
0.150***
(0.037)
0.065*
(0.038)
0.037
(0.044)
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
24 Jess DIAMOND and Ulrike SCHAEDE
pay incentives reward tenure, so that employees often stay with one company for the first 30 years
of their work life. Regular rotations and rigid tournament rules for promotions also mean that midcareer entry into a regular position at a new company can be difficult (Abegglen 1984, Lincoln and
Nakata 1997, Schaede 2008). These factors combine into a significant and long-lasting impact of a
worker’s initial job on his/her future employment trajectory, and it has been argued this impact is
particularly pronounced in Japan (Kondo 2007; Esteban-Pretel et al. 2011).
With growing youth unemployment, this determinative rigidity in early-career employment
choices is a grave concern. It has given rise to the so-called ‘lost generation’ phenomenon, referring
to those who graduated from college during the 1990s when a scarcity in regular employment openings limited their career options, allegedly for life. Most studies on this issue have focused on the
impact of entering the labor force in non-standard employment (e.g.,Brinton 2010). If the rigidity
also holds true for self-employment, young people may decide against self-employment at an early
stage in life to avoid getting ‘stuck’. It is sometimes argued that the determinative quality of Japan’s
labor market is a barrier to entrepreneurship in Japan.
Our data allow us to explore this situation. We use the retrospective panel and look at the local job
openings ratio at the time of an individual’s entry into the labor market to see how much it affects the
probability that he or she will be in self-employment later in life. We run the following probit model:
Pr y k , ( X k X c z k ) it i it | ( ) 0 1 = + Φ b b2 0i i + + b b 3 4 t i0 5 + + b b t i 6 t
where Φ(·) is the cumulative distribution function for the standard normal distribution; yit is a binary variable which is ‘1’ if individual i is in self-employment in year t, and ‘0’ otherwise; ki0 is the local
job-openings ratio at the time that individual i initially entered the labor market; and Xit is a vector
of relevant explanatory variables. We include year fixed effects (zt), cohort fixed effects (ci0), and the
current local job-openings ratio (kit).
The results are reported in Table 9. For people with only a high school degree, poor local labor
market conditions at the time of first entry into the labor market significantly increase the likelihood
of being in self-employment at some point in one’s life. For male college graduates, the result is
not significant. Yet again, the one exception to the pattern is female college graduates, for whom a
healthy local labor market at time of entry actually increases the likelihood of future self-employment.
With these results we can also calculate the odds. On average, a 10 percentage point decrease in
the local job market (a fall in the local job-openings ratio from say, from 1.1 to 1.0) at the time of
labor market entry increases the probability of being self-employed by 0.35% for high school males,
0.12% for college males, and 0.15% for female high school graduates. However, for female college
graduates, such a change in the initial local job openings ratio decreases the probability of future
self-employment by 0.32%. The base rate of self-employment in our sample is approximately 8% for
women, 13% for college males, and 18% for high school males. The standard deviation of the local
job openings ratio ranges from 0.16 (in Hokkaido) to 0.67 (in Chubu). Therefore, a back-of-theenvelope calculation yields a non-trivial impact of labor market conditions: a one standard deviation
deterioration in the local job openings ratio at the time of labor market entry in an area like Chubu
may increase the future probability of self-employment by as much as 13% for both, high school
men [(0.035×0.67)/0.18] and women [(0.015×0.67)/0.08]. Male college graduates could see as
much as a 6% [(0.012×0.67)/ 0.13] increase. Yet, female college graduates face a very large 27%
[(0.032×0.67)/ 0.08] decrease in the probability of future entry into self-employment if the local
labor market sours.
We repeated this exercise for long-term self-employment by measuring the effect of the local
job openings ratio at the time of labor market entry on the probability of being in long-term
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 25
self-employment in the future. The results are similar to those shown in Table 9 and are not reported
here. Now the results are significant only for male high school graduates, and the estimated effect for
male college graduates is small and insignificant. The same back-of-the-envelope calculation as above
suggests that a one standard deviation decrease in the local job openings ratio at the time of labor
market entry increases the future probability of long-term self-employment by as much as 15% for
high school graduates yet decreases it by 8% for female college graduates.
In terms of magnitude, the average marginal effect of the local job openings ratio at the time of
labor market entry is comparable to the average marginal effect of the job openings ratio in the most
recent survey year; for men, it is larger. In other words, a poor job market at the time of school graduation is as determinative for future self-employment as are current labor market conditions.
Overall, our findings are similar to those for non-standard employment and underscore the rigidity in Japan’s career trajectories. Insofar as self-employment entails asset-specific investments (such as
equipment or skill formation), rigidities may be even more severe: whereas non-standard work may
lead to skills that are more easily transferable to regular employment, self-employment investments
may raise the opportunity costs of switching.
These rigidities may also offer one possible interpretation for the age effect results. In the retrospective panel (until 2004), young people, concerned about getting ‘stuck’, were more likely to
assume paid positions, while middle-aged workers dissatisfied with their job situation may have been
less concerned with the ‘one-way’ nature of switching into self-employment. However, for the years
Table 9. Effect of Initial Labor Market Conditions on Future Self-Employment.
Men Women
High School College High School College
(1) (2) (3) (4)
Probit coefficients
Initial local job
openings ratio
−0.152***
(0.032)
−0.062
(0.044)
−0.111**
(0.039)
0.207**
(0.068)
Local job openings
ratio
−0.018
(0.053)
−0.050
(0.074)
−0.133*
(0.067)
0.414***
(0.114)
Constant −6.179***
(0.262)
−5.472***
(0.354)
−2.135***
(0.623)
−2.092***
(0.539)
N 16,219 12,686 14,375 6,686
Pseudo R2 0.071 0.081 0.067 0.146
Average marginal effect
Initial local job
openings ratio
−0.035***
(0.007)
−0.012
(0.008)
−0.015***
(0.005)
0.032***
(0.011)
Local job openings
ratio
−0.004
(0.012)
−0.009
(0.014)
−0.018*
(0.009)
0.065***
(0.018)
*** p < 0.01 ** p < 0.05 * p < 0.1 at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
26 Jess DIAMOND and Ulrike SCHAEDE
2004 through 2007, we saw that the age variable switched signs. Perhaps the job market situation for
the young, especially those with only a high school degree, has become so dire that previous tradeoff
considerations are no longer as relevant.
6. Conclusions
This paper analyzes the recent situation of self-employment in Japan, using a private Keio University
household survey that allows us to draw connections between working in self-employment and
economic conditions and individual characteristics such age, education, work history, family situation, and assets. We believe ours is the first study that explores these factors separately for men
and women.
The most persistent finding is that switching across job categories remains limited. Yet, while
Japan’s labor market remains very sticky, in some sense there is perhaps more switching than one
might have thought. By looking at those instances of switching, we shed new information on some
stipulations that were previously explored mostly for the US. For the 50 years prior to 2004, our
results suggest that lower education (high school only) predicts entry into self-employment. We find
no support for the ‘experience through job-hopping’ notion for Japan. Moreover, in this period,
more Japanese were likely to enter self-employment in bad economic times, and they were older.
These results may favor the ‘push’ notion of an involuntary entry into self-employment due to limited options. Growing up in a household where the father was self-employed greatly increased the
likelihood for the son entering self-employment.
For the more recent years of 2004 through 2007, we support some of the standard notions suggested in finance: more assets and home ownership are associated with self-employment, as is more
debt (for high school graduates). Marriage is not a factor, but for women, having young children is.
Most noteworthy, in the recent period, younger people were more likely to switch into self-employment than older at a time when local job markets were not changing much.
While this may bespeak of new forms of entrepreneurism in Japan, this new trend could also trigger serious concerns under the heading of ‘lost generation’: by being pushed into self-employment
due to limited options, young people may miss out on regular career opportunities because of a lack
of training and the labor market’s determinative rigidity. We offer support for this concern by showing that the probability of becoming a long-term self-employed ranges between 6% and 15% if the
labor market is poor at the time of school graduation.
There is one important exception to most of our results: educated women. In almost all analyses,
they go against the grain. They are more likely to enter self-employment in good times and even
though young children are definitely a determining factor, these women may also be single professionals, with less credit constraints. Unlike for the other groups, the father’s profession does not
appear to figure into their choice. The job market at the time they enter the workforce has little effect
on their future careers. A positive interpretation of these findings is that educated women appear to
be the ones with the most options in our sample.
For men and women with high school degrees, our results do not suggest one clear profile.
While the estimated coefficients reveal consistent patterns, more often than not, they are not
precisely estimated. We hope that future iterations of the Keio survey will add additional data
that allow us to shed further light on this important segment of Japan’s labor market. But clearly
we find evidence of the great diversity in backgrounds, motivations as well as constraints among
Japan’s self-employed.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
Self-Employment in Japan: A Microanalysis of Personal Profiles 27
References
Abegglen, James C. 1984. The Strategy of Japanese Business. Cambridge: Ballinger.
Amaral, A. Miguel, Rui Baptista, and Francisco Lima. 2011. ‘Serial entrepreneurship: impact of human capital on time to
re-entry’. Small Business Economics 37(1): 1–21.
Bates, Timothy. 1995. ‘Self-employment entry across industry groups’. Journal of Business Venturing 10(2): 143–156.
Borjas, George. 1986. ‘The Self-Employment Experience of Immigrants’. Journal of Human Resources Fall 21(4): 485–506.
Blanchflower, David. 2000. ‘Self-employment in OECD countries’. Labour Economics 7(5): 471–505.
Blundell, Richard and Bond, Stephen. 1998. ‘Initial Conditions and Moment Restrictions in Dynamic Panel Data Models’.
Journal of Econometrics 87: 115–144.
Brinton, Mary C. 1989. ‘Gender Stratification in Contemporary Urban Japan’. American Sociological Review 54 (4):
549–564.
——. 1993. Women and the Economic Miracle in Japan: Gender and Work in Postwar Japan. Berkeley: University of California
Press.
——. 2010. Lost in Transition: Youth, Work and Instability in Postindustrial Japan. Cambridge: Cambridge UP.
Cheng, Mariah Mantsun. 1997. ‘Becoming Self-Employed: The Case of Japanese Men’. Sociological Perspectives 40:581–600.
Diamond, Jess. 2009. Non-Standard Employment and Earnings in Japan. Working Paper: University of California, San
Diego.
——. 2011. Employment Status Persistence in the Japanese Labor Market. Working Paper: University of California, San Diego.
Edwards, Linda N. and Elizabeth Field-Hendry. 2002. ‘Home-Based Work and Women’s Labor Force Decisions’. Journal of
Labor Economics 20: 170–200.
Esteban-Pretel, Julen; Nakajima, Ryo, and Tanaka, Ryuichi. 2011. ‘Are contingent jobs dead ends or stepping stones to regular jobs? Evidence from a structural estimation’. Labour Economics 18(4): 513–526
Evans, David S. and Boyan Jovanovic. 1989. ‘An Estimated Model of Entrepreneurial Choice under Liquidity Constraints’.
Journal of Political Economy 97(4): 808–827.
Evans, David S. and Linda S. Leighton. 1989. ‘Some Empirical Aspects of Entrepreneurship’. American Economic Review
79: 519–535.
Genda, Yuji. 2004. Jobu kurieishon (Job Creation in Japan). Tokyo: Nihon Keizai Shimbunsha.
Genda, Yuji and Ryo Kambayashi. 2002. ‘Declining Self-Employment in Japan’. Journal of the Japanese and International
Economies 16:73–91.
Holthus, Barbara G. 2010.‘Parental Evaluation of Institutionalized Childcare: Survey Results from Japan’. Working Paper,
German Institute for Japanese Studies Tokyo. Available at http://www.dijtokyo.org/articles/holthus10_01.pdf. Last
accessed 6 December 2012
Koike, Kazuo. 1983. ‘Workers in small firms and Women in Industry’. In Contemporary Industrial Relations in Japan, eds. T.
Shirai. Madison: University of Wisconsin Press: pp. 89–115.
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): 379–402
Lazear, Edward P. 2004. ‘Entrepreneurship’. American Economic Review 94: 208–211.
Lincoln, James R. and Yoshifumi Nakata. 1997. ‘The Transformation of the Japanese Employment System: Nature, Depth,
and Origins’. Work and Organizations 24(1): 33–55.
Masuda, Tatsuyoshi. 2006. ‘The Determinants of Latent Entrepreneurship in Japan’. Small Business Economics 26: 227–240.
Patrick, Hugh T. and Thomas P. Rohlen. 1987. ‘Small-Scale Family Enterprises’. In The Political Economy of Japan, Part
1: The Domestic Transformation, eds. K. Yamamura and Y. Yasuba. Stanford: Stanford UP: pp. 331–384.
Robinson, Peter and Sexton, Edwin. 1994. ‘The effect of education and experience on self-employment success’. Journal of
Business Venturing 2: 141–156.
Schaede, Ulrike. 2008. Choose and Focus: Japanese Business Strategies for the 21st Centuries. Ithaca: Cornell University Press.
Sorensen, Jesper B. 2007. ‘Closure and Exposure: Mechanisms in the Intergenerational Transmission of Self-Employment’.
Research in the Sociology of Organizations 25: 83–124.
at University of California, San Diego on June 25, 2014 http://ssjj.oxfordjournals.org/ Downloaded from
28 Jess DIAMOND and Ulrike SCHAEDE
Soos, Piroska. 1990. ‘Self-Employed Evasion and Tax Withholding: A Comparative Study and Analysis of the Issues’. UC Davis
Law Review 24: 107–193.
Westhead, Paul and Mike Wright. 1998. ‘Novice, portfolio, and serial founders: are they different?’ Journal of Business
Venturing 13(3): 173–204
- You must be logged in to reply to this topic.