Fig. 3. The left figure shows that given
𝜑
2
=
0,
𝜑
1
increases
from 0 to 0.5, and the right figure switches parameters. We
reveal that the magnitude of
𝜑
1
or
𝜑
2
increases the statistic
value of the ADF test declines, indicating when the econ-
omy gets more deadweight loss that it will cause the gap
between V
J
and V
W
to become unstable (persistent). A sim-
ulation with 1000 replications is done, and the parameters
are set up by
𝜆 =
0.01,
𝜏 =
0.01,
𝜎 =
0.02, and r
=
0.02,
respectively; we have V
J
∕
V
W
and its ADF test. Without loss
of generality, we set k
∗
=
1 and the range is
(
1
−
𝛿
1
,
1
+
𝛿
2
)
,
where
𝛿
1
=
𝜑
1
and
𝛿
2
=
𝜑
2
∕(
1
−
𝜑
2
)
.
156
W.D. Chen
Economic Modelling 73 (2018) 152–162
of
𝜕𝛽∕𝜕𝜑
1
<
0. In two particular cases, if
𝜑
1
=
0, then equation
(8)
will
be the wage equation as in
Mortensen and Pissarides (1994)
, and
𝛼 = 𝛽
.
On the contrary, if
𝜑
1
=
1, then we see that the coefficient of productiv-
ity equals zero, implying a disconnection between wage and productiv-
ity and showing the model is unstable; see
Fig. 3
. In general,
𝜑
1
∈ (
0
,
1
)
,
which displays the role of a decelerator for the market mechanism. If
the value of
𝜑
1
increases, then the market becomes more inefficient.
Considering an extraordinary case, if
𝜑
1
>
1 indicates the cost is exten-
sive and greater than the transferring surplus, then it means when work-
ers ask for a wage raise that this will cause a loss of them, because the
value is more massive than the gain, which will make the coefficient of
productivity become negative.
In the same way, we regard another scenario with downward wage
rigidity. If the surplus flows from workers to firm, then the firm offers
the transferring cost. We can express the relationship as V
(
J
,
W
) =
V
(
J
+ (
1
−
𝜑
2
)
dB
2
,
W
−
dB
2
)
, where dB
2
is the transfer at the upper
boundary for the wage, and
𝜑
2
denotes the cost to cut salaries. We then
have
(
1
−
𝜑
2
)(
1
−
𝛼)
W
=
𝛼
J, where
𝛽 =
𝛼
1
−
𝜑
2
(
1
−
𝛼)
. Thus, workers with
an equilibrium wage can be displayed in the following relationship:
w
=
(
1
−
𝛼)(
1
−
𝜑
2
)
1
−
𝜑
2
(
1
−
𝛼)
b
+
(
𝛼
1
−
𝜑
2
(
1
−
𝛼)
)
y
+
(
𝛼
1
−
𝜑
2
(
1
−
𝛼)
)
𝜅𝜃.
(9)
We have
𝜕𝛽∕ 𝜕𝜑
2
>
0. If
𝜑
2
=
0, then
𝛽 = 𝛼
and the model returns to a
competitive model. In another case, if
𝜑
2
=
1, then
𝛽 =
1, implying the
employees enjoy the full benefit from productivity growth, and
𝜑
2
∈
(
0
,
1
)
implies the cost is smaller than the transferred surplus when firms
cut salaries.
According to equation
(4)
, we can simulate the relationship between
V
W
and V
J
. Without loss of generality,
Fig. 3
a shows that if we fix
𝜑
2
=
0.00 and let
𝜑
1
increase, then the relation between V
J
and V
W
becomes persistent, which implies that a higher ratio of deadweight
loss will cause the economy to be unstable. In the same way, if we fix
𝜑
1
and allow
𝜑
2
to change, see
Fig. 3
b, then we have similar results.
According to
Fig. 3
, we know that if
𝜑
2
or
𝜑
1
is great, then their
relationship will become unstable, and the gap between the right- and
left-hand sides will show persistence. We realize why the relationship
between productivity and wage could be unstable, because the persis-
tent components dominate their relationship. Thus, we can detect the
capability of the market mechanism by the persistent portion of the gap
between the wage and equilibrium. If the ratio is high, then this market
is associated with a high level of market failure. When the coefficient
of productivity is low and accompanied by a sizable persistent portion,
then the market has upward wage rigidity, and company owners occupy
the primary surplus.
4. Market failure in Japan labor markets
This section focuses on the relationship between wages and produc-
tivity, especially under the situation of upward wage rigidity. We use
the MOLS method to estimate the long-run relationship for the wage
equation in Japan labor markets, in which we consider the structural
changes associated with different government policies. According to the
previous discussion, we realize that if the market exhibits upward wage
rigidity, then this will decrease the coefficient of productivity and make
the market unstable, which will reflect in the gap between the real wage
and equilibrium.
From historical data, we realize that Japan has a low unemploy-
ment rate; the average value is 2.73% from 1953 until 2016. How-
ever, since 2000, the unemployment rate has dramatically fluctuated.
Because of economic restructuring, the ratio hit 5.4% in 2002 and then
declined, until the financial crisis when the unemployment rate rose
again and reached 5.1% in 2009. After the great recession the unem-
ployment rate began to fall; from 2015 to 2016 the unemployment rate
has held around 3.3%. However, even during the financial crisis, the
unemployment rate was relatively low versus other countries; such as
the U.S., where its average unemployment rate was 5.82% from 1948
until 2016. We should note that one important reason for such a low
unemployment ratio is the hiring of a lot of temporary workers. In 2014,
non-regular employees hit 37.4% of the Japanese workforce, and nearly
11% of Japanese employers retained dispatched workers. The IC indus-
try employed the highest percentage of dispatched workers (26.9%),
while the FI industry took on the second most significant rate (19.1%).
Since temporary workers are not like regular employees who have
‘lifetime’ employment protections, the dispatched agencies usually pay
a relatively low salary. To alleviate job insecurity, the Democratic Party,
the Socialist Party, and The People’s New Party presented a reform bill
on employment insurance to the ruling party on March 6, 2009. Fur-
thermore, on September 30, 2015, the Japan government amended the
Worker Dispatch Law by loosening existing restrictions primarily on the
length of time.
We apply our model to the 21 industry markets
1
of Japan, and the
data come from the Ministry of Health, Labour, and Welfare. The period
is from January 2000 to October 2015, and the index for January 2000
is 100.
Considering the 2008 financial crisis and the revision of the Dispatch
Worker Law, we set dummy variables to display the structure changes
for the labor markets. Simultaneously, according to
Phillips and Lore-
tan (1991)
,
Saikkonen (1991)
, and
Stock and Watson (1993)
, we can
estimate their long-run relationship for the wage equation by using the
MOLS method as follows:
w
t
=
𝛽
0
+
𝛽
1
y
t
+
𝛽
2
𝜃
t
+
𝛽
3
D
2
+
𝛽
4
D
3
+
𝛽
5
D
2
y
t
+
𝛽
6
D
3
y
t
+
𝛽
7
D
2
𝜃
t
+
𝛽
8
D
3
𝜃
t
+
p
∑
s
=−
p
𝛾
1s
Δ
y
t
−
s
+
p
∑
s
=−
p
𝛾
2s
Δ
𝜃
t
−
s
+
u
t
⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏞⏟
g
t
,
(10)
where D
2
and D
3
present the periods associated with the financial crisis
(from August 2007 to May 2009) and after the revision of the Dispatch
Worker Law (June 2009 to October 2015), and the cross-product terms
display the coefficient shifts due to the events that took place. Notation
g
t
denotes the gap between the observation and the long-term expected
equilibrium, which will show persistency if rigidity takes place. We
assume the disturbance term u
t
is an AR process:
u
t
=
𝜙
1
u
t
−
1
+
𝜙
2
u
t
−
2
+ · · · +
𝜙
p
u
t
−
p
+
a
t
.
The t statistics can be multiplied by s
T
∕
𝜆
to use the standard t table,
where s
T
is MSE in equation
(10)
,
𝜆 = 𝜎∕(
1
−
𝜙
1
−
𝜙
2
− · · · −
𝜙
p
)
, and
𝜎
2
=
E
(
a
2
t
)
. If we get the long-run coefficients, then we can put the
coefficients in equation
(10)
to get the gap between the observation
and equilibrium, in which we have to remove the short-run impacts
that are associated with the differential items for each industry. We can
display the gap for the jth industry as follows:
w
jt
= ̂
𝛽
j0
+ ̂
𝛽
j1
y
jt
+ ̂
𝛽
j2
𝜃
jt
+ ̂
𝛽
j3
D
2
+ ̂
𝛽
j4
D
3
+ ̂
𝛽
j5
D
2
y
jt
+ ̂
𝛽
j6
D
3
y
jt
+ ̂
𝛽
j7
D
2
𝜃
jt
+ ̂
𝛽
j8
D
3
𝜃
jt
+
g
jt
,
(11)
where g
jt
is the gap that denotes the deviation in the wages away
from the effective equilibrium, where the coefficients are obtained from
equation
(10)
. Through equation
(11)
, we can get the gap between the
real wage and the expected value.
We illustrate the first two industries with the highest number of
dispatched workers, information and communication (IC) and financial
1
The 21 industries include Construction (CN(L), CN(S), Finance and Insurance (FI(L),
FI(S)), Manufacturing (MN(L), MN(S)), Utilities (UT(L), UT(S)), Wholesale and Retail
Trade (WR(L), WR(S)), Mining and Quarry, Stone and Gravel (MQ(L),MQ(S)), Compound
Services (CS(L), CS(S)), Education and Learning Support (EL(L),EL(S)), Information and
Communication (IC(L),IC(S)), Transport and Postal Acts (TP(L), TP(S)), and Medicine,
Health Care and Welfare (MH(L)), where (L) and (S) indicate the employee numbers are
large-scale (more than 30) and small-scale (more than 5) respectively).
157
W.
D
.
Ch
en
Ec
o
n
o
m
ic
Mo
d
ell
in
g
73
(2018)
152–162
Table 1
We estimate the wage equation associated with structure changes by MOLS.
a
The data used from 1/2000 to 12/2015 for 21 industries in Japan.
Section
Intercept
y
t
b
𝜃
t
D
2
D
3
D
2
y
t
D
3
y
t
D
2
𝜃
t
D
3
𝜃
t
CN(L)
4.254*
c
(0.514)
d
0.056 (0.108)
0.186* (0.054)
0.194 (0.483)
0.373 (0.309)
−
0.016 (0.103)
−
0.052 (0.065)
−
0.291 (0.153)
−
0.283* (0.070)
CN(S)
4.099* (0.339)
0.100 (0.071)
−
0.041 (0.036)
0.225 (0.319)
0.176 (0.204)
−
0.042 (0.068)
−
0.034 (0.043)
−
0.036 (0.101)
0.060 (0.046)
FI(L)
4.066* (0.681)
0.104 (0.143)
0.391 (0.072)
0.359 (0.640)
0.082 (0.410)
−
0.101 (0.136)
−
0.012 (0.086)
0.302 (0.203)
−
0.420* (0.092)
FI(S)
4.535* (0.504)
0.008 (0.106)
0.243* (0.053)
0.310 (0.474)
0.037 (0.303)
−
0.088 (0.101)
−
0.007 (0.064)
0.268 (0.150)
−
0.248* (0.068)
MN(L)
5.412* (0.274)
−
0.183* (0.058)
0.295* (0.029)
0.191 (0.258)
0.009 (0.165)
−
0.047 (0.055)
0.000 (0.035)
0.189* (0.082)
−
0.126* (0.037)
MN(S)
4.480* (0.120)
0.015 (0.026)
0.255* (0.040)
0.068 (0.169)
0.035 (0.096)
−
0.017 (0.040)
−
0.002 (0.020)
0.133 (0.123)
−
0.093 (0.072)
UT(L)
4.364* (0.588)
0.044 (0.124)
0.147* (0.062)
0.614 (0.553)
−
0.128 (0.354)
−
0.126 (0.118)
0.046 (0.074)
−
0.056 (0.175)
−
0.403* (0.080)
UT(S)
4.190* (0.535)
0.087 (0.112)
−
0.002 (0.057)
0.762 (0.503)
0.051 (0.322)
−
0.174 (0.107)
−
0.004 (0.068)
0.191 (0.159)
−
0.232* (0.073)
WR(L)
6.133* (0.598)
−
0.328* (0.126)
−
0.046 (0.063)
−
1.018* (0.563)
−
0.316 (0.360)
0.186 (0.120)
0.055 (0.076)
0.477* (0.178)
0.182* (0.081)
WR(S)
5.078* (0.432)
−
0.100 (0.091)
−
0.286* (0.046)
−
0.448 (0.406)
−
0.185 (0.260)
0.062 (0.087)
0.018 (0.055)
0.518* (0.129)
0.336* (0.059)
MQ(L)
6.117* (1.070)
−
0.342 (0.225)
0.467* (0.113)
0.165 (1.007)
0.668 (0.644)
−
0.021 (0.214)
−
0.090 (0.136)
−
0.279 (0.319)
−
1.041 (0.146)
MQ(S)
7.643* (1.025)
−
0.617* (0.216)
−
0.749* (0.108)
−
1.713 (0.964)
0.751 (0.617)
0.285 (0.205)
−
0.192 (0.130)
0.878 (0.305)
0.276 (0.139)
CS(L)
2.347* (0.822)
0.517* (0.173)
−
0.765* (0.087)
−
1.419 (0.773)
0.987 (0.494)
0.331* (0.165)
−
0.198 (0.104)
−
0.481 (0.244)
0.085 (0.112)
CS(S)
2.524* (0.688)
0.468* (0.145)
−
0.591* (0.073)
−
1.233 (0.647)
0.559 (0.414)
0.267 (0.138)
−
0.118 (0.087)
−
0.121 (0.205)
0.176 (0.094)
EL(L)
4.771* (0.455)
−
0.016 (0.096)
−
0.493* (0.048)
−
0.171 (0.428)
−
0.231 (0.274)
−
0.043 (0.091)
0.006 (0.058)
1.142* (0.135)
0.252* (0.062)
EL(S)
4.389* (0.417)
0.073 (0.088)
−
0.698* (0.044)
0.013 (0.392)
−
0.169 (0.251)
−
0.081 (0.084)
−
0.014 (0.053)
1.093* (0.124)
0.374* (0.057)
IC(L)
5.074* (0.485)
−
0.102 (0.102)
−
0.082 (0.051)
−
0.633 (0.456)
−
0.200 (0.292)
0.123 (0.097)
0.035 (0.061)
0.247 (0.144)
0.194* (0.066)
IC(S)
5.425* (0.438)
−
0.176 (0.092)
−
0.140* (0.046)
−
0.544 (0.412)
−
0.166 (0.264)
0.098 (0.088)
0.025 (0.055)
0.345* (0.130)
0.220* (0.060)
TP(L)
4.491* (0.509)
0.025 (0.107)
−
0.204* (0.054)
−
0.336 (0.479)
−
0.364 (0.307)
0.043 (0.102)
0.053 (0.064)
0.416* (0.152)
0.478* (0.069)
TP(S)
4.812* (0.447)
−
0.051 (0.094)
−
0.137* (0.047)
−
0.388 (0.420)
−
0.367 (0.269)
0.057 (0.090)
0.057 (0.057)
0.352* (0.133)
0.318* (0.061)
MH(L)
4.810* (0.392)
−
0.036 (0.082)
−
0.352* (0.041)
−
0.300 (0.368)
−
0.493 (0.236)
0.009 (0.078)
0.067 (0.050)
0.785* (0.117)
0.331* (0.053)
a
The MOLS model is set by w
t
=
𝛽
0
+
𝛽
1
y
t
+
𝛽
2
𝜃
t
+
𝛽
3
D
2
+
𝛽
4
D
3
+
𝛽
5
D
2
y
t
+
𝛽
6
D
3
y
t
+
𝛽
7
D
2
𝜃
t
+
𝛽
8
D
3
𝜃
t
+
∑
p
s
=−
p
𝛾
1s
Δ
y
t
−
s
+
∑
p
s
=−
p
𝛾
2s
Δ
𝜃
t
−
s
+
u
t
, where u
t
is a serial correlation process.
b
The results reveal that a high degree of upward wage rigidity exists.
c
An asterisk * denotes rejection of the null hypothesis at the 5% significance level.
d
The value in the parentheses is the standard deviation.
158
W.D. Chen
Economic Modelling 73 (2018) 152–162
insurance (FI), as examples. According to the number of employees, we
classify the companies into large-scale (number of employees is more
than 30) and small-scale (more than 5 and smaller than 30). According
to equation
(10)
we can estimate the long-run relationship for the IC
industry as follows.
Large
−
Scale:
w
t
=
5.074
∗
[
11.199
]
−
0.102
[−
1.067
]
y
t
−
0.082
∗
[−
1.713
]
𝜃
t
−
0.633
[−
1.485
]
D
2
−
0.200
[−
0.732
]
D
3
+
0.123
∗
[
1.351
]
D
2
y
t
+
0.035
[
0.615
]
D
3
y
t
+
0.247
∗
[
1.833
]
D
2
𝜃
t
+
0.194
∗
[
3.153
]
D
3
𝜃
t
+
g
t
,
(12)
Small
−
Scale:
w
t
=
5.425
∗
[
8.534
]
−
0.176
[−
1.314
]
y
t
−
0.140
∗
[−
2.083
]
𝜃
t
−
0.544
[−
0.910
]
D
2
−
0.166
[−
0.434
]
D
3
+
0.098
[
0.771
]
D
2
y
t
+
0.025
[
0.313
]
D
3
y
t
+
0.345
∗
[
1.825
]
D
2
𝜃
t
+
0.220
∗
[
2.545
]
D
3
𝜃
t
+
g
t
,
(13)
where
[
.]
denotes the t value that has been adjusted by serial corre-
lation, and an asterisk * indicates the significance levels of 5%. We
can see that most coefficients of productivity for IC are negative. With
the large-scale companies as an example, for the first period (January
2000–July 2007) its value is
−
0.102; for the second period (August
2007–May 2009) it is 0.021 (0.021
= −
0.102
+
0.123); and the coeffi-
cient in the third period is
−
0.067 (
−
0.067
= −
0.102
+
0.035). Because
the coefficients for IC are small and even negative, it implies that severe
upward wage rigidity exists. In traditional approaches, we know that a
negative coefficient does not fulfill the assumption requirement for the
bargaining power to be between 0 and 1. However, referring to equa-
tion
(8)
, we know that the transfer cost will affect the coefficient of
productivity; if the transfer cost is enormous and when
𝜑
1
>
1, then
the coefficient becomes negative. That means wage increases could
cause the total surplus to shrink, because the surpluses of both firms
and workers decrease. In other words, we can say that companies could
decrease payments to increase productivity so as to compete with for-
eign exporters. We should note here that the coefficients in the above
equations present a co-movement relationship that does not imply the
influence direction is from the right-hand side to the left.
We are aware that programmers in the IC industry usually need flex-
ible working time. Some of them even work at home. It is thus not nec-
essary to ban workers by a fixed working time length, but this is entirely
different from workers in other sectors, such as the financial insurance
industry, who usually offers immediate legal service that needs a fixed
working time length. This intrigues our attention as to whether or not
the Work Dispatch Law has the same effect on different industries. In
the following context, we show the estimation of the FI industry.
Large
−
Scale:
w
t
=
4.066
∗
[
6.387
]
+
0.104
[
0.778
]
y
t
+
0.391
∗
[
5.803
]
𝜃
t
+
0.359
[
0.600
]
D
2
+
0.082
[
0.215
]
D
3
−
0.101
[−
0.794
]
D
2
y
t
−
0.012
[−
0.151
]
D
3
y
t
+
0.302
[
1.595
]
D
2
𝜃
t
−
0.420
∗
[−
4.855
]
D
3
𝜃
t
+
g
t
,
(14)
Small
−
Scale:
w
t
=
4.535
∗
[
9.617
]
+
0.008
[
0.077
]
y
t
+
0.243
∗
[
4.877
]
𝜃
t
+
0.310
[
0.700
]
D
2
+
0.037
[
0.132
]
D
3
−
0.088
[−
0.932
]
D
2
y
t
−
0.007
[−
0.119
]
D
3
y
t
+
0.268
∗
[
1.909
]
D
2
𝜃
t
−
0.248
∗
[−
3.864
]
D
3
𝜃
t
+
g
t
.
(15)
We realize that the coefficients of FI are similar to IC, and they
are all close to zero, implying severe upward wage rigidity. One thing
a little bit different is that during the financial crisis, the coefficients
of productivity (D
2
y
t
) for the FI industry decreased, but those for
the IC industry increased. That means upward wage rigidity became
Fig. 4. Applying the block bootstrap method with 300 times of re-sample replacement.
The Boxplots show the quartiles for each factor. With the confidence interval, (Q
1
-1.5
Δ
Q,
Q
3
+
1.5
Δ
Q) where
Δ
Q
=
Q
3
−
Q
1
, we have 12 stationary and 9 non-stationary factors.
The data are between January 2000 and December 2015.
more severe in the financial insurance industry during the economic
crisis, which is reasonable, because a lot of workers lost their jobs,
and a lot of salaries were cut. However, this situation did not appear
in the IC industry. It could be that IC is an efficient industry, and
its workers have flexibility and are used to temporary work. Though
the financial crisis burst forth, their tasks were difficult to reduce,
because they are essential for business security.
Table 1
shows the
estimation results for the 21 industries. Most of the coefficients of
productivity are close zero, implying there exists broad upward wage
rigidity.
After estimating the long-run relationship for the wage equation,
we can obtain the gaps between the wages and equilibrium. Following
Chen (2016a
,
b)
, we apply a multiple-factor panel data model to detect
the degrees of market failure across different industries. According to
equation
(10)
, we obtain the gap between wage and equilibrium for
each industry by the following equation:
g
jt
=
w
jt
−
𝛽
j0
−
𝛽
j1
y
jt
−
𝛽
j2
𝜃
jt
−
𝛽
j3
D
2
−
𝛽
j4
D
3
−
𝛽
j5
D
2
y
jt
−
𝛽
j6
D
3
y
jt
−
𝛽
j7
D
2
𝜃
jt
−
𝛽
j8
D
3
𝜃
jt
.
If a market is free and efficient, then g
jt
will show a stationary ten-
dency, and the wage will fluctuate around the equilibrium; otherwise,
g
jt
demonstrates a persistent property, indicating wage rigidity. Con-
sider a N
×
1 vector of residuals with common hidden factors. Some of
these factors are persistent, and some are mean reversions. We set a
one-period lagged model for the gap terms:
Δ
g
t
=
𝜻
1
g
t
−
1
+
u
t
,
for t
=
1
, … ,
T
,
(16)
where
Δ
g
′
t
= [Δ
g
1
,
t
, Δ
g
2
,
t
, … , Δ
g
N
,
t
]
, g
′
t
−
1
= [
g
1
,
t
−
1
,
g
2
,
t
−
1
, … ,
g
N
,
t
−
1
]
,
and u
′
t
= [
u
1
,
t
,
u
2
,
t
, … ,
u
N
,
t
]
. Here,
𝜻
1
is an N
×
N matrix, N is the cross-
section dimension, and T is the time dimension. The observation g
j
,
t
represents the gap of the jth market at time t, and
Δ
g
j
,
t
is its first differ-
ence, for j
=
1
, … ,
N.
Following
Chen (2016a
,
b)
and according to equation
(16)
, we can
obtain the common factors hidden in the gaps and then classify per-
sistent and mean-reverting components through the confidence inter-
vals estimated by the Bootstrap approach, as seen in
Fig. 4
. We have
12 mean reversion and nine persistent factors in the Japan labor mar-
kets.
Fig. 5
shows these factors whose proportions are above 5% in
the total sum of squares, including the stationary and non-stationary
components. Since the gaps contain stationary and non-stationary com-
ponents, we can measure the proportion of stability for each indus-
try. If the ratio of stationarity is high, then the market is more effi-
cient.
Table 2
shows a summary for all industries. Among them, the
159
W.D. Chen
Economic Modelling 73 (2018) 152–162
Fig. 5. The first five factors are stationary factors and the last three are non-stationary, whose proportions of the sum of squares are above 5%.
information and communication industry with large-scale IC(L) firms
had stable proportions over 90%, implying high efficiency in the
market.
To evaluate the change of market efficiency, we measure their sta-
bility ratios by three different times. The first period from January
2000 to July 2007 is between the dot-com collapse and the global
financial crisis, which encompasses Japan’s Lost Decade. The second
one is from August 2007 to May 2009 and is associated with the
global financial crisis. The third one is between June 2009 and Octo-
ber 2015, after the revision of the Worker Dispatch Law to the new
Amendment Act. We are interested in the effect of the Worker Dispatch
Law.
Table 3
shows the proportion of instability for the three periods. We
illustrate the large-scale companies in the finance and insurance FI(L)
industry as a typical example, because most industries exhibit similar
patterns. During the first period (January 2000 to July 2007) the pro-
portion of instability for FI(L) is 16.87%, while during the second period
and suffering from the financial crisis (August 2007 to May 2009) it is
29.10%, showing that the portion of instability drastically increased.
This dramatic change shows that the deadweight loss is rising quickly.
Because the market size of the finance and insurance industry vastly
shrunk during the financial crisis, a lot of workers turned toward tem-
porary work, which largely reduced wages and made the market inef-
ficient. We can see its productivity coefficient decreased from 0.104
to 0.003
(
0.003
=
0.104
−
0.101
)
. This is consistent with our statement
that when the coefficient of productivity becomes smaller, the market is
worse and has more failure, implying there is widespread upward wage
rigidity. After the revision of the Dispatch Worker Law (June 2009 to
October 2015), the portion of instability returns to 0.1987. It expresses
that the effect of the Dispatch Worker Law is positive. Because the gov-
ernment asked the dispatched agencies to pay workers with the rule of
“the same pay for the same work,” it mostly increased salaries. That also
reflects in the coefficient of productivity, as the coefficient increases
from 0.003 to 0.092 (
=
0.104
−
0.012).
Through the result analysis for the 21 industries, we reveal that
most ratios of instability quickly rose during the financial crisis. Firms
hired a lot of dispatched workers.
2
This phenomenon improved after
the implementation of the Dispatch Worker Law in 2009, as it indeed
enhanced the salaries and reduced the amount of dispatched work-
ers; see
Fig. 1
a. It is also reflected in the relationship between the
ratios of instability and the coefficients of productivity, as the rates
of instability decrease when the coefficients of productivity increase.
However, because the salaries of dispatched workers usually are lower
than equilibrium, when firms lack a labor force they often hire tem-
porary workers to reduce their production cost, which will increase the
degree of upward wage rigidity. In the empirical study, we also discover
that Japan government’s regulation on the dispatched agencies to raise
wages indeed improved the market efficiency and boosted dispatched
workers’ salaries.
2
Utilities (UT) and Compound Services (CS) are rarely affected by a financial crisis,
because they produce necessary goods and services.
160
W.D. Chen
Economic Modelling 73 (2018) 152–162
Table 2
We estimate the model with the data from January 2000 to December 2015, in which we measure the wage equation for each
industry, and find that some of the industries’ productivity coefficients are negative, because their gaps between wage and
productivity are persistent; this is contradicted by the conventional assumption. We measure the ratios of the persistent and
mean-reversion parts
a
for the 21 industries, and the ADF test
b
is used to verify their accuracy.
Industry
January 2000–December 2015
Unstable Part
Stable Part
̂𝜌−
1
̂𝜎
𝜌
̂𝜌−
1
̂𝜎𝜌
Rate
̂𝜌−
1
̂𝜎
𝜌
̂𝜌−
1
̂𝜎𝜌
Rate
CN(L)
−
0.121
0.058
−
2.103
0.186
−
1.156
0.108
−
10.729
0.814
CN(S)
−
0.135
0.055
−
2.454
0.276
−
1.048
0.101
−
10.333
0.724
FI(L)
−
0.441
0.077
−
5.736
0.161
−
1.286
0.107
−
11.98
0.839
FI(S)
−
0.481
0.078
−
6.155
0.15
−
1.273
0.107
−
11.933
0.85
MN(L)
−
0.448
0.07
−
6.383
0.387
−
1.115
0.101
−
11.03
0.613
MN(S)
−
0.353
0.066
−
5.377
0.504
−
1.07
0.107
−
9.953
0.496
UT(L)
−
0.314
0.069
−
4.571
0.242
−
1.273
0.112
−
11.372
0.758
UT(S)
−
0.301
0.077
−
3.93
0.349
−
1.347
0.113
−
11.906
0.651
WR(L)
−
0.167
0.065
−
2.592
0.403
−
1.232
0.104
−
11.859
0.597
WR(S)
−
0.155
0.055
−
2.84
0.562
−
1.025
0.106
−
9.71
0.438
MQ(L)
−
0.418
0.072
−
5.783
0.129
−
1.374
0.109
−
12.557
0.871
MQ(S)
−
0.219
0.066
−
3.309
0.246
−
1.048
0.1
−
10.489
0.754
CS(L)
−
0.164
0.066
−
2.494
0.353
−
1.06
0.101
−
10.531
0.647
CS(S)
−
0.2
0.071
−
2.827
0.5
−
1.107
0.102
−
10.891
0.5
EL(L)
−
0.351
0.071
−
4.927
0.5
−
1.158
0.113
−
10.242
0.5
EL(S)
−
0.423
0.076
−
5.591
0.379
−
1.047
0.109
−
9.617
0.621
IC(L)
−
0.476
0.079
−
6.053
0.055
−
1.017
0.132
−
7.704
0.945
IC(S)
−
0.472
0.076
−
6.196
0.179
−
1.137
0.105
−
10.833
0.821
TP(L)
−
0.154
0.064
−
2.408
0.386
−
1.728
0.155
−
11.118
0.614
TP(S)
−
0.201
0.061
−
3.307
0.469
−
1.382
0.113
−
12.252
0.531
MH(L)
−
0.21
0.064
−
3.273
0.563
−
1.309
0.113
−
11.627
0.437
a
The proportions are measured by the sum of squares.
b
The best BIC is used to choose the optimal lag length, where the critical value for the Dickey-Fuller t statistic is
−
4.02, at the
significance level of 2.5% for the two right-hand variables, excluding the constant term.
Table 3
We measure the ratios of instability (mean reversion part) among three periods, from January 2000 to
December 2007, January 2008 to February 2010, and March 2010 to October 2015, respectively. Most
industries are like the FI industry, where efficiency declined during the financial crisis and then rose after the
revision of the Dispatch Worker Law of 2009.
Proportion of Instability
a
January 2000–July 2007
August 2007–May 2009
June 2009–October 2015
CN(L)
0.3258
0.6666
0.2965
CN(S)
0.5149
0.6728
0.3723
FI(L)
0.1687
0.2910
0.1987
FI(S)
0.1796
0.2766
0.1599
MN(L)
0.3752
0.7171
0.4258
MN(S)
0.4982
0.9241
0.5812
UT(L)
0.2935
0.3054
0.3366
UT(S)
0.4832
0.4261
0.3904
WR(L)
0.5769
0.7365
0.5060
WR(S)
0.7906
0.8850
0.6199
MQ(L)
0.1684
0.3352
0.1273
MQ(S)
0.4608
0.4858
0.2075
CS(L)
0.4923
0.4459
0.5503
CS(S)
0.5834
0.5647
0.6995
EL(L)
0.5581
0.8360
0.5730
EL(S)
0.4198
0.5765
0.4712
IC(L)
0.0484
0.1051
0.0872
IC(S)
0.2438
0.2794
0.1530
TP(L)
0.4673
0.8483
0.5699
TP(S)
0.6057
0.8008
0.6782
MH(L)
0.7958
0.9129
0.5151
a
The proportion of the non-stationary part in the gap between real wage and equilibrium is used to display
the situation of stability. A larger value means a more efficient market.
This research investigates the effect of the dispatched worker sys-
tem in the Japan labor market. With a dynamic equilibrium model, we
analyze the changes in market efficiency during three different time
periods. As the dispatched agencies usually pay salaries below the equi-
librium, this has shaped the wage distribution to be sticky toward a low
level. It also shows that even while the Japan labor market holds a flat
unemployment rate, the market is still inefficient. Upward wage rigid-
ity widely exists in Japan labor markets, and one of the primary reasons
is the dispatched worker system. The evidence also supports the revi-
sion of the Dispatch Worker Law of 2009, as the government proposed
the “the same pay for the same work” and the “least hiring length,”
which indeed did increase market efficiency. Most industries in Japan
have significantly increased their effectiveness since June 2009. This
article provides an insightful interpretation for why the relationship
161
W.D. Chen
Economic Modelling 73 (2018) 152–162
between productivity and wages is often insignificant or even negative.
The results highlight the imbalance of the welfare shares and reveals
the importance of the government policy.
References
Afonso, O., 2016. Effects of labour-market institutions on employment, wages, r&d
intensity and growth in 27 oecd countries: from theory to practice. Econ. Modell.
53, 48–62
.
Aoyagi, C., Ganelli, G., 2015, June. Does revamping Japan’s dual labor market matter?
Econ. Syst. 39 (2), 339–357
.
Bentolila, S., 1992. The macroeconomic impact of flexible labor contracts, with an
application to Spain. Eur. Econ. Rev. 36, 1013–1047
.
Bentolila, S., 1997. Sticky labor in Spanish regions. Eur. Econ. Rev. 41, 591–598
.
Berg, G.J.V.D., Ridder, G., 1998. An empirical equilibrium search model of the labor
market. Econometrica 66, 1183–1221
.
Boeri, T., Jimeno, J.F., 2016. Learning from the great divergence in unemployment in
europe during the crisis. Lab. Econ. 41, 32–46
.
Booth, A.L., 2014. Wage determination and imperfect competition. Lab. Econ. 30, 53–58
.
Burdett, K., Mortensen, D.T., 1998. Wage differentials, employer size, and
unemployment. Int. Econ. Rev. 39 (2), 257–273
.
Cahuc, P., Malherbet, F., 2004. Unemployment compensation finance and labor market
rigidity. J. Publ. Econ. 88, 481–501
.
Chang, J.J., Hung, H.W., 2016. Trade unions, unemployment, economic growth, and
income inequality. Macroecon. Dyn. 20 (1), 404–428
.
Chen, W.D., 2016a, February. Detecting multiple factors in panel data: an application on
the growth of local regions in China. Appl. Econ. 48 (37), 3558–3568
.
Chen, W.D., 2016b, August. Policy failure or success? detecting market failure in China’s
housing market. Econ. Modell. 56, 109–121
.
Chu, A.C., Cozzi, G., Furukawa, Y., 2016. Unions, innovation and cross-country wage
inequality. J. Econ. Dynam. Control 64, 104–118
.
Dobbelaere, S., Kiyota, K., Mairesse, J., 2015. Product and labor market imperfections
and scale economies: micro-evidence on France, Japan and The Netherlands. J.
Comp. Econ. 43 (2), 290–322
.
Dobbelaere, S., Mairesse, J., 2013. Panel data estimates of the production function and
product and labor market imperfections. J. Appl. Econom. 28 (1), 1–46
.
Dolado, J.J., Felgueroso, F., Jimeno, J.F., 2000. Youth labour markets in Spain:
Education, training, and crowding-out. Eur. Econ. Rev. 44, 943–956
.
Elgin, C., Kuzubas, T.U., 2013. Wage-productivity gap in OECD economies. Economics:
Open-Access Open-Assess. E-J. 7, 2013–2021
.
Faccini, R., Ortigueira, S., 2010, August. Labor-market volatility in the
search-and-matching model: the role of investment-specific technology shocks. J.
Econ. Dynam. Control 34 (8), 1509–1527
.
Gaston, N., Kishi, T., 2007, December. Part-time workers doing full-time work in Japan.
J. Jpn. Int. Econ. 21 (4), 435–454
.
Hoffmann, F., Shi, S., 2016. Burdett-Mortensen model of on-the-job search with two
sectors. Rev. Econ. Dyn. 19, 108–134
.
Jahn, E.J., Pozzoli, D., 2013, October. The pay gap of temporary agency workers - does
the temp sector experience pay off? Lab. Econ. 24, 48–57
.
Jovanovic, B., 1979. Job matching and the theory of turnover. J. Polit. Econ. 87 (5),
972–990
.
Jovanovic, B., 1984. Matching, turnover, and unemployment. J. Polit. Econ. 92 (1),
108–122
.
Kahn, L.M., 2012, January. Temporary jobs and job search effort in Europe. Lab. Econ.
19 (1), 113–128
.
Karabarbounis, L., Neiman, B., 2014. The global decline of the labor share. Q. J. Econ.
129 (1), 61–103
.
Krusell, P., Rudanko, L., 2016. Unions in a frictional labor market. J. Monetary Econ. 80,
35–50
.
Kurozumi, T., Zandweghe, W.V., 2012, April. Learning about monetary policy rules
when labor market search and matching frictions matter. J. Econ. Dynam. Control
36 (4), 523–535
.
Mishel, L., Gee, K.F., 2012, Spring. Why aren’t workers benefiting from labour
productivity growth in the United States? Int. Prod. Mon. 23, 31–43
.
Mortensen, D.T., Pissarides, C.A., 1994. Job creation and job destruction in the theory of
unemployment. Rev. Econ. Stud. 61, 397–415
.
Mortensen, D.T., Pissarides, C.A., 1998. Technological progress, job creation and job
destruction. Rev. Econ. Dyn. 1, 733–753
.
Mortensen, D.T., Pissarides, C.A., 1999. Unemployment responses to ‘skill-biased’
technology shocks: the role of labour market policy. Econ. J. 109, 242–265
.
Phillips, P.C.B., Loretan, M., 1991. Estimating long-run economic equilibria. Rev. Econ.
Stud. 58, 407–436
.
Pissarides, C.A., 1986. Unemployment and vacancies in Britain. Econ. Policy 1 (3),
499–559
.
Pissarides, C.A., 1994. Search unemployment with on-the-job search. Rev. Econ. Stud.
61, 457–475
.
Pissarides, C.A., 2000. Equilibrium Unemployment Theory. MIT Press, Cambridge
.
Saikkonen, P., 1991. Asymptotic efficient estimation of cointegration regressions.
Econom. Theor. 7, 1–21
.
Stock, J.H., Watson, M.W., 1993. A simple estimator of cointegrating vectors in higher
order integrated systems. Economerica 61, 783–820
.
Tanaka, A., 2013, December. The causal effects of exporting on domestic workers: a
firm-level analysis using Japanese data. Jpn. World Econ. 28, 13–23
.
Zanetti, F., 2011, June. Labor market institutions and aggregate fluctuations in a search
and matching model. Eur. Econ. Rev. 55 (5), 644–658
.
162
Document Outline
Do'stlaringiz bilan baham: |