IN THE COVERED AND UNCOVERED SECTORS: Evidence from Individual M icro-Level Data
6.2. Previous Minimum W age Studies Using Individual M icro-Level Data
In this section, the previous empirical studies on the effects o f m inim um wage on
wages and em ploym ent using individual m icro level data are review ed separately in
the case o f developed and developing countries. In general, the m ain difference
betw een the m inim um wage in developed and developing countries is that the degree
o f com pliance w ith the m inim um wage policy is higher in developed countries, while
in developing countries, including Indonesia, the com pliance is likely to be low, given
the lack o f enforcem ent and the greater proportion o f the informal uncovered sector
(see Jones, 1997). In Indonesia, m ore than 20% o f total paid em ploym ent is paid
below the m inim um wage level (paid em ploym ent in the uncovered sector).
6.2.1. Developed Country Studies
There are large num bers o f empirical studies in developed countries on the impact o f
m inim um wages using individual m icro level data. One o f the earliest influential
studies o f the m inim um wage effect is the study o f the “natural experim ents”
conducted by Card and Krueger (1994) in the U nited States. Using data from 410 fast-
food restaurants, Card and Krueger investigated the im pact o f an increase in N ew
Jersey’s m inim um wage on N ew Jersey and Pennsylvania (nearby states) em ploym ent
by com paring the em ploym ent, wages, and prices at those restaurants before and after
an increase in m inim um wage. The m ethods used in their study included differences-
in-differences and regression-adjusted model. In practice, they em ployed a N ew
Jersey dum m y variable and the wage gap betw een the new m inim um wage and the
initial wage to m easure the m inim um wage variable. As a result, contrary to the
standard com petitive model, they found that an increase in N ew Jersey’s m inim um
wage raised the em ploym ent in the state, supporting the m onopsony model. Evidence
in support o f a m onopsony m odel was also found for California when Card (1992a)
exam ined the effect o f the state m inim um wage on the young and less skilled workers.
He found that an increase in the state m inim um wage by 10% was associated w ith an
increase in the em ploym ent-population rate o f 2-6%.
A nother influential study in the United States was conducted by N eum ark et al (2004).
U sing an individual level panel data set from 1979 to 1997, N eum ark et al (2004)
em ployed the com plete specification o f the m inim um wage ratio mw, -m w ,
mwt - \
for
each w orkers’ group position in the wage distribution to m easure the m inim um wage
effect. Besides investigating the current effect o f the m inim um wage, they also
estim ated a lagged effect to exam ine the w orkers’ condition one year after an increase
in m inim um wage. As a result, although wages o f low-wage workers increased, they
found that the em ploym ent and hours o f w ork o f the workers who initially received
the m inim um wage level in the previous year (t-1) declined as the m inim um wage
increased, indicating that the m inim um wage was binding for these groups o f worker.
Com pared to the other groups o f worker, these groups were recognized as the most
affected by an increase in m inim um wage.
To com pare the evidence in the United States with other developed countries, Abowd
et al (1997) exam ined the transition probability am ong individual’s em ploym ent status
for youth em ploym ent in the United States from 1981 to 1987 and France from 1981
to 1989 using the individual panel data set. In contrast w ith a decrease in the United
States real m inim um wage, France experienced an increase in the real m inim um wage
that the effect o f a change in the real m inim um wage level in those countries was quite
sim ilar. In the U nited States, a decrease in the m inim um wage by 1% w as associated
w ith an increase in the probability o f the young m en being em ployed o f 2.2% , while
an increase in the French m inim um wage o f 1 % was associated with a reduction in the
probability o f the young m en being em ployed o f 2.5% , supporting the standard
com petitive model.
Currie and Fallick (1996) introduced a “wage gap” variable to identify a group o f
youth workers that was directly affected (“bound”) by an increase in the 1979 and
1980 federal m inim um wage in the U nited States. In this case, the wage gap variable
is defined as the difference betw een the individual’s earning in the previous period
(t-1) and the new m inim um wage level (t), given their previous earnings (t-1) are not
below the old m inim um wage level (t-1). As control groups, Currie and Fallick (1996)
included those individuals with earnings less than the old m inim um wage level, those
individuals with earnings greater than the new m inim um wage level, and those
individuals em ployed in the uncovered sectors. The dependent variable was a binary
variable, w here em ploym ent is equal to 1 if the individual was em ployed in both year
t-1 and year t, otherwise em ploym ent is equal to 0. Using both OLS and fixed effects
estim ation, they concluded that workers who were “bound” were about three percent
less likely to rem ain em ployed in the follow ing year as the m inim um wage increased
in 1979 and 1980.
Zavodny (2000) extended Currie and F allick’s (1996) model by estim ating the effect
o f m inim um wage on the change in wage and the probability o f rem aining employed.
Com paring the data used by Currie and Fallick (1996), firstly, Zavodny (2000) used
CPS sam ple w ith greater sample periods (1979-1993). Secondly, individuals were
classified as the workers affected by the m inim um w age based on an increase in the
real m inim um wage rather than the nom inal m inim um wage. As a result, she found
that an increase in m inim um wage (the larger the wage gap) was m ore likely to
increase the average real wage o f the affected w orkers by US$ 2.31. In addition, using
probit estim ates, she concluded that an increase in m inim um wage decreased the
probability o f rem aining em ployed for the affected youth em ploym ent by 2.2%,
indicating a sm aller effect than in Currie and Fallick (1996).
Y uen (2003) exam ined the effect o f m inim um wages on youth em ploym ent using
C anadian individual level panel data from 1988 to 1990. Similar to Currie and
F allick’s (1996) m ethod, Yuen (2003) em ployed a dum m y variable o f the individual
directly affected by an increase in m inim um wage as the m inim um wage measure. The
dum m y variable is equal to 1 if the individual earning in the previous period is
betw een the old and new m inim um wage level (mwjt-i < Wjt-i < mwjt), otherwise the
dum m y variable is equal to 0. As pointed out by Yuen (2003), the advantage o f using
the individual panel data set is that it m akes it possible to investigate the individual
transition probability from employed status in the previous period to non-em ployed
(or em ployed) status in the following period. His results indicated that an increase in
m inim um wage o f 8.4% was likely to decrease the num ber o f low wage teenager
w orkers by 7% and to decrease the num ber o f low wage young adults by 10%.
A sim ilar study was conducted by Kawaguchi and Yam ada (2007) in Japan. In this
case, they used Japanese individual level panel data with the fixed effect estim ator to
low w age fem ale workers affected by an increase in m inim um wage were 20 to 30
percent less likely to be em ployed in the follow ing year, com pared to low wage
fem ale workers not affected by an increase in m inim um wage.
B ased on the previous m inim um wage studies in developed countries, as discussed
above, we can see that the empirical results o f the m inim um wage effects generally
can be view ed from two m ain fram ew orks, i.e. the standard com petitive and the
m onopsony m odels. In addition, it is interesting to note that, although the m inim um
w age studies are conducted in the same country (for exam ple the United States or the
U nited K ingdom ), there is no clear consensus about the direction and size o f the
m inim um wage effects, depending on the m ethodology used and the object o f study.
This is actually one o f the reasons w hy the m inim um wage studies are still a
challenging topic in the international literature, both in developed and developing
countries. The evidence in developing countries is discussed in the next section where
the m inim um wage com pliance is likely to be low, given the lack o f enforcem ent and
the greater proportion o f the uncovered (inform al) sector. In developing countries,
therefore, the tw o-sector m odel m ight be m ore relevant rather than the standard
com petitive or m onopsony m odels, showing a m ore com plete analysis o f workers in