WAGE LEADERSHIP IN
TURKISH MANUFACTURING INDUSTRY
Erkan ERDIL
Middle East Technical University
Department of Economics
06531 Ankara, Turkey
E.mail: erdil@metu.edu.tr
1. Introduction
The
main motivation of this study is to clarify the institutional mechanisms in
wage determination processes for Turkish manufacturing industry. In most of the
studies on labor markets, the institutional mechanisms are either totally
ignored or frankly treated without any legitimate theoretical background. As a
starting point, we are going to deal with the classical dichotomy between
market and institutional forces in the determination of wages. However, it is
not a real dichotomy indeed. The latter approach to the labor market generally
uses the main premises of the former and introduces some extensions such as
rigidities or market imperfections. The orthodox model assumes perfect
competition and unique equilibria, on the other hand, the institutionalists
point to pervasive market power and to indeterminancy even under competition.
For instance, Common (1924) argued that in the labor market, the employer
possessed superior information, resources, and bargaining power as compared to
individual workers; Webbs (1920) claimed that the labor contract was
indeterminate regardless of the degree of competition, leaving its terms to be
specified by custom, bargaining, or law (Jacoby,1990:164-65). In the orthodox
view, an individual’s wants are taken as given and exogenous to the realm of
economic need satisfaction dominated by efficiency forces. The
institutionalists hold instead that market exchange is mediated by social institutions
that determine and are at the same time determined by individual wants and
behaviors. The orthodox view takes the utilitarianist assumption that homo
economicus is guided by rational self-interest, whereas the institutionalists
derive from pragmatism and other resources their belief that economic theory
has to be based on behavioral and psychological (or self-interest) factors
rather than on assumptions about economic behavior per se. Finally, in the
orthodox view, economic theory is synchronic: an abstraction from reality that
isolated its transhistorical and universal aspects. On the other hand, the
institutionalists emphasize that the abstractions of economic theory are
neither timeless nor placeless but instead are an ideal type.
In
this study, two models of wage inflation will be offered: a typical bargaining
model and the wage leadership model as a representative version of the
spillover models of institutional tradition. In the literature, the spillover
forces have been described under numerous names as wage leadership, relative
deprivation, key bargains, pattern wage bargains, and others. All of these
forces concentrate on the role of wage relativities in the wage determination
process. The wage leadership hypothesis postulates that the level of money
wages in any one sector of economy is determined by a comparison to a given set
of wages in the wage leading sector. The theme behind the wage leadership
hypothesis is that a certain key group of industries act as pattern setters in
the wage determination and there is a causal ordering in the wage setting
patterns in that wage increases in the key sectors lead to wage increases in
the other sectors of the economy. The role of wage leader could be played by
different sectors in course of time or by the same sector continuously over
time depending upon the subjective conditions of a country under consideration.
In the first part, first,
the general properties of the bargaining models will be analyzed. Second, a
typical bargaining model of wage inflation will be presented. Finally, the wage
leadership formulation as a representative version of spillover models of wage
inflation is derived. In order to find the factors affecting the wage imitation
behavior, sectoral unionization rate and sectoral concentration ratio will be
offered as possible sources of wage imitation. Tests for the effects of these
variables will complete the analysis. Therefore, this part provides the tools
of analysis for the next chapter concerning the empirical application. In the
next part, empirical application part of the study, both the bargaining and the
wage leadership models will be estimated by Seemingly Unrelated Regression
(SUR) method of estimation and the results of these models will be compared.
The last part will discuss the policy relevance of the wage leadership in the
light of the empirical findings. It may be the case that an economy with strong
wage leading behavior has more likely to control the wage growth if it has an
access to control the wages in the leading sectors. Therefore, macro wage
policy will reduce a sector-specific policy. Finally, the analysis will end
with the general discussion of results and concluding remarks.
2. Models Of Wage Inflation
In
this part of the study, three versions of wage inflation models will be
formulated. First, a bargaining model of wage inflation proposed by Nickell and
Wadhwani (1990), Layard et al. (1991), Nickell et al. (1994), and Lever and
Marquering (1995) is presented. Second, a market model which also includes some
bargaining elements is discussed. Finally, by employing the basic premises of
these models, a wage leadership formulation as a representative version of
spillover models of wage inflation is obtained. The models discussed here
should not be considered as antagonist to each other rather they might be
treated as complements to one another. The aim of all these discussions is to
formulate a model that can be estimated in the next chapter.
2.1. A Typical Bargaining Model of Wage Inflation
The
bargaining model used in this study has a close resemblance the models proposed
by Addison and Burton (1979), Plowman et al. (1986), and Bemmels and Zaidi
(1990).
In order to derive the
model, assuming that all firms in an industry have the same Cobb-Douglas
production function[1], the rate of growth of
labor demand in an industry (
) is written as[2]
(2.1) = 
where 
is the value added in sector i 
is the level of real wages in the ith sector.
Moreover,
labor supply to any sector is a multiplicative function of its real wage and
those in alternative occupations. Then, the sectoral labor supply function can
be written as
(2.2) 
where γ0 is a parameter that summarizes the effect of all
other aspects of net advantages on the labor supply decision and it is assumed
to be held constant and γi>1.
Then let's define Wa (a = 1,….,k) as the alternative wage set that
is the wages in alternative occupations. Taking the logarithmic time derivative
of equation (2.2), an equation
determining the rate of growth of labor supply to ith sector is
found as
(2.3) 
where γa < 0.
By
equating (2.1) and (2.3), the market equilibrium growth
rates of real wages in the ith sector is found as
(2.4) 
where β0 = 1/(1 + γi) > 0 and βa = - γa
/(1 + γi) > 0.
We further assume that
market do not clear instantaneously. In other words, during a given period only
a fraction of θ of the
difference between the current equilibrium rate of growth of the ith
real wage 
and the past period's actual growth rate 
is made up
(2.5) 
where 0 < θ < 1.
By substituting (2.4) into (2.5), it is
found that
(2.6) 
where ε = 0, π = θ/(1 + γi), Φa
= -θγa /(1 + γi), and ρ = (1 - θ).[3]
Then, we further assume
that all individuals are without money illusion that is expected rate of inflation
equals to the actual rate of inflation.
Thus, it may be possible to add the current expected rate of growth of
the price level to both sides of the equation (2.6).[4]
(2.7) 
What is implicitly
assumed in equation (2.7) is that
both employers and employees have the same sort of price expectations. Finally,
we add the inverse of the probability of being unemployed to the equation (2.7). Thus, we have[5]
(2.8) 
In
the next section, we will add the sectoral wage imitation behavior to the model
given in (2.8). Therefore, we will
add another significant component of wage determination in the bargaining
process. The resulting model, then, can be named as the wage leadership model.
2.2. The Wage Leadership Model
We
will first give a brief outline of the wage leadership model before introducing
how the presence of such behavior is examined. Following Addison and Burton
(1977), we postulate that nominal wage changes are mainly transmitted from one
sector of the labor market to another not merely by a market mechanism but by a
spillover mechanism. This mechanism has the following common form:
(2.9) 
, 
i =1,…,n (i ≠ k, s = 0) and lir,t-s ≥ 0.
In (2.9), 
is the proportional rate of money wage increase in the ith sector; 
is the proportional rate of change of money wage increase in
the reference sector(s); lir,t-s
stands for the a coefficient stating the magnitude of the spillover effect of
the rate of rth sector
wage change in the period t-s on the
current rate of ith sector
wage change; s and t are time subscripts; and finally h is the time horizon of
the spillover system.
The
relation given in (2.9) can also be
expressed in matrix notation; the array of spillover coefficients include what
Tobin (1972) has termed the wage pattern matrix of the spillover system. The
wage leadership hypothesis assumes that reference wage sets are neither large
nor variable across the labor market; all wage comparisons are supposed to be
made with respect to one singular leading sector or selected key group of
mutual leading sectors so that only Lth
column or L columns (L = 1,…, r) of the wage pattern matrix
contain nonzero elements (Burton and Addison, 1977:336). In such a situation, (2.9) can be rewritten as
(2.10) 
, i ≠ L and liL,t-s > 0
In other words,
(2.11) 
, i ≠ K and liK,t-s > 0
where
(2.12) 
, L = 1,…, r.
where f denotes the operation like
mean, median whereby the index of the rate of wage inflation in the key group 
is arrived at by
participants of the nonkey sectors (ibid, 337).
It
is possible to formulate and test the wage leadership hypothesis in several
steps:
· determining the leading sector(s),
· formulating and estimating a system
of equations,
· testing wage spillovers,
· identifying the determinants of wage
leading-following behavior.
The
first step to test the wage leadership hypothesis is to determine the leading
sector(s). In the literature, most of the studies chose the leading sector on
apriori base, i.e. Eckstein and Wilson (1962), McGuire and Rapping (1966, 1968,
1970), Reuber (1970), and Driehuis (1975). Sometimes, certain criteria are used
by some other studies in choosing the leading sectors. Edgren et.al. (1973),
for instance, used the tradable export-oriented sectors and Eatwell et.al
(1974) applied the fastest productivity growth measure. Rarely, statistical
methods such as testing the covariance structure and factor analytic approach
are employed; Mehra (1976), Plowman et.al. (1986), and Bemmels and Zaidi (1990)
are the exceptions. All of the above methods to chose the leading sector(s)
have some sort of deficiencies as explained in the previous chapter. Graafland
and Verbruggen (1993) applied modified Sims’ method in which both the one-year
and two year-lagged percentage change in wages for each sector are regressed on
the percentage change in wages in the other sectors.
In this study, we will
make use of a somewhat different method to determine the wage leading sector(s)
and employ wage leads in the model in addition to lagged wages. This model
will, in fact, be a causality analysis and details of the procedure are
explained in the appendix.
In
order to choose the leading sectors, the following equation will be estimated
by OLS for each sector in the economy.[6]
(2.13)
Ni,t = b0 + b1,iNi,t-1 + b2,iNi,t-2 + b3,iLi,t + b4,iLi,t-1 + b5,iLi,t-2 + b6,iLi,t+1 + b7,iLi,t+2 + ut
where i stands for industry, t for time,
and
WLÞ the
percentage change in the hourly wage rate of the leading sectors,
WNÞ the
percentage change in the hourly wage rate of the following sectors, and
utÞ the
independently, identically distributed disturbance term with N(0,s2).
Then,
the hypothesis of b6 + b7 > 0 will be
tested by the Wald. When the estimates of b6 + b7 > 0, this
means that changes in Lth
sector wages do not cause changes in Nth
sector wages. Hence, wages in sector N
could be leading to the wages in sector L.
After the determination of the
leading sector(s), the second step is to estimate a system of equations in
order to test the wage spillovers. The model will be estimated is found by
imposing (2.12) into (2.8).
(2.14) 
LÞ the percentage change in the hourly
wage rate of the leading sectors,
The addition of L to model given in equation (2.8) is the wages in the leading
sectors. This can be done in two ways; either weighting these wages in leading
sectors by employment rates and adding them as one variable, or adding them as
separate variables. Although the second way gives more certain results, it may
cause degrees of freedom problem. Therefore, using employment-weighted wages in
the leading sector(s) is a more appropriate way to introduce this variable.
However, if wages in one or two sectors are identified as leading wages, they
may be used separately.
After estimating equation (2.14), whether the real wage increases
in the leading sector(s) produce wage increases in the nonleading sectors will
be tested. The coefficient on variable L which can be called as the leadership
coefficient should be positive if the wage leadership hypothesis holds. If the
nonleading sectors precisely follow the leading sector(s), it should be equal
to unity. If they follow weakly, the wage leadership coefficient will be
positive but less than one.
The
next step is to identify the determinants of wage leading-following behavior.
For this purpose, two variables will be used, namely sectoral unionization rate
(UR) and sectoral concentration ratio
in terms of the first four largest firms’ sales (CR4).
One basis for the wage leadership
hypothesis is the idea that wage spillovers from one sector to another will
basically take place through labor market institutions rather than market
forces. If this is the case, it is expected that wage leadership and wage
following behavior should be most evident in highly unionized sectors. This
suggests that each of the leadership coefficients should be a function of the
extent of unionization in that sector. That is,
(2.15) μi = a0,i + a1,i URit
URÞ unionization rate in industry i.
In order to test this point
of view of the wage leadership hypothesis the model in equation (2.14) is reestimated with equation (2.15) imposed. The estimating equation
for this case will be
(2.16) 
Then the hypothesis to be tested will
be
(2.17) H0 :
a1,i = 0
H1 : a1,i ¹ 0
Furthermore, the transmission of
wage changes from leading sectors to following sectors may be a result of
product market structure. The wage changes first appear in the concentrated
sectors and may be transferred from these sectors to the following sectors via
labor market institutions. This means that leadership coefficient is a function
of sectoral concentration ratios in the following way.
(2.18) μi = θ0,i +
θ1,i CR4i
CR4Þfour-firm concentration ratio in
industry i.
Again this hypothesis can
be tested by imposing equation (2.18)
into equation (2.14) and
reestimating it.
(2.19) 
Then the hypothesis to be tested will
be
(2.20) H0 :
θ1,i = 0
H1 : θ1,i
¹ 0
The wage leadership hypothesis implies
a causal ordering from the leading sectors to the following sectors. In other
words, increases in wages in the leading sectors induce wage increases in the
nonleading sectors, yet wage increases in the nonleading sectors do not cause
wage increases in the leading sectors. This unidirectional causality is an
important component of the wage leadership hypothesis. If there were
bi-directional causality, this would indicate that wages in different sectors
are simply intercorrelated with no explicit pattern of leading and following.
This may also imply that the positive results for the leadership coefficients
reflect little more than the role of these sectors as part of the alternative
wage set as defined in the neoclassical theory. The unidirectional causality
denoted by the wage leadership hypothesis can be tested with a model designed
by Granger (1969). Thus, the following equation is estimated
(2.21) L,i,t = g0,i + g1,iT + g2,iLi,t-1+ g3,iNi,t-1 + ht
where T is a linear time trend which is included in the regression to
make the series stationary. In order to test the causality it should be tested
that the g3,i coefficients are significantly different from zero.
That is,
(2.22) H0 : g3,i = 0
H1 : g3,i ¹ 0
If they are not significantly
different from zero, then it is proved that the wage increases in the
nonleading sector do not cause the wage increases in the leading sector.
In
conclusion, this part provides the tools of analysis for the next section
concerning the empirical application. First, we present the general properties
of the bargaining models in detail. Then, a bargaining model is proposed.
Finally, the wage leadership model as an extension of the bargaining models that
includes the impact of spillover forces in the context of wage determination is
formulated. In order to estimate the last model, we have to define how to
determine wage-leading sector(s). This study proposes somewhat a cumbersome but
a very reliable method for the determination of leading sector(s) since for
each sector a separate equation is estimated and finally unidirectional
causality is also tested for the proof of the first step. Moreover, our aim is
not only to verify relevancy of the wage leadership model but also to search
for the possible institutional factors behind this behavior. For this end,
sectoral unionization rate and sectoral concentration ratio are offered as
possible sources. Tests for the effects of these variables will complete the analysis.
It is natural to think about some other relevant variables to our models but
our models reflects the basic premises of looking at labor markets.
3. Wage Determination Process In Turkish Manufacturing Industry
In
this part of the study, the previously proposed models of wage inflation will
be analyzed econometrically after they are adjusted for the availability of the
data. First, the bargaining model of wage inflation is studied. Second, the
wage leading sectors in Turkish Manufacturing Industry are determined. Third,
we estimate the wage leadership model. Finally, the extensions of the wage
leadership model in the context of the determinants of wage leading and
following behavior are studied. All the models are estimated by the Seemingly Unrelated Regression (SUR)
method. This study is concentrated only on the manufacturing industry because
of data limitations on other sectors and reliability of the manufacturing data.
Nevertheless, we have still some problems with the existing data on
manufacturing industry. The data sources and possible shortcomings of the data
are discussed in the appendix.
3.1. A Bargaining Model of Wage Inflation
In
order to assess the relevancy of the bargaining model of wage inflation for 26
sectors of the Turkish Manufacturing Industry, the following equation is
estimated by SUR. The estimation results can be seen in Table 1.[7]
(3.1) Wi,t
= b0 + b1,ivai,t + b2,iwa,t
+ b3,iwi,t-1
+ b4,imwt
+ b5,iPet
+ b6,i(U)-1t-1
+ et
where i = 1,...,29, t = 1971,...,1994, et
~
N(0,s2),and
WiÞ the
rate of growth in nominal hourly wages in industry i,
vai
Þ the
rate of growth in real value added per hour in industry i,
wa
Þ
alternative wage set,
wi
Þ the rate
of growth in real hourly wages in industry i,
mw Þ the
real minimum wage paid in the economy,
PeÞ
expected prices,
U-1Þ inverse
of unemployment rate,
The estimation results are given by
Table 1.
The labor productivity variable,
growth rate of value added, is significantly related with the growth rate of
money wages in only 10 sectors. What is more interesting is the existence of a
negative relation between these two variables in 3 of the industries with
significant coefficients. Therefore, the changes in the rate of productivity
growth do not generally explain the changes in the rate of growth of money
wages and the signs of some significant estimates are against the predictions
of the neoclassical theory.
For the Turkish
manufacturing industry, we have two variables on the opportunity cost of labor,
alternative wage set and minimum wages. In 16 out of 29 sectors, the
coefficients for alternative wage set are significant with positive signs as
expected, the only exception is sector 369 (Manufacture of other non-metallic
mineral products) with a negative sign.
Moreover, minimum wage
variable is significant in 12 sectors. These two variables have a significant
effect in 24 sectors in the explanation of the growth rate of money wages.
Therefore, it is possible to claim that opportunity cost is an important
variable in the determination of money wage changes.
The
coefficients of past real wage changes strongly confirm the presuppositions of
real wage adjustment lag hypothesis. Only one coefficient is insignificant and
all the significant coefficients have positive sign.
Price
expectations are also an extremely important factor in the determination of
money wages changes. In all industries, the coefficients of this variable are
significant and positive.
The
final important result of table 6.1 is the disapproval of Philips-type relation
in Turkish manufacturing industry. Only in 5 industries, the coefficient of
this variable is significant and negative as expected.