SLIDE 1 Long-term effect of displacement in Japan in Japan
Mayu FUJII, NIPSSR Ryo KAMBAYASHI, IER, Hitotsubashi University 16th May, 2013 the Conference for the launch of the OECD Analytical Report
SLIDE 2 Sketch of talk
- 1. The role of displacement in the Japanese economy
- 2. What we know about the Japanese displacement
- 3. New dataset
- 4. Preliminary results
a. Methodology: how to set the control group? b. Basic results c. How to interpret the tenure effect
- 5. Remarks and discussions
SLIDE 3
T ti Turnover ratio (% of employment stock, 1971-2011, ETS published data)
SLIDE 4
S ti ti b Separation ratio by reason (% of employment stock, 1971-2011, ETS published data)
SLIDE 5
S ti ti b Separation ratio by reason (% of employment stock, 1971-2011, ETS published data)
SLIDE 6
1 The role of displacement in the 1.The role of displacement in the Japanese economy
Compared to the stock of unemployment, hiring/separation has been relatively stable. The average frequency of involuntary separation is not hi h (3 9% b 1971 d 2011) so high (3.9% as average btw 1971 and 2011). Traditionally, we have regarded the displacement as a minor issue in our labor markets.
SLIDE 7 1 The role of displacement in the 1.The role of displacement in the Japanese economy
Since 1990s, the frequency of involuntary separation has grown up, and this trend seems to be corresponded to the increase in stock of unemployment. Th d l l b k i J The dual labor market in Japan. Displacement has become a major issue
- Q. Does the displacement create a long‐term inefficiency?
- Q. If it does, are there policies to remedy the negative effect?
SLIDE 8 2 What we know about the Japanese 2.What we know about the Japanese displacement
Data constraint
- Previous literature heavily depends on cross‐sectional
datasets datasets
- ETS; Employment Trend Survey, an establishment survey to capture
labor flow in each establishment.
- ESS; Employment Status Survey, a household survey to retrospectively
track the one‐year transition of employment status.
We use the information from governmental administration; pension enrolment records.
SLIDE 9 2 What we know about the Japanese 2.What we know about the Japanese displacement
Timing of displacement (Kambayashi and Kato, 2012)
- ESS v.s. DWS in the U.S.
- Long tenure protects workers from displacement in Japan
(but not in recent US).
- Demographic attributes (gender age) affects the probability
Demographic attributes (gender, age) affects the probability to be displaced in Japan (but not in the US).
Displacement in Japan is not randomly allocated among workers, but concentrates in particular group (elder, female, less tenured workers).
SLIDE 10 2 What we know about the Japanese 2.What we know about the Japanese displacement
Effects of displacement (Bognanno and Kambayashi, 2013;
Kambayashi and Kato, 2012; Bognanno and Delgado, 2008; Abe et al., 2002)
- Rehiring probability of displaced workers within one year is
- Rehiring probability of displaced workers within one year is
also affected by demographic attributes such as age, gender, and education.
- One‐shot wage penalty of displacement is higher in aged and
long tenured workers. University graduates are likely to lose wage more. wage more.
Duality of labor market may have affected the effects of y y
- displacement. But in long‐term?
SLIDE 11
2 What we know about the Japanese 2.What we know about the Japanese displacement
Therefore, it is important to confirm the long‐term effect of displacement in Japan, in order to deepen the argument on duality (flexibility) of labor market.
SLIDE 12 3.New datasets
LOSEF (The Japanese Longitudinal Survey on Employment and Fertility)
d l ( l b
- Random sampling survey in 2011 (sample size 6000 between
age of 30 and 59).
- With pension enrolment records since their first job
With pension enrolment records since their first job
- Pension enrolment records include the whole history of base
monthly salary in April (as the contribution of pension) and the name of employer.
- When they change their employer, we ask industry,
- ccupation contracting term title of each job and reason
- ccupation, contracting term, title of each job, and reason
and training of each job change (max. 45 years).
SLIDE 13 3.New datasets
LOSEF (The Japanese Longitudinal Survey on Employment and Fertility)
h ld f ( l b h f h ld
- Household information (marital status, birth of children,
cohabitation with parents) can be used in each year.
- Because the information on monthly salary is based on the
Because the information on monthly salary is based on the administrative records, it is relatively more accurate compared to other retrospective survey.
- Microdata may be available for academic use, with an English manual in
http://takayamaonline.net/pie/stage3/English/d_p/E2012.html
SLIDE 14 3.New datasets
Downside of LOSEF
- LOSEF salary history covers only one of three main pension
systems in Japan systems in Japan
- Kosei Nenkin Hoken (Employee’s pension) 35m in 2011
- Kyosai Nenkin Hoken (Public sector’s pension) 4m in 2011
- Kokumin Nenkin Hoken (Self‐employed and others) 30m in 2011
- Sample is heavily biased to standard workers in private sector.
Nonstandard workers are not likely to participate in KNH (but Nonstandard workers are not likely to participate in KNH (but they are likely in Kokumin Hoken). The condition to participate in KNH has gradually widen. Recently it is more l k l l d d d k h b f likely to include nonstandard workers than before.
SLIDE 15 3.New datasets
Downside of LOSEF
- Monthly salary is so called “Standard monthly
compensations” compensations”
- excludes bonus and extra allowance
- classified into 47 class (‘under 63000JPY’ to ‘over
classified into 47 class ( under 63000JPY to over 1.17mJPY’) with various widths. The widths also have been changed from time to time.
SLIDE 16 S ti ti Separation ratio
(% of employment stock, 1971-2011, LOSEF and ETS published data)
SLIDE 17 Di l t ti Displacement ratio
(ratio to employment stock, 1971-2011, LOSEF and ETS published data)
SLIDE 18 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
Measure the wage penalty of displacement for longer‐ term. Sample restriction
i. Exclude if experienced voluntary quit. ii E l d if di l d ithi th t ii. Exclude if displaced within three year tenure.
- iii. Exclude if worked as nonstandard worker at least in one
point of career (tentative). po t o ca ee (te tat e)
- iv. Exclude if multiple displacement (tentative).
v. Exclude if unemployment spell exceed one year (tentative).
SLIDE 19 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
“0” if continuously employed “d0=1” if displaced “dnt=1” t year before displacement, “dpt=1” t year later from displacement y p p y p
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 dn1 d0 dp1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 dn1 d0 dp1 dn1 d0 dp1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 dn1 d0 dp1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
SLIDE 20 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
“0” if continuously employed “d0=1” if displaced “dnt=1” t year before displacement, “dpt=1” t year later from displacement y p p y p
1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 dn1 d0 dp1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 dn1 d0 dp1 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007
SLIDE 21 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
“0” if continuously employed “d0=1” if displaced “dnt=1” t year before displacement, “dpt=1” t year later from displacement y p p y p
ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya dn1 d0 dp1 yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb dn1 d0 dp1 yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc
SLIDE 22 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
“0” if continuously employed “d0=1” if displaced “dnt=1” t year before displacement, “dpt=1” t year later from displacement y p p y p
ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya dn1 d0 dp1 yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb dn1 d0 dp1 yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc d0 d0 yb yc ya yb yc OLS estimates v.s. average of average of
SLIDE 23 4 Preliminary results 4 .Preliminary results
- a. Methodology: how to set the control group?
“0” if continuously employed “d0=1” if displaced “dnt=1” t year before displacement, “dpt=1” t year later from displacement y p p y p
y ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya ya dn1 d0 dp1 yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb yb dn1 d0 dp1 yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc yc d0 yb yb d0 v.s. average of v.s. average of FE estimates average of yc yc g
SLIDE 24 4 Preliminary results 4 .Preliminary results
sample LOSEF sample dependent variable estimated model coeff.
s.e. p‐value
coeff.
s.e. p‐value
coeff.
s.e. p‐value
coeff.
s.e. p‐value
d 3 0 032 0 012 0 015 0 049 LOSEF log of monthly salary OLS FE FE OLS dn3 ‐0.032
0.082 0.697
‐0.012
0.062 0.843
0.015
0.040 0.706
0.049
0.037 0.192
dn2 ‐0.059
0.082 0.473
‐0.059
0.062 0.343
‐0.026
0.040 0.514
0.003
0.037 0.937
dn1 ‐0.029
0.091 0.751
‐0.050
0.069 0.473
‐0.029
0.045 0.518
‐0.003
0.042 0.940
d0 ‐0.157
0.082 0.054
‐0.191
0.062 0.002
‐0.153
0.040 0.000
‐0.128
0.037 0.001
dp1 ‐0.120
0.082 0.140
‐0.166
0.062 0.008
‐0.127
0.040 0.002
‐0.103
0.037 0.006
dp2 ‐0.089
0.082 0.273
‐0.145
0.062 0.020
‐0.107
0.040 0.008
‐0.082
0.037 0.029
dp3 ‐0.087
0.082 0.287
‐0.150
0.062 0.016
‐0.115
0.040 0.004
‐0.087
0.037 0.019
dp4 ‐0.081
0.082 0.318
‐0.150
0.062 0.016
‐0.118
0.040 0.003
‐0.088
0.037 0.019
year age tenure sample size no no no 16878 16878 16878 yes yes yes yes no yes yes no no 16878 sample size
F value 0.6245 1237.6 1459.18 16878 16878 16878 16878 0.3532
SLIDE 25
Wage penalty of displacement Wage penalty of displacement
(1971-2011, LOSEF)
SLIDE 26 4 Preliminary results 4 .Preliminary results
13% penalty is plausible? (Bognanno and Kambayashi, 2013)
Around 6% from all sample
(max. -30%) and/or nonstandard workers?
SLIDE 27 4 Preliminary results 4 .Preliminary results
As results of conservative estimation,
- At the point of displacement, they are likely to lose about
13% of monthly salary 13% of monthly salary.
- The wage penalty of displacement does not disappear even
after four years of reemployment. y p y
- The salary loss estimated from FE is smaller than from OLS,
which indicates displacements are biased to less (time‐ invariant) unobservable ability.
SLIDE 28 4 Preliminary results 4 .Preliminary results
- c. How to interpret the tenure effect
When we control for tenure…
sample dependent variable LOSEF log of monthly salary p estimated model coeff.
s.e. p‐value
coeff.
s.e. p‐value
dn3 ‐0.015
0.062 0.804
0.002
0.037 0.963
dn2 ‐0.063
0.062 0.313
‐0.045
0.037 0.224
g y y OLS FE dn1 ‐0.025
0.069 0.721
0.066
0.041 0.107
d0 ‐0.126
0.063 0.045
0.144
0.039 0.000
dp1 ‐0.108
0.063 0.084
0.147
0.039 0.000
dp2 ‐0.096
0.063 0.127
0.146
0.039 0.000
dp3 ‐0 109
0.063 0.083
0 118
0.039 0.002
dp3 0.109
0.063 0.083
0.118
0.039 0.002
dp4 ‐0.116
0.063 0.064
0.096
0.039 0.013
year age tenure yes yes yes yes yes yes sample size
F value 0.6258 1483.89 16878 16878
SLIDE 29 4 Preliminary results 4 .Preliminary results
- c. How to interpret the tenure effect
A typical trajectory of salary
d0 w/o tenure d0 with tenure
SLIDE 30 4 Preliminary results 4 .Preliminary results
- c. How to interpret the tenure effect
d0 without tenure may indicate
- loss of human capital by displacement
d0 with tenure may indicate
l i f h i l d i h i j b
- accumulation of human capital during the previous job
- Declining coefficient d0>dp1>dp2>dp3>dp4 with tenure may
imply the tenure effect is weaker in the second job than in the imply the tenure effect is weaker in the second job than in the first job.
SLIDE 31
- 5. Remarks and discussions
By using new panel dataset, we find around 13% decline in monthly salary at the displacement. This wage penalty may not disappear even after several years of reemployment as standard workers.
When we consider the possibility of reemployment as nonstandard workers the average wage penalty may amplify nonstandard workers, the average wage penalty may amplify.
SLIDE 32
- 5. Remarks and discussions
More control at the displacement
the difference between gender, age, period industry, occupation change small sample size may cause problems…
Extension of sample
multiple displacement multiple displacement control group definition longer effect of displacement (e.g. 10 yrs) g p ( g y )
