THEORETICAL REVIEW Human Capital Theory
DO WAGES REFLECT PRODUCTIVITY UNDER MINIMUM WAGES SETTING?
Evidence from Indonesia Firm Level Survey
Rifai Afin
Universitas Trunojoyo Madura Abstrak
Penelitian ini menginvestigasi pengaruh pendidikan pada produktifitas tenaga kerja dan upah pada skema kebijakan upah minimum pada sektor manufaktur di Indonesia. Menggunakan data tingkat perusahaan, studi ini mengidentifikasi faktor yang mempengaruhi output dan upah pada tingkat perusahaan dan menggunakan data tingkat regional untuk menginvestigasi pengaruh upah minimum pada perubahan upah.
Temuan pertama, tenaga kerja dengan 9 dan 12 tahun pendidikan mengalami diminishing marginal produktifitas dan tingkat pendidikan lain konsisten dengan human capital theory dan sebagian besar varibel kontrol secara positif berpengaruh terhadap output dan upah. Kedua, pendidikan dan variabel lain mempengaruhi output dan upah dengan pola yang sama yang mengindikasikan dengan kuat bahwa pengusaha membayar upah berdasarkan kinerja para pekerjanya. Ketiga, upah minimum secara statistik signifikan mempengaruhi upah. Hal ini menunjukkan bahwa upah minimum adalah pedoman bagi pengusaha dalam memberikan pembayaran minimum atas jasa pekerja, tetapi jika pekerja ingin lebih maka harus berbuat lebih baik.
Kata kunci: Produktivitas,Upah, Kebijakan Upah Minimum
policy. Considering economic situation In competitive market, wage is a
INTRODUCTION
such as price or inflation, economic result of market clearing process of the
unemployment, labour market, the intersection between
growth,
and
government formulate nominal wage labour supply and demand function.
Market competition Nominal wage is as a result of
moderately.
determines wage through interaction multiplication of product price and
between employer, in the demand side, marginal productivity of labour or in
and workers in the supply side but in the other words is the value of marginal
minimum wage law context, this may productivity of labour. Increasing
not always work.
nominal wage might reflect increasing Wage is the value of labour price or marginal productivity of labour,
productivity. Theoretically, Education which is in the demand side, can make
both formal and informal, working movement along demand function of
experience, age, and wage with its labour downward. In the contrary,
endogeneity have responsibility for labour supply function has positive
workers quality reflected in their slope which means that the higher the
productivity. Many studies prove the wage the higher the motivation of
effect of education, labour to join in labour market.
significant
experience, and job training on earnings Government, as the regulator, sets
that is commonly called returns to nominal wage that satisfies both firm
education for examples Angrist and and labour through minimum wage
Newey (1991), Vaillancort (1995),
Rifai Afin, Do Wages Reflect Productivity Under Minimum Wage Setting? 71
Kling (2001), Blackburn and Neumark minimum wage set by the local (1995), Altonji (1993), Harmon and
government compared to productivity Walker (1995), Ashenfelter and
and nominal wages?, and does the Zimmerman (1997), and Bonjour,
minimum wage setting affect wages?. Cherkas, Haskel, Hawkes, and Spector
This paper is structured as (2003). This evidence shows that
follows: the following section describes education has important role to generate
theoretical foundation on how education earnings but according to Chevalier,
affects productivity and earnings and Harmon, Walker and Zhu (2004), this is
how minimum wage policy affects not clear whether the effect of education
wages. Section three explores data and on wages is because educated workers
empirical investigation method by have high productivity or signalling of
econometric model, education level on worker’s ability.
proposing
estimation techniques, the strategy how Some other studies such as Besen
the outcome of empirical methods (1968), Griliches (1968), Khaldi (1975),
answer the questions, and data sources. Lockheed, Jamison, and Lau (1980),
Section four explores the results of Pudasaini (1983), and Weir (1999),
empirical study examined in section investigate effect of education on
three. Last section, five, brings us to the productivity.
concluding remarks. Most studies on the relationship of education, wage, and productivity have
sort of lack the interconnection of this THEORETICAL REVIEW three components. In the one side, some Human Capital Theory
Human investment firstly has not studies partially focus on the obviously considered as a kind of relationship between education and real capital accumulation process until labour productivity but in the other side Schultz (1961) and Becker (1962) concern about the effect of education on formally identified education as an earnings by arguing that earnings activity affecting future income. represent labour productivity. Jones Education is basically general concept (1994) tried to identify the effect of of which does not influence income; the education on both productivity and truth is skills that generate income. earnings in industrial sector. This paper Correa (1963) defines skills as the contributes to the further question by learned responses that contribute to the investigating
the interrelationship efficiency of behaviour. Education is between the triangle of education, only a formal process of acquiring skills productivity, and earnings as well as in educational institutions like schools how minimum wage set by the or colleges. Even though workers have government behaves in the two labour improved their productivity, this does performance indicator, productivity and not automatically raise wages unless earnings. More specifically, this paper workers could convincingly threaten to has several questions to be addressed. get a job elsewhere (Cahuc and Firstly, how is the effect of education Zylberberg, 2004). In equilibrium and other control variables on labour competitive goods and labour market, productivity and earnings? Secondly, do the labour income or wage is the same other factors, besides education, have as value of marginal productivity of the same pattern as predicted by level of labour which is called marginal education on affecting productivity and productivity theory. Regarding this wages? Third, where is the position of theory, Addison and Siebert (1979) the interrelationship efficiency of behaviour. Education is between the triangle of education, only a formal process of acquiring skills productivity, and earnings as well as in educational institutions like schools how minimum wage set by the or colleges. Even though workers have government behaves in the two labour improved their productivity, this does performance indicator, productivity and not automatically raise wages unless earnings. More specifically, this paper workers could convincingly threaten to has several questions to be addressed. get a job elsewhere (Cahuc and Firstly, how is the effect of education Zylberberg, 2004). In equilibrium and other control variables on labour competitive goods and labour market, productivity and earnings? Secondly, do the labour income or wage is the same other factors, besides education, have as value of marginal productivity of the same pattern as predicted by level of labour which is called marginal education on affecting productivity and productivity theory. Regarding this wages? Third, where is the position of theory, Addison and Siebert (1979)
human capital is that the assumption of ability of
education
or
accumulation has significant effect on employers to measure marginal cost and
increasing real productivity of labour marginal revenue is implausible and this
and wages as well. At this level, the is supported by empirical evidence by
connection of education to productivity Lester (1946) who strongly attacked the
and wages seems not straightforward marginality concept. The argument to
like general views on this particular refuse marginal productivity idea is the
topic. In the first line, a problem starts assumption of profit maximization.
from education effect on productivity. They argue that this assumption is
In this step, trouble arises when we invalid in the general areas of economy
define education such as general or especially in the public sector. This
specific training (Becker, 1975) and argument seems to be a common issue
skills measure (level of education or that shows the limitation of market
scoring test). In the second line, the clearing process. The third objection of
problem is how to represent accepting marginal productivity theory
productivity in terms of wages like is about the nature of technology. They
previously discussed that there are four argue ‘’specific marginal product value
objections to marginal are not related to specific employment
main
productivity theory so that it is levels but to ranges of employment’’. At
presumably more complicated. this rate, the marginal productivity of
We have discussed the effect of labour positively depends on the
human capital in micro economic employment ranges (discontinuity of
perspective accompanied by some labour demand function).The last but
contradictory ideas and empirical not least of the main objection of
the other side, marginal productivity theory. According
studies.
In
macroeconomic studies on human to Oi (1962), labour is not perfectly
capital theory have slowly but sure have variable but has fixed cost in part, such
been increasing. Some economists as hiring and training cost, which is
investigate human capital theory in called labour as a quasi-fixed factor,
macroeconomic perspective e.g. Nelson consequently, employer rationally
and Phelps (1966), Jorgenson and decides allocation of labour cannot only
Griliches (1967), Griliches (1963), and
be based on the relation of wages and Lucas (1988) that pioneered to marginal value of product but also fixed
investigate theoretically and empirically cost that must be paid in the future of
how the quality of labour affects that quantity of labour. These four main
economic progress through increasing objections of marginal productivity
productivity. According to Aghion and theory are very critical in this particular
Hewitt (1998), the human capital theory research.
can be divided into two different views Regardless the objections to
which are Nelson-Phelps approach and marginal productivity theory, many
Lucas approach. Delsen and Schone- studies, some of them have already been
wille (1999) argues that in the Nelson mentioned in the section I, concern
and Phelps approach, economic growth about empirical test on the factors
is more depending on the accumulation affecting marginal productivity and
of human capital endowment relative to wages in micro economic level that is
the current human capital accumulation. basically from marginal productivity
This means the higher the level of This means the higher the level of
quality of the data. Cohen and Soto use Moreover, in this approach, the only
the survey data in which the factor of production in the production
classification of education system is function is physical capital or past
uniform. At this stage, we may conclude innovation. Therefore, if the physical
that the effect of education on economic capital increases economic growth, it
growth faces more technical problems means that that human capital which is
rather that debating among existing included in the innovation has
theories.
significant effect on growth. In the other side, Lucas (1988)
Signalling Theory
proposes model focusing on mechanism The most courageous theory of the effect of human capital on
challenging human capital theory is economic growth through schooling and
signalling theory. Basically, signalling learning by doing. Workers who are just
theory assumes that there is no clear hired perform their job depending on
information on individual worker when the skills that they bring before being
employers hire them and in the short run hired but in the meantime they learning
after hiring them. This means there is and experience that make them perform
asymmetric information in the process better. According to Delsen and
of transaction in the job market. In other Schonewille (1999), Lucas identifies
words, education does not inform that there is direct effect of human
employer the true ability of workers capital on production process, and
therefore the relationship of education assuming that human capital is another
and productivity and earnings might not factor of production process besides
be causal (Cahuc and Zylberberg, labour and physical capital. Thus, the
2004). According to Spence (1973) higher the level of human capital stock
employers only identify workers based the higher the economic growth.
on their personal data, such as education Estimating the effect of human capital
level, race, sex, and age rather than effect on the production function which
measuring marginal product of workers. is based on the Lucas approach will be
Besides that, there are always some more straightforward that Nelson and
potential sources of information that are Phelps approach.
needed by employers to categorize their However, empirical tests of
wages such as previous job position, human capital theory in the
wages, criminal and or reference from macroeconomic context result in
previous employer, medical condition. different conclusion among studies.
Let us discuss what Spence (1974) Some studies such as Benhabib and
had done in explaining signalling theory Spiegel (1994), Barro (2001), and Islam
formally. Suppose there are two groups (1995) conclude that there is no
of people with different job and significant effect of human capital
different productivities and both of them indicator on economic growth. In the
can do investment in education as well contrary, Temple (1999), Ciccone and
as the costs of education between those Papaioannou (2005), Cohen and Soto
two groups are different. The years of (2006) in their studies summarize that
schooling for group 1 and 2 are E1 and human capital has significant impact on
E2 and schooling costs for both groups economic growth. Some arguments that
are CE1 and CE2 respectively and C2 < have been proposed regarding these
C1 by assumption. For separating the C1 by assumption. For separating the
will happen in the signalling theory. In (1-a1) or a2 be the proportion of people
signalling, productivity of workers does in group 2 and there are two kinds of
not depend on the education, so that we jobs in the labour market for
can write f i1 =f i2 =f i for group i = 1, 2. If simplifying. Productivity of group i in
E* is the number of schooling that job j with education E is represented by
satisfies the inequalities: the f ij (E).
f 2 f 1 f 2 f 1 Some assumptions need to be
(1) c 1 c 2 imposed to make the model works.
Firstly, employers make decision on By assumption, f2>f1 and c2>c1 and hiring workers based on their
the wages offered if f1 if E<E* and f2 if observation on obtained education of
E ≥E* the equilibrium can be written in workers without knowing their
the Table 1 as follows. productivity. Secondly, in the labour
The essential finding from this side, workers find the job that is suitable
model is that the private and social with their expected productivity
returns to education differ. As its conditional on their education level and
consequence, group two invest more in finally, in the equilibrium of the job
education whereas the optimum is just market, labour receives income in
when E=0 for both groups. Wages which their productivity is used as a
would be 1 f 1 2 f 2 for every group. It basis. In this such situation, individual
benefit group 1 but not group two therefore invest in education based on
because group 1 must invest more in what they expect from future income
education. In signalling, people pursue and allocates their educational fund as
more education not because of they will much as the amount of expected
be paid more but to distinguish among income.
others.
Table 1. Pure Signalling Equilibrium
Group
Wages Education E=0
Productivity
Education
Spending Group 1
E = E*
Cost
f1 f1 E1 f1 0 Group 2
f2 f2 E2 f2 C2E* Note: Bold indicate equilibrium productivities
Table 2. Pure Human Capital Equilibrium
Group
Wages Education E=0
Productivity
Education
Expenditure Group 1
E = E*
Cost
f1 f1 E1 f1 0 Group 2
f2 f2 E2 f2 C2E* Note: Bold indicate equilibrium productivities
In human capital model, assume on this particular theoretical and that f ij ( E ) f ( E ) for both group and
empirical study.
both kinds of jobs and consider that
f ( 0 ) f 1 and f ( E *) f 2 as well as E
DATA AND METHODOLGY
satisfies the equation 2.1 above then the
Research Framework
equilibrium is similar to in pure The strategy which is used in this signalling model as shown in Table 2
study emerges from the basic idea that and the only difference from pure
employer should know how qualified signalling is that off diagonal terms so
their employees represented by their that it has another implication which is
productivity that their wages can be wages. In human capital equilibrium,
based on. By identifying the effect of wages is offered at different levels
education on real productivity and depending on the level of education.
wages and make comparison of those Educated worker is more productive.
effects we may come to the conclusion that, intuitively, employers, rationally,
Minimum Wage Setting
pay their workers based on their In market clearing mechanism,
productivity regardless initial working wage is determined by the value of
contract, employers can make revision marginal productivity of workers
of contract based on the evaluation of themselves. Employers offer wages
workers actual productivity. If there is a based on their observation to their
worker who is caught shirking or the worker’s performance. The government
worker performance is not like what could do intervention labour market
they expect, the employer will make process through setting the minimum
some revision on wage payment so that wage policy.
the wage must always represent the minimum wages as anchor for price
Keynesian views
actual productivity of workers. The level. Minimum wages force the wage
question is whether the effect of structure and cause movement of
education on productivity increasing income distribution among workers. In
with the level of education and does the the contrary, classical view of minimum
effect of education on productivity is wages argues that minimum wages have
followed by the proportional increase in negative effect on employment. Some
wages. The next question is where does empirical studies confirm those theories.
the minimum wages that has been set by Bryan, Salvatori, and Taylor (2012) find
the government lies on between value of that minimum wage negatively affect
productivity and received wages. the earnings for young workers even
though they still question the imprecise
Data Sources
estimation results. Maloney and Mendez This study explores cross section (2004) support empirically that
data of manufacturing survey and labour minimum wages strongly affect real
individual survey and regional level wages. Meyer and Wise (1983) find that
panel data. First part of this study is there is no earnings effect because of
estimating factors affecting output and minimum wage changes. Bazen and
wages from characteristics of the firms, Martin (1991) find that minimum wages
and in this part, cross section data from has increased real wages of youth
firm level survey are used. The data of employment but it could have negative
manufacturing can be collected from effect on youth employment as well.
two kinds survey those are economic Until now, there is no single consensus two kinds survey those are economic Until now, there is no single consensus
and the only census of manufacturing typically designed but in economic
that can be accessed is census in 2006 census we will find more information
containing 25,694 firms with medium such as level of worker’s education and
and large scale of production. firm administration status but this
Table 3. Variable Statistic Summary
Minimum Maximum Variable
Std. Dev. value Value Male workers with < 6 year schooling (MNSYS) 20.95455 31.64908
Mean
1 125 Female workers with < 6 year schooling (FNSYS)
1 112 Total workers with <6 year schooling (TNSYS)
5 159 Male workers with 6 year schooling (MSYS)
2 515 Female workers with 6 year schooling (FSYS)
5 1480 Total workers with 6 year schooling (TSYS)
17 1995 Male workers with 9 year schooling (MNYS)
2 869 Female workers with 9 year schooling (FNYS)
4 2262 Total workers with 9 year schooling (TNYS)
15 3131 Male workers with 12 year schooling (MTYS)
5 2167 Female workers with 12 year schooling (FTYS)
1 4509 Total workers with 12 year schooling (TYS)
9 6393 Male workers with university degree (MUNIV)
1 245 Female workers with university degree (FUNIV) 17.27273 27.37775
1 123 Total workers with university degree (TUNIV)
2 287 Wages
791857 9.22E+07 Wages and any additional payment
2.01E+07 2.73E+07
793748 1.28E+08 Ratio of imported raw materials to total raw materials (IMP)
2.51E+07 3.47E+07
0.534171 0.304058 0.003029 0.993053 Output
4857154 2.14E+09 Capital/assets
3.96E+08 6.83E+08
1.84E+11 5.85E+11
39929 2.73E+12
In the other side, annual needs some estimation techniques and manufacturing survey does not contain
we cover this by implementing OLS that kinds of information. Mostly, other
(Ordinary Least Squares). The estimated studies use only labour individual
models of production function (Cobb- survey that cannot describe the true
Douglas) are described below: productivity of labour; moreover it
(2) cannot be compared to wages. The
Q = f (X, Z, S)
second part of this study is estimating Q represents for output variable, X the effect of minimum wage on low job
denotes input variables, Z is firm position or under supervisor wages in
Characteristics, and S denotes schooling manufacturing sector using panel
variables. In this part, the strategy used regional level data which contain eight
is estimating separately output and years and four regions those are region
wages function between male and
1 (West Java, Jakarta, Banten), region 2 female workers and total workers per (Middle Java, and Jogjakarta), region
level of education and the role of non three (East Java and Bali), region 4
labour inputs. Separation of gender aims (Sumatra, Kalimantan, Irian Jaya or
to see clearly whether gender has papua, and Sulawesi). This panel data
important role on production while are provided by Central Bureau of
separating other factors such as capital Statistics (Badan Pusat Statistik). Some
and energy. The mathematical model 2 supporting data for the qualitative
can be derived into econometric models assessment are taken and explored from
in the logarithm as follows:
some documents of Ministry of Labour
LnQ 0 1 lCapital i 2 lMNSYS i 3 lFNSYS i
and Transmigration.
4 lMSYS i 5 lMSYS i 6 lFSYS i 7 lMNYS i 8 lFNYS i 9 lMTYS i 10 lFTYS i 11 lMUNIV i
Econometric Model
12 lFUNIV i 13 lenergy 14 IMP i
There are three types of estimated
( 3 models for this study. The first is output )
i Dcontrol
models for measuring marginal There are 13 control variables included productivity of workers. The second is
in the models to make sure that there is wages function that has the same
no omitted variable bias. Since this predictors with output function. Both
study aims to identify how well models are the same, only have different
marginal productivity proxy wages and, dependent variables, because, by
the model for estimating output and assumption, marginal productivity
wages is the same model even though reflects wages but even though they
they could have different meaning on have the same explanatory, some of
the model. The wages model is as variables have different meaning. Jones
follows:
(1994) also runs similar model to
Lnwages 0 1 lCapital i 2 lMNSYS i
compare output function and wages and
3 lFNSYS i 4 lMSYS i 5 lMSYS i 6 lFSYS i
only removes capital in the wages
7 lMNYS i 8 lFNYS i 9 lMTYS i 10 lFTYS i
function. The empirical models applied
11 lMUNIV i 12 lFUNIV i 13 lenergy 14 IMP i
in this study are taken from the
production function and wages function.
i Dcontrol i ( 3 . 3 ) ( 4 i )
Both types of model are estimated in 1 The third model is used for
several forms of nested and non nested estimating the effect of minimum wages models containing five models each on earnings in manufacturing sector. type function. Running these models Bazen and Martin (1991) applied several forms of nested and non nested estimating the effect of minimum wages models containing five models each on earnings in manufacturing sector. type function. Running these models Bazen and Martin (1991) applied
approach, graphical analysis are used to also Neumark, Sehweitzer, and Wascher
describes how wages, education, (2000) approach using two time
productivity, and minimum wages difference for estimating minimum
interact and intuitively interpreted, wage effect on wage distribution. In this
quantitative approach applies some case, difference in wages is used as
econometric models and tests the effect dependent variable and current
of education on productivity and wages. minimum wages and its lag 1, price level, and lag 1 of wages as explanatory
Qualitative Assessment
variables. If price level is significant it Labour Education Level means that the wages changes are partly
Education per labour force caused by price changes and as its
represents the availability of quality of consequence, wages changes do not
human resources in manufacturing fully increase welfare. The model is
sector. Table 4.1 draws the distribution written below:
percentage of worker’s education level
of total labour and manufacturing sector 4 PI i , t 1 wages i , t
lwages it 1 MW it 2 MW i , t 1 3 PI it
5 1 i u it (5)
which shows, on average, manu-
facturing is more educated than total ∆lwages is the wages changes, MW is
national workers but the pattern is minimum wage, PI denotes price level,
quietly the same, most workers, almost αi is the unknown intercept for each
80 percent have secondary school. entity, u it is the error term. The model is
There are some meanings that can be already specified in fixed effect model.
proposed based on this fact. Firstly, manufacturing sector may tend to use
RESULTS AND ANALYSIS
superior technology relative to other
Empirical Facts of Education,
sectors,
with
higher potential
Earnings, and Minimum Wages in
productivity, profits and wages (private
Manufacturing Sector
returns).Tertiary is university degree This chapter is divided into two
workers, Ssec and diploma is senior parts, those are qualitative and
high school and diploma degree of quantitaive or formal test assessment of
workers, Jsec is Junior school degree, wages, education, and productivity, as
elem denotes elementary school, <Elem well as minimum wages policy both in
denotes having no education degree. total labour markets and manufacturing
sector generally known as returns to
Graph aph 1. Distribution of Worker’s Education Leve vel
Tertiary
Ssec and
Jsec
Elem
<Elem m
Diploma
Total Labour
Manufacturing
Source: Calculated fro from Labour Individual Survey and Economic Cens nsus of Manufacturing , 2006 06
productivity is increasi asing from 2006- consequence of the first a t argument, in
The second
me meaning,
as
2007 with the increase se 54 percent in the context of returns to ed education and
2006 and 77 percent in in 2007, and in human capital theory, m manufacturing
2006 decrease at 26 p percent , while may have the higher ave verage labour
wages increase 28 perce rcent in 2006 and income, because the higher er the level of
39 percent in 2007. A And follow the education the higher the pro productivity of
productivity changes at 2 t 27 percent. workers will generates hig higher income.
In perfect comp petitive market, We will come again to t this issue in
wages are fully determi mined by worker section econometric approac oach .
productivity but graph 2 2 tell us different thing.
That is th the change of Labour Productivity and Wa Wages
productivity is always f s followed by the Strengthening eviden dence of the
changes in wages paid b d by the firm and effect of education on labour
the increase of produ ductivity is not productivity in manufactu cturing sector
responded by the incr crease of wages can be investigated from the the firm labour
proportionately. Some r e reasons may be services expenditure in term erms of labour
suspected, regardless ss the market productivity changes. Reas easonably, the
structure, the measur sure of labour increase of productivity ty should be
productivity used in tabl able 2 ignores the followed by the increase ase of wages
contribution of other f r factors such as expenditure, in other word ords, employer
capital, raw materials als, and energy pays workers based on th their working
because the value of f productivity is performance. Graph 2 depi picts the trend
derived from value of production of percentage productivity ity and wages
divided by the number ber of production from 2006-2008. Percentag tage change of
workers.
Graph 2. Percen entages of Change in Wages Expenditure and P d Productivity
Source: Calc alculated from Manufacturing Economic Census, 20 2006 -2008 This method cannot estimat ate the role of
district and municipal, w which are based workers and other factors o s of production
on proper living need eds. Ministry of accurately and independ ndently. The
labour and transmigrat ration issues the second reason, technically, lly, each firm,
minister rule no 17/ Me Men/8/2005 about has some positions of for th their workers,
technical explanation o of proper living not only production but but also non
needs of workers that m must be included production department t such as
in the labour services es payment. The marketing, finance, human an resources,
amount of payment is c s calculated from
the annual field surv urvey which is production but it is diff ifficult to be
that contribute to the he value of
conducted by the team b built by the local included in productivity ity measuring
government involving ing employers, process. The third reason n is minimum
aca cademics, and wages policy. Pricing pol olicy such as
labour
union,
Formal ally, provincial minimum wages distor torts market
government.
government announce m minimum wages mechanism in th e labour ma markets. Three
based on the lowest wag ages in distric and parties in the labour mar arkets, firms,
municipal in that province p after labour, and government, hav have their own
accomplishing price su survey in each expectation to the price ce of labour.
local area of districts. Government intervention in in the labour
Local government nt set the nominal markets by setting minim nimum wages
wages each year and i it is common ly affects labour markets thro rough demand
increasing reflecting the price of and supply side. In the n next section,
commodity movement. nt. Study of the estimation on productivity a y and wages is
effect of minimum wage ages in Indonesia, done by applying econome metric models.
Smeru (2001), shows t s that this affects Even though there is dif differences in
labour markets especi ecially from the changes but the pattern is th the same
demand side. This study dy concludes that the increase of 10 perce rcent of minimum
Minimum Wages Policy wages decreases the e labour force Indonesia governm ment set the
absorption in labour r markets by 1 minimum wages in the e local level,
percent. Since decentr ntralisation fiscal percent. Since decentr ntralisation fiscal
wages increase gradual ually from 2006- higher. Something that t cannot be
neglected is political reason son behind this
Graph 3. Minim imum Montly Wages Average of Provincial Le Level
Source: Calculated d from Data of Ministry of Labour and Transmig igration and BPS, 2009
Graph 4. M Montly Minimum and Manufacturing Wages
2008 Manufacturin turing
Minimum Wages Average
Total Labour our
Source: Calculated d from Manufacturing Economic Census and Mi Ministry of Labour and Transmigration n
Table 4 depicts the trend of translog but ignoring the education level minimum wages and manufacturing
and gender by aggregating skilled and sector. Clearly pictured that wages in
unskilled labour. The method proposed manufacturing sector is higher, on
by Moretti, translog estimation, may average, than the wages which are set
work properly in the small scale model by government and total labour. In the
which means the model contains previous part, told us that labour
relatively small number of dependent education level in manufacturing sector
variables. Translog approach derives is higher than total labour, which
main variables into additional standard probably factor that generates higher
variables in the translog model such as labour income in manufacturing or
squared and multiplication between temporarily we may conclude that
main variables. In this paper, I use productivity of labour in manufacturing
estimation instead of sector is higher than total labour.
direct
transformation logarithm, translog. Ordinary Least Squares (OLS) is applied to all models to get unbiased
Econometric Results
and consistent parameters. Estimation Results As clearly shown in the table A1, This is the second part of generally, all models indicate the same empirical facts that contains the results; all main parameters which are estimation results of econometric related to the effect of education on models followed by some tests of the labour productivity are statistically models. Those tests are hetero- significant except for female workers skedasticity, multicollinearity, model with low level of education, not specification, and normality which are completed and only six years schooling the critical tests for cross section data. we will come again to this result in next The first part is estimation results of section. In addition, some formal tests production function which is shown in to check the models performance such the table A1 and A2 in appendix. as Breusch-Pagan /Cook-Weisberg test In the output function, there are for heteroskedasticity, ovtest for four models to be alternatives which Regression Specification Error Test have the same dependent variable but (RESET), Variance Inflation Factor different predictors. The predictors are (VIF) for collinearity diagnostic, and estimated separately among the models Saphiro and Wilk test for normality of based on the gender comparison per error disturbance have been done and level of education and factors of presented in Table . 4. According to the productions besides labour such as Table 4 there is no serious problem on capital and energy used. Moretti (2004) the estimated models in terms of the estimates the effect of education on
assumption tests. productivity at firm level using both
regression
direct estimation on Cobb-Douglas and
Table . 4. Summary of Classical Assumption Tests
Tests
Multicolinearity Output
(Min and Max Models (Probability Chi Square)
Homoskedasticity
Model Specification Normality
(Probability Z) VIF) 1 0.057
(Probability F)
1.3 and 2.2 Wages
Note: the level of significant of tests is 5 percent and VIF is under 10
In the other side, wages functions Firstly, in all models labour in all levels (see table A3 and A4) perform as we
of education have significant effect but expected as well. The estimation
for junior and senior high school strategy applied in the wages function is
workers, they tend to have diminishing providing four models with two kinds of
marginal productivity. It means dependent variables those are wages and
additional labour with 9 and 12 years wages plus and two different groups
education contributes to the decrease of gender separation like what output
output changes. Lewis (1954), in the function does in output model
theory of dualism, supported this estimation. Separating wages and wages
finding by arguing that in countries plus function is aimed to know whether
where unlimited supplies of labour and additional bonus and payment besides
large population relative to capital exist, wage is also attached to the
marginal productivity of labour could be performance in production process. In
negligible, zero, or even negative. Even order to obtain the coefficient of
though this argument still has some parameters, I apply the same technique
further question, how we measure the as estimating output function, OLS. All
relativity of one production factor to the wages functions show that there is
others is still need to be redefined. In significant and positive effect of labour
other words, diminishing marginal per level of education. The assumption
labour productivity in manufacturing tests for OLS are also run and the tests
sector especially for labour with junior result in no deviation from hypothesis
and senior high school certificate could which means estimated parameters are
exist because labour with that level of unbiased and consistent which are
education is relatively more abundant presented in table 4.
than other production factors that they can work with, such as capital and
Results Interpretation energy. In Indonesian manufacturing The estimation results for
sector, the composition of drop out (<6 production function have some findings.
year schooling) workers only 2.2 year schooling) workers only 2.2
The third international aspect is foreign with 9 years schooling, 49 percent
investment status that does not workers with 12 years schooling, and
surprisingly
have positive and
2.5 percent with college and university significant effect on output. In the other degree. This composition, at least,
side, domestic status of firm has also emphasize our results that is why
diminishing in our sample. Foreign workers with 9 years and 12 years
companies, which commonly use more schooling might have diminishing
advanced technology, still have more marginal productivity. Jones (1994)
capability of doing more. found the same results when estimated
Other variables that are important both production and wages function
role for manufacturing output are which resulted in inconsistency when
location and standardised product. the results show that workers with lower
Location, as we expected, positively level of education are not always have
significant which means industrial state higher productivity. Moreover, Jones
increases manufacturing output. Some argues that in that case, education has
studies that support this results are Lall, indirect effect on labour productivity.
Shalizi, and Deichmann (2001), and Fan According to two first findings, it is
and Scott (2003) that have found difficult to say that production function
empirically that agglomeration has clearly point out how human capital
positive effect on industry performance theory works. Pudasaini (1983)
and also Fu and Ross (2010) who prove confirmed that, in his results, there was
that industrial agglomeration has the decline contribution to output by the
positive effect on wages. higher educated workers even though on
Administrative factors which are average the role of education on output
legal forms and standardisation of is positively significant.
product have also important role on Other variables characterizing the
output and wages. Legal status of the manufacturing sector output are
firm such as Government Company, international aspects those are export,
cooperatives, limited import, and foreign investment.
corporation,
partnership, etc has positive effect on Imported raw materials have positive
both output and wages and it shows that effect on output which means that the
the parameter of corporation status is higher imported raw materials the
the highest among other legal status. higher the output. Even though not
This is not surprising finding. In the many respondents or about 20 percent
other side, standardisation significantly of respondents import their raw
affects output. Intuitively, this is easy to materials, it does matter. In other words,
understand that highly standardized imported raw materials have relatively
products are usually produced by high high productivity in production process.
performing firms so that product In the other side, export has also
standardization is a good way to boost positively significant impact on output
manufacturing productivity. supporting study done by Sjoholm
In the wages function side, the (1997) that had proved the positive
results show us what we expected effect of import and export from
before. Firstly, the effect of education Indonesian manufacturing data even
and training on wages tells us that the though there is no clear consensus
pattern of parameters per level of among theoretical and empirical studies
education are the same as the pattern of education are the same as the pattern of
when
marginal
productivity of labour at a level of education decreases the wages decrease as well. Regardless job contract that is signed before knowing true productivity of workers, the expected of employers to workers performance is correct. In the other side, trained workers are paid more than that non-trained. Most firms in Indonesia apply some condition to their workers for some months before they get an extended contract and get more payment. One of the conditions is training both on the job and off the job training. For workers who have passed the training they get some beneficial including higher salary, bonus, and insurance. Employers, basically, will not know the true marginal productivity of workers but the signal in the labour markets which is easily recognised that most workers, as previously mentioned, who search jobs in the labour markets have junior and senior high school certificate so that employers decided to
accept such typical workers that make, in the production function, capital and other production factors optimally used for that education level are decreasing. From the graph 4.5, which is based on model 3 estimation results, it is clearly illustrated that the changes of the number of low education level than the changes in their productivity compared to high skilled labour with university degree and the effect of number of workers with junior and senior high school certificate on output is relatively lower than others and it is followed by the effect on wages. It exists because of the diminishing marginal productivity. This argument is supported by the fact those level of education are the biggest composition in manufacturing sector, but it does not mean that the nominal value of received wages for 9 and 12 year school is lower than 6 and <6 year schooling because it represents percentage of changes (variables estimated in logarithm). The pattern of those parameters from two functions is depicted in the Graph . 5 below:
Graph 5. The Pattern of Effect of Education on Marginal Productivity and
9 years 12 years university degree
Effect on Output
Effect on Wages
Secondly, capital and energy have status. Table A.4 summarised that those positive effect on wages which means
three variables are positive different capital and energy are complementary
from zero. Firms located on industrial factors of workers. Capital and energy
state pay higher than that of outside are complementary factors in our
industrial state. Indonesia government sample, machinery use energy in the
built some industrial states in some production process and if the capital and
cities such as Surabaya, Pasuruan, energy combination substitutes workers
Tangerang, and Bekasi which are in so the more intensive the combination
most crowded island, Java, and used the less workers will be hired and
spreading them out to other island such the expenditure for wages decreases as
as Batam in Sumatra Island, and Timika consequences of using machinery and
in Irian Jaya and some other are still in energy intensively. Generally, we may
progress. Most of firms located in conclude that workers are needed to
industrial states are big companies and operate machinery and for some cases
well publicly known as companies that some job positions related to machine
have high standard products. To sum operation cannot be replaced by other
up, graph A1 depicts the impact of machines such as installation and
control variables on output and wages maintenance.
based on model 3 estimation. The graph International
shows us that most of the effects of important role in determining industrial
factors
play
those variables on wages follow the wages. Import, exports, and foreign
trend of their marginal productivity. investment have positively significant effect on wages. In empirical literatures,
Effect of Minimum Wage Setting on there is no consensus on what effect of Manufacturing Wages import and export on wages. One When the effect of education and supporting finding is from Martins and other control variables on output and Opromolla (2009) using firm level data wages has the same pattern, so where proved that export and import have the minimum wages does lie?. Section positive impact on workers salary. In
4.1.3 provides qualitative explanation of the contrary, Alvarez and Opazo (2008) minimum wage and manufacturing used firm level data and found that labour earnings, this part apply imported goods have negative effect on econometric model instead. Table 4.3 domestic firm wages. The other shows us the effect of minimum wage international aspect is status of on manufacturing wage changes. By investment, which is foreign ownership, focusing on fixed and random effect does positively affect on wages. It is models, the results show that minimum generally
wages and lag of wages are different management system is better than that from zero at 10 percent level. of domestic so that it is not surprising Considered by the sensitive results of results that foreign pays more than fixed and random effect model, domestic firms. This result is shown in specification test, Hausman test, need to table A3 conclude that foreign
be applied. According to Hsiao (2003), ownership firms pays more than the issue of specification test is not domestic firms. whether or not individual effect fixed or Other variables that are expected random, but more important thing is to have positive effect on wages are individual effect can be considered as a location, standardisation, and Legal random draws from a common be applied. According to Hsiao (2003), ownership firms pays more than the issue of specification test is not domestic firms. whether or not individual effect fixed or Other variables that are expected random, but more important thing is to have positive effect on wages are individual effect can be considered as a location, standardisation, and Legal random draws from a common
distribution of individual effect and 2
LM T r ij
i 2 j attributes can be viewed as identical 1 across panels and suggested Hausman
Where r 2 ij is the i jth residual correlation test as an alternative method to identify
coefficient. The Breusch and Pagan satisfying model. Hausman test result (1980) test statistic is distributed 2
suggests that fixed model is more ,
appropriate in this case at 10 percent where d = Nc(Nc − 1)/2, under the null level. To confirm the result of Hausman
cross-sectional test, we test whether the variance among
hypothesis
of
The other test, panels is zero, in order to perform the
independence.
heteroskedasticity, is run by command test, Breusch and Pagan Lagrangian
xttest3 following the null hypothesis multiplier test has been applied. 2 specifying that
i for i=1..Nc where Nc Due to fixed effect model is more
is the number of cross sectional data. appropriate for this case, further tests
Let T i e it be the estimator of
for fixed model result must be applied.
t Baum (2001) argued that fixed effect 1 the ith cross-sectional unit’s error
model estimated in stata command variance, based upon the T residuals e requires OLS point estimator and its
i it
available for that unit. Then define: interval perform under classical
assumption which probably the error 2
Ti
i T i T 1 ( e it i )
disturbance is independently and
identically distributed and in the case
panel data these assumptions could be as the estimated variance of . The rejected in some ways. Two tests for
modified Wald test statistic, defined as fixed effect, Breusch and Pagan for
^ Nc 2
( 2 ) independence and modified Wald test
for groupwise heteroskedasticity, are i 1 V i suggested to check the assumptions.
[Nc] under the 2 Breusch-Pagan
will be distributed
null hypothesis. Calculation shows us contemporaneous
test
for
the results of Breusch-Pagan tests that explored by Zellner’s seemingly
correlation
are
result in high probability chi square, unrelated regression (SUR) estimator.
0.43 and 0.63 for Breusch-Pagan Command ‘’xttest2’’ in Stata tests the
Lagrange Multiplier and modified Wald hypothesis that the residual correlation
test meaning that the two tests strongly matrix, calculated over data common to
that there no all cross-sectional units, is an identity
concluded
heteroskedasticity and no serial matrix of order Nc, where Nc is the