Collective Bargaining Effects on Wage Dispersion in Western Europe \ A. Ben-Moyal

(1)

1 | P a g e

Academic year 2008/2009 Supervisor: Dr. N.A. Lillie

Collective Bargaining Effects on Wage Dispersion in

Western Europe \ A. Ben-Moyal

Department of Economics and Business, University of Groningen The Netherlands

(2)

2 | P a g e

Introduction

Coordination through collective bargaining is recognized since the 1960’s as an influential determinant of labor market outcomes and macroeconomic performance in national political economies (Blau and Kahn, 2002). Following a post world war period in which distribution of wages tended to become more compressed, most OECD countries have experienced some increase in wage inequality since the 1970’s together with a declining role of labor institutions. The magnitude of the changes, however, varies significantly across these countries. Thus, with countries entering the 1980’s at very different levels of wage inequality, the persistence of cross national diversity remains a suspect of the data that will be presented during this research. Such pattern of rising wage inequality and a rapid declining role of labor market institutions provide a very good testing ground for evaluating the role of labor market institutions in explaining the evolution of wage structure. Thus, the main question presented in this paper is whether unions support economic equity or just defend skilled insiders.

(4)

4 | P a g e

distribution. Their test shows that the move in Sweden towards decentralized bargaining structure in recent years have actually improved individual bargaining position. Their data, however, does not allow them to distinguish between those two effects and it rely on annual surveys of members of affiliated trade unions. Thus, it seems that country specific differences play a role in judging the direction of bargaining effects on wage dispersion.

Collective bargaining is a process of negotiations which takes place between employee’s union representatives and employers to the purpose of mutual acceptance of employment issues such as wage, hours, and working conditions. We assume that since the negotiations deals with scarce resources there is bound to be a conflict of interest between them. The extent to which collective bargaining covers employee’s wages as well as how many employees are represented by a union may affect the bargaining outcome. Three main dimensions of collective bargaining (CB) levels can be observed: union density, bargaining coverage, and centralization of collective bargaining. Union density covers the employees who are officially members of a union. Bargaining coverage also covers the employees which are covered by an agreement but not officially member of a union. Centralization of collective bargaining measures the level of which collective bargaining is conducted in a specific country. The first level is the firm level, where employees and firms negotiate terms of work and wage at the individual firm level. Countries such as the U.S, Canada, Japan, and UK are among this group and considered to have a relatively decentralized wage bargaining systems. At the other extreme level, the bargaining procedure is conducted at the national level, covering the entire economy or most part of it. Countries such as the Nordic countries (Sweden, Norway, Finland, Iceland, and Denmark) were historically among these countries and still considered to have a relatively centralized wage bargaining system. Most EU countries belong to the intermediate level of wage negotiation which is conducted mainly at branch or Sectoral level.

(5)

5 | P a g e

1. Does overall earnings dispersion tends to fall as the percentage of employees who are members of a union increase?

2. Does an increase in bargaining coverage tend to cause overall earnings dispersion to fall? 3. Does earnings dispersion tend to fall as unionization become highly centralized or

decentralized?

4. Do the answers to questions 1-3 differ significantly for Nordic and non-Nordic countries? The rest of the paper is organized as follows: sections 2 explore the literature on collective bargaining and discuss the nature of unions, collective bargaining, and associated costs. 3 main parameters are identified in the literature and used in this research: centralization of collective bargaining, bargaining coverage, and union density. Then we show evidence from the U.K, where major reforms were taken by the Thatcher administration in the 1980’s to reduce labor union power. We show that this era is associated with an increase in wage dispersion. This section also explores the Nordic labor institutions and their importance on wage formation. Income policy has been a central element in the labor market regimes in all the Nordic countries since the 1950’s along with rapid state intervention in bargaining and conflict resolution. State intervention among Nordic countries is different from one country to the other and it is based on different historical traditions concerning the role of the state in the industrial relations system. Section 3 shows the hypotheses derived from the literature, the data that is used in this research and a discussion of the patterns which can be observed from the data. This research use data from the OECD and the Visser classification. As the data reveal there is a general trend towards increased wage inequality while trade union density and collective bargaining coverage show negative trend over time. Section 4 focuses on the methodologies and models used throughout this research. Panel data is used to cover data over time and individual countries and for that 4 basic models are used. The common feature of all the models is that they are all fixed effects models. Section 5 show the empirical results derived from the data and methodology, and section 6 discusses the results. Section 7 concludes the paper.

Literature

(6)

6 | P a g e

regarding union power in different countries have affected the wage gap between the high and the low tier of the wage distribution.

2.1 Rise of the Unions

The rise of unions can be seen as the outcome of asymmetry in contracting between individual workers and employers regarding basic labor rights, work conditions, pay structures, wages, and different employment relations. As to wages, the evidence presented here suggests that in most countries union membership cause compression of the wage distribution. One possibility might be that labor unions increase the wages of the bottom and mid bottom of the earnings distribution by renegotiating their terms of employment and pay structure. This causes wages of the low and mid tier to rise and therefore get closer to the mean pay structures in the economy, which induce a wage equalization effect. Another possibility might be that unions control for the wages of the high tier to moderate inflation and to reduce the wage gap. Therefore, wage coordination might be a goal for labor unions.

Labor unions are involved in a variety of activities. Aidt and Tzannatos (2002) make a distinction between 3 aspects of union behavior:

1. Participatory benefits of unions

Unions are seen as communication channels for the preferences of workers to management and a way to participate in the establishment of working conditions. Unions can also increase productivity as they provide a communication channel to management attention.

2. Unions as political organizations

According to Pencavel (1995), union’s political activities can have large distributional costs as it comes from the expense of nonunion workers and consumers. For example when a union is successful in getting government regulation, an economic distortion and a dead weight loss to society is created. However, according to Aidt and Tzannatos (2008) we cannot assume that total costs of a negotiation breakdown are higher under collective bargaining system than under individual contracting system. Furthermore, given economies of scale in the dissemination of information, there is a reason to believe that collective bargaining system (as a system of dispute resolution) may be less costly than individual contracts.

3. The monopoly cost of unions

(7)

7 | P a g e

to avoid an industrial conflict. Union’s monopoly power depends on their size and strength. When a labor market is endowed with much non unionized labor supply, union’s bargaining position is weaker. When unions succeed they will impose, what Aidt and Tzannatos (2008) define as monopoly costs of unions, such as higher prices due to top wage demands and higher wage mark-ups. Therefore a bigger union endowed with unionized workers might have a positive effect on their earning distribution.

2.2 Collective Bargaining Characteristics

By reviewing the literature (Calmfors 1993, Visser 2000) 3 main characteristics can be observed for collective bargaining: centralization of collective bargaining, bargaining coverage, and union density. The following section examines those characteristics and their plausible influence on wage equality.

Centralization of Collective Bargaining

1

Centralization of collective bargaining is one of the institutional aspects of the labor market that determines the degree of bargaining coordination. Centralization of collective bargaining is defined by Calmfors (1993) as “the degrees of inter-union and inter-employer co-operation in wage setting”. There are several reasons, found in the literature, which help explain why centralization of collective bargaining is likely to produce better economic results. First, wage compression can lead to a reduction in wage dispersion under centralization of collective bargaining system since firm-specific conditions are less likely to enter wage contracts. The benefit associated with this is the encouragement of new and more efficient firms. The costs however can be associated with economic misallocation of resources and a lower output. Secondly, issues such as general training of workers and health are more likely to occur at collective bargaining at higher levels. Efficient overall training can lead to overall economic growths. Some opponents claim, tough, that efficient bargaining and cooperation between unions and firms is more efficient under decentralized bargaining. Thirdly, proponents of centralized bargaining system stress the fact that imperfect information can lead to more strikes. Centralization, therefore, increases the level of information about demand conditions, which, in turns, reduce the probability to strike.

As to the effect of centralization of collective bargaining on wage dispersion, by centralizing the bargaining process wage setters are forced to bear a larger share of the cost of their

1_{Throughout this research centralization of CB is measured on a five point scale from individual firm level CB at}

(8)

8 | P a g e

actions, as more workers become included in the bargaining coalition, and this creates incentives in favor of wage restraint which induce wage equality (Aidt and Tzannatos, 2008). According to Rueda and Pontusson (2000), centralization of collective bargaining facilitates the reduction of sectoral wage differences since more firms and sectors are included in a single wage settlement. Centralizing the collective bargaining process causes policy makers to consider the whole economy and make sure that an increase in one sector would not cause unemployment in another sector or heavy cost to the whole economy. In order to understand this we need to imagine a country where all employees are organized in unions and each firm negotiates with company’s union. Then when one union succeeds to raise its employee’s wage, every increase of their real wage will cause economic cost as they impose economic cost on others. However they will only bear a small fraction of the total economic cost. On the contrary, in a country where all employees are organized in a single union, wage setters are forced to bear a larger share of the cost of their actions, since all workers are included in the union. This creates an incentive for wage restraint.

This notion was previously tested by Calmfors and Diffill (1988) which hypothesized that both highly centralized and decentralized bargaining systems produce real-wage moderation and therefore reduction in wage variance between sectors. In the case of decentralize bargaining system, market forces will tend to restraint wage growth as labor and capital will flow to the high paid sector and therefore reduce it back. According to Calmfors and Diffill only in the intermediary levels of collective bargaining at the industry level there will be a high probability of lower wage restraint since both market forces and internalization effects could be too weak. The point is that when wage bargaining concerns a whole industry, the probability of transferring pay rises to consumers through an increase of the relative output price is great, and as a result it will hold back the rise of the sector’s real product wage2. Because employment is set by the real product wage, the employment loss from a given increase of the real consumption wage is reduced. This will tend to reduce the incentives for wage moderation on the union side. On the employer side, the output price increases reduce the profit decrease from a given increase of the real consumption wage (Calmfors and Diffill, 1993). Hence, it follows from Calmfors and Diffill that the relationship between economic performance and centralization of collective bargaining can be non monotonic which means U or hump shaped (see graph 2.2a).

(9)

9 | P a g e

Graph 2.2a: Hump shaped hypothesis

The hump shaped hypothesis should be taken with some considerations, however. First, if we assume perfect competition in the product market then indeed individual firms will not be able to raise their relative price under decentralized wage setting. However in a more realistic case monopolistic competition will dominate and a relative price increase will indeed be the outcome of wage increase in the individual firm. Second, the statement that only intermediary extent of wage setting at the industry level will produce worse macroeconomic outcomes than very low or high degree of centralization lies on the assumption that wage restraint is eliminated if domestic producers bargain together. However this statement ignores international competition, since foreign competitors are not encompassed by domestic wage increase. In graph 2.2a this scenario is presented by curve 4 where we can observe the relationship between the extent of centralization and the aggregate real wage. As we can observe, there is no hump shape curve such as in a closed economy or even open economy. Thus this hypothesis was rejected by many researchers and was abandoned.

Union Density and Bargaining Coverage

(10)

10 | P a g e

(head count) who are members of a union, as a percentage of all workers, unionized and none unionized. Union density might be related to wage dispersion since the bigger the union is, the more power it has to affect the wages of its members. Bargaining coverage, on the other hand, measures the real rather than the potential extent to which employees are subjected to union negotiated terms. Bargaining coverage is operationally defined as the number of full time employed workers, unionized or not, which have their pay and employment conditions determined by a collective agreement, as a percentage of all full time employed workers, unionized and none unionized. Bargaining coverage is less straightforward than union density, since some employees are automatically covered by an agreement even when they are officially not member of the union. On the other hand some employees may not be covered even if bargaining coverage takes place. First, union and employees may be too weak to include all employees which belong to their domain of action. Secondly, bargaining parties sometimes exclude certain groups such as part time employees. This group usually contains large proportion of women and young people. Unions may also exclude employees which earns above a certain threshold. Thirdly, certain categories of employment may also be omitted from coverage such as employees in the public sector (i.e. police and army forces). Chart 2.2b and 2.2c show bargaining coverage and union density in Europe3. The charts reveal that the average level of bargaining coverage in Europe is almost as twice as high as the average density level. Chart 2.2b shows the enormous difference in bargaining coverage levels across Europe. Chart 2.2c reveal that the highest levels of union density can be found in the Nordic countries (about 80% in Sweden) and lowest in France, Spain, Poland and Hungary. Thus there is a great difference between union density and bargaining coverage across Europe. Therefore it is in the interest of this paper to find which one of these parameters is more significant for wage dispersion.

3

(11)

11 | P a g e

Chart 2.2b: Bargaining coverage in Europe

Chart 2.2c: Union density in Europe

Source: OECD (2004).

2.3 Evidence from the U.K

(12)

12 | P a g e

between wage council and non wages council workers fell from 0.38 log points in 1979 to -0.43 log points in 1992.

Graph 2.3 shows the difference between wage inequality over the last 30 years between Sweden and the U.K. As it clearly shows, while Sweden maintained a relatively stable wage inequality over the years the U.K has experienced a relatively steady increase in wage inequality from 1970-2009. According to the graph, the major increase in wage inequality occurred since the beginning of the 1980’s until the beginning of the 1990’s when, politically speaking, Margaret Thatcher was prime minister. According to Gregory (2000), under the Thatcher era (1979-1990) new laws came into effect in Britain which imposed a legal framework on union activities, progressively restricting union powers. Union privileges at the workplace, for example, were decreasing. Closed shops, where employees had to be a union member, were abolished by law. Even the ‘check off system’ whereby the employer deducts union fees from the employee was limited and had to get an authorization from the employees. Those measures increased union costs and as a result from 1979 onwards union member fell from 11.7 million in 1979 to 7.2 million in 1996. This represents a fall in union density from 50 % to 23%. A direct result of the declining role of unions was the declining role of collective bargaining. By 1984 66% of British employees worked in establishments where unions were recognized and by 1994 it fell to 34%.

Table 2.3: Changes in the structure of male wages 1979-93, by wages Council coverage

(13)

13 | P a g e Graph 2.3

Source: my calculations using data from the OECD 2006.

2.4 Nordic Institutions

This section spills light on the historical background and organizational structure in the Nordic countries, which should be seen as explaining variables in the analysis of collective bargaining systems and income policies. As graph 2.3 and chart 2.2c show, Nordic countries tend to have relatively lower income dispersion and a high union density. According to Elvander (2002), in the last 100 years the state has played a significant role in collective bargaining and conflict resolution in Nordic countries. Denmark, for instance, has started its industrialization process as early as 1870 and developed on a craft based small industry basis. Hence, modern rules and institutions have developed in Nordic countries earlier and high degree of social partners in the form of binding agreements that become the basis for legislation. Elvander (1974) claim that there is a clear relation between income policies and bargaining systems since income policy has the character of public addition to the bargaining system. In his paper “collective bargaining and income policy in the Nordic countries” Elvander describes four types of income policies4, each representing different levels of intervention commonly used by the Nordic countries each to a different extent.

4

Income policy is defined here as the degree of state participation in the process of wage formation. 1,5 2 2,5 3 3,5 4 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 W a g e i n e q u a li ty

Wage inequality among Nordic countries and the U.K

(14)

14 | P a g e

1. Political rhetoric this is mostly vague recommendations to the bargaining parties with the aim of encouraging the employee organizations to exercise wage restraint. The given reason is that such an act would prevent an unwanted inflation for instance. 2. Neutral expert reports this group does not have a coercive power. There sole aim is to

give information which might affect wage negotiations. This board of experts is appointed by the government and their task is to analyze the economic situation and give recommendations.

3. Wage control the government appoints an agency for supervision of wage developments. The control agency tries to influence the market forces and is able to apply sanctions.

4. Stabilization agreements this is the most ambitious form of income policy as it tries to stabilize the price level and usually includes all groups of employees and producers with an active participation of the government.

(15)

15 | P a g e

of a strong repression of Communism during the inter-war years. This might explain the extensive Finish government intervention in income policies.

Elvander (1974) concludes that a social democratic government is most likely to satisfy the demands for a strong government in relation with the initiation of an incomes policy. Social democrats are in an advantage when it comes to getting the support of the trade union movement for a policy of wage restraint. They do it mostly through secret meeting and contacts while non socialists tends to act in a more open way which might be less effective in getting trade union support for income policy measures. Social democracy presents the mobilization of industrial workers such that workers organizations exercise significant power in the labor market and in the government. Unions negotiate the terms of work for most of the labor force and exert a strong influence in politics through closed connections to a social democratic party. Therefore, according to Elvander (2002) (1974) Nordic countries are in the highest probability of having the lowest levels of wage inequality since they score high on centralization level and a high union density. To summarize, the evidence from the literature, the U.K and the Nordic countries reveal that collective bargaining might play a significant role in shaping the wage in the economy.

Hypotheses, Data, and Stylized Facts

The research questions mentioned above together with the factors that shape collective bargaining mentioned in the literature review lead to the following hypotheses that will be researched through the remaining of this article:

1. An increase in union density causes a decrease in wage dispersion.

2. An increase in bargaining coverage causes a decrease in wage dispersion.

3. A higher level of centralization of collective bargaining causes a decrease in wage dispersion.

4. There is no difference between Nordic countries and Western European countries as to unionization effects5 on wage dispersion.

5

(16)

16 | P a g e

3.1 Data

As derived from the literature, there are 3 important indicators to collective bargaining strength in a country: collective bargaining coverage, centralization of collective bargaining, and union density. Therefore, they are included in the regression that will investigate their effect on the spread of wages. The database that will be used in this research is compiled from data from the OECD, Euro Stat and the Visser classifications. The sample covers annual data from 10 European countries; Belgium, Germany, Denmark, Finland, France, Italy, the Netherlands, Norway, Sweden, and the U.K. The selected time period is from 1980-2007. Since not all data is widely available for some years, some countries such as Austria, Cyprus, Spain, Ireland, and Luxembourg were omitted from the sample countries. The OECD statistics portal offer a wide range data regarding wages in OECD. The dependent variable in this analysis is a measure of the distribution of gross income from employment. The particular measure that is being used in this research is the ratio of earning at the 90th percentile to earnings at the 10th percentile. The 90-10 ratio is a measure of the distance between these two points. It is important to keep in mind that this measure ignores self employment, income from capital and is restricted to full time employees (GAFO6 and GWGO7classification). Therefore, the 90-10 ratio fits this research well since self employed and income from capital is not covered by unions and collective bargaining. This matches the measuring of union density and bargaining coverage which also include only full time employed workers.

Data on union density, bargaining coverage, centralization of collective bargaining, and the level of bargaining is obtained from the Visser classifications. Visser uses various national and comparative sources such as the OECD (2006), EIRO (2004), and survey data for Germany, the U.K, and the Netherlands. The indicator on centralization of collective bargaining measures the level at which bargaining between employers and unions take place. The indicator can take any of five values: 1. Company and plant level predominant. 2. Combination of industry and company level. 3. Industry level predominant. 4. Predominantly industrial bargaining. 5. Central level agreements (government level). Union density measures the fraction of union members in the in the employed labor force. The data on union density and bargaining coverage is taken from Visser (2006) and based on various national and comparative sources.

6

GAFO- yearly gross income excluding casual payments for full year employees working full time.

7

(17)

17 | P a g e

3.2 Stylized facts

By looking at the data that was collected it is possible to observe the trends in wage dispersion, union density, centralization of collective bargaining, and bargaining coverage within and among sample European countries. Table 3.2a summarizes trends in earnings dispersion between 1980 and 2007. It appears that among most countries there is a general trend towards increased wage inequality. The U.K, which is known to have a relatively decentralized collective bargaining system, had an earning dispersion of 3.195 in the 1980’s and by 2007 scored 3.6 in inequality. Even Sweden which is known to have a relatively centralized bargaining system grew in inequality from 2.05 in 1980’s to 2.30 in 2007.

(18)

18 | P a g e

Table 3.2a Trends in earnings dispersion 1980-2006

Country 1980’s 1990-1994 1995-1999 2000-2007 Total change

Belgium 2.41 2.28 _… _… ... Germany 2.86 2.79 2.87 3.11 0.25 Denmark 2.17 2.16 _… _… _… Finland 2.49 2.39 2.36 2.41 -0.08 Italy 2.29 2.35 2.40 .. 0.11 France 3.18 3.21 3.07 2.95 -0.23 Netherlands 2.51 2.60 2.85 2.92 0.41 Norway .. .. 1.96 2.11 _… Sweden 2.05 2.11 2.23 2.30 0.25 U.K 3.19 3.39 3.45 3.61 0.42

Source: my calculations, using data from the OECD 200 ... Data not available

Table 3.2b: Trade union density and collective bargaining coverage in sample countries

Country Trade Union Density Collective Bargaining coverage

1980-1989 1990-1999 2000-2007 1980-1989 1990-1999 2000-2007 Belgium 52.28 54.51 52.07 96.3 96 96 Germany 34.23 29.85 22.425 75.9 68.2 63 Denmark 77.88 76.21 71.8 72.4 70.8 81 Finland 69.9 78.18 73.03 77.7 82.51 88.75 France 14.39 9.27 8.14 88.9 94.61 95 Italy 43.72 37.71 33.79 84.7 82 80 Netherlands 28.79 24.6 21.85 73 85.2 83.62 Norway 57.51 56.9 54.71 70 71.8 72 Sweden 81.23 84.14 76.9 74 89.52 91.71 U.K 46.34 34.57 29.48 64 40.93 35.26

Source: my calculations, using data from the OECD 2006.

Table 3.2c: Centralization of collective bargaining rank in sample countries 1980-2007

(19)

19 | P a g e

Norway 3.5 4.5 4.5 4.5

Sweden 4.5 3 3 3

U.K 1 1 1 1

Centralization:

1. company and plant level predominant

2. combination of industry and company level

3. industry level predominant

4. predominantly industrial bargaining

5. central level agreements.

Source: my calculations, using data from the OECD 2006.

Methodology

(20)

20 | P a g e

the countries for some specific year then this can have an enormous effect on the estimated parameters. However using different methodology such as cross sectional data is also not optimal since we will not make an efficient use of the data because cross sectional model is meant for one cross section while fixed effects model takes into account the same cross section evolving over time.

Most previous research into the effects of unionization on wage inequality is concerned with bi variant correlation analysis. However there are several studies that include a fixed effects regression. Rueda and Pontusson (2000) study the effects of multiple independent variables on wage inequality. They try to determine whether effects on wage inequality are the same for different types of economies (social market, liberal market, and mixed market). They use a fixed effects model with a lagged time variable. To deal with the problem of biased coefficients due to the use of fixed effects with combination with lagged dependent variables they compute instrumental variable estimator suggested by Visser (2000). In this research, we do not make a use of lagged variables in order to avoid such a bias, and since we do not have any additional variables that can serve as instrumental variables. As an effect, we do assume changes in the independent variables have an immediate effect on the spread of wages.

(21)

21 | P a g e

In order to empirically research for the relation between spread of wages and collective bargaining the following model will be investigated:

(1)

Where yjt is the dependent variable which measures the spread of wages, xkjt is the independent variable which measures passage of time, union density, collective bargaining coverage, and levels of centralization of collective bargaining, j refers to the cross sectional units (countries), t to the time units, k to the number of independent variables, αj refers to separate intercepts for each country, βk refers to the slope of the explanatory variables, and εjt is a random error term normally distributed around a mean of 0 with a variance of σ2. For convenience table 4.1 lists all the symbols used throughout this research.

The first model that will be used is a fixed effects model with a fixed year effect as it allows intercept to vary across time and individual countries and each year has a separate effect on wage dispersion. The regression to be estimated in this model is:

(2)

Centralization of collective bargaining is divided into 5 levels, where level 5 pertains to highly centralized bargaining structure and level 1 to highly decentralized bargaining structure. In this model, level 5 is taken to be the base case, while 4 other dummy variables indicate when the level of centralization is different. Hence, the coefficients for the other levels measure the difference in wage dispersion between their level and the highest level of centralization of collective bargaining.

A time variable dummy was included in order to allow for changes over time which affects the entire panel. The inclusion of time variable dummies was done in order to make sure that changes in average in wage dispersion over the years would not affect results. Hence, any structural changes over time are controlled for so the average within a year across countries stays zero. However, modeling year effects using time dummies has several disadvantages. First, it has a lot of coefficients to estimate, which requires more data. Second, it does not allow for prediction since future year dummy coefficients are unknown.

k kjt j jt k

x

a

jt

y

=

∑

β

+

+ ε

1 2

cov

3 1 4 2 5 3 6 4 jt j t jt jt

(22)

22 | P a g e

Table 4.1: List of symbols used throughout the research

j cross sectional country

t Time units (t = 1980, … , 2007)

Sprjt Spread of wages of country j in time t

αj Intercept of the spread of wages of country j

δt Time effect of time period t

β Coefficients

εjt Error term

adjcovjt Collective bargaining coverage

leveljt Variable for the level of country j

(1=highly decentralized company level, 5= highly centralized level)

udjt Union density

level1jt, level2jt,

level3jt, level4jt

Dummy variables for the level of centralization 1, 2, 3 and 4 resp.

yeart Year at time t, ranging from 1980 up until 2007

norjt Dummy variable indicating whether country j is a Nordic country

wstjt Dummy variable indicating whether country j is a western European sample country

The second model that will be used is a fixed effects model with a linear year effect. This model assumes linear relationship between spread of wages and time. Hence across time we expect wage dispersion to rise in a linear way. The regression to be estimated in this model is: (3)

The third model is a model which uses a single variable to represent the level of centralization of collective bargaining instead of four separate dummies. This model is useful in order to check if there are any advantages in using separate dummies over a single variable. The regression to be estimated in this model is:

(4)

spr

jt

= α + β

j 1

ud

jt

+ β

2

adj

cov

jt

+ β

3

centralization

jt

+ β

7

year

+

ε

jt

The fourth model is a model which controls for country group specific effects. This way it possible to test whether the effects of union density, bargaining coverage and centralization of collective bargaining has a different effects on the spread of wages for different countries. In the original model (second model), we have small amount of parameters to estimate (17) but

1 2

cov

3 1 4 2 5 3 6 4 7

jt j jt jt

(23)

23 | P a g e

we cannot really test if there are country specific differences. If we use a full country specific model we can test for country specific differences. However, there is large number of

parameters (41) which requires a large database. Hence, we decided to group the countries into 2 basic groups. According to the literature, the Nordic countries score low on spread of wages and high on union density and centralization of collective bargaining. Therefore we have decided to group them together. The rest of the countries will be grouped as Western Europe group. Also according to the stylized facts it is possible to observe that this division is logic if we look at the spread of wages and union density within and among those countries. The regression to be estimated in this model is:

(5) 1 2 3 4 5 6 7 8 9 10 11 12 cov 1 2 3 4 cov 1 2 3 jt j t jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt jt

spr year ud nor adj nor level nor level nor level nor level nor ud wst adj wst level wst level wst level wst

= α + δ + β ⋅ + β ⋅ + β ⋅ +

β ⋅ + β ⋅ + β ⋅ + β ⋅ +

β level4_jt⋅wst_jt+

ε

_jt

Where udnor has the same value as ud for Nordic countries and 0 for western European countries. While udwst has the same values as ud for western European countries and 0 for Nordic countries. Other variables with nor or wst suffix are constructed similarly. In this model, we combine the advantages of country specific parameters while keeping the number of estimations limited (23). Note that there is no difference in intercept of Nordic and Western European countries since dummy for Nordic countries is indistinguishable from the combined country specific dummies of Nordic countries. Level 5 of the centralization parameter (national level) is the base case for both Western and Nordic countries. There is no variable distinguishing between the highest levels for Nordic versus Western European countries to avoid linear dependency.

Empirical results

(24)

24 | P a g e

clearer idea as to the fitness of the model to the data. Additionally, to check for the presence of multicollinearity, pair wise correlations between the independent variable representing bargaining coverage, union density, centralization of collective bargaining and year are given by table 5.0. The table shows that correlations between pairs of variables are reasonably low8. All tables, figures, and graphs are to be found in the appendix section.

Table 5.0: Multicollinearity test: bivariate correlations

5.1 Model 1: Fixed effects model with a fixed year effect

Table 5.1 show results of the regression analysis. For a better overview of the results for the year dummies, in table 5.1 coefficients for the year dummies are averaged over periods of time. These ‘period coefficients’ are the average coefficient value of the year dummies over the time period they represent. Averages have been taken over 5 years for convenience. Note that the year dummy representing the first year in our dataset (1980) was dropped because we use it as a base case. One of the requirements of linear regression is that independent variables are not linear dependent. If we include all year dummies then they are linearly dependent on all country dummies.

Results show that union density has a significant9 (p value < 0.01) negative effect on wage dispersion as expected by hypothesis 1. Bargaining coverage does not show significance (p value > 0.05) which means that expected negative effect of bargaining coverage on the spread of wages of hypothesis 2 was not found. Centralization of collective bargaining at level 1 and 2 show a positive significance (p value < 0.05). The more decentralized unionization the more wages are spread out as expected by hypothesis 3. As to time effect, it is possible to observe from the averaged coefficients that there is a trend towards higher variability of wages as the coefficients increase over time. Graph 5.0 shows the same trend. As to diagnostics check, the histogram of the residuals (see graph 5.1) shows a fairly normal distribution. The JB testfor normality confirms there is no reason to believe the distribution is anything other than normal (p > 0.05). Although the JB test show that the distribution of the errors are not significantly different from normal distribution, the errors still seem to have a relation with time (see graph

8_{Absolute value of higher than 0.8 would indicate risk of collinearity (Adkins & Carter Hill, 2008)} 9

The term significant pertains to 1% level throughout the empirical results

year --0--00.0...1111113113233222 00.00...0000111818688666 ---0-0.00...0001011212222222 1111..0..00000000000000 0

cent 000.0...22222222222626 66 000.0.3..336366363033000 11.11...0000000000000000

adjcov 000.0.0..0005555111616 66 111.1...0000000000000000

ud 111.1...0000000000000000

(25)

25 | P a g e

5.2) as well as country (see graph 5.3). From this we can observe that there might be a difference between countries as to how wage dispersion is related to union density, centralization of collective bargaining, and bargaining coverage. Another possible explanation is auto-correlation. In fact, the residuals do show a correlation of 0.5568 with their previous value. An LM test shows there is no heteroskedasticity (p > 0.5) based on the union density and bargaining coverage, but residuals do seem to be different for separate countries (p < 0.01).

Table 5.1 Descriptive regression relating characteristics of the collective bargaining system on spread of wages using fixed effects model with a fixed year effect, 1980-2006

Spread of wages Average year 1980-1985 -0.00179 Average year 1985-1990 0.02413 Average year 1990-1995 0.04132 Average year 1995-2000 0.11906 Average year 2000-2007 0.16919

Trade union density -0.01530**

(0.00411)

Collective bargaining coverage -0.00184

(0.00223) Centralization level 1 0.366** (0.1097) Centralization level 2 0.111 (0.0619) Centralization level 3 0.07254 (0.04518) Centralization level 4 0.0695 (0.0447)

Year dummies Yes

Country dummies Yes

Number of observations 195

R-squared 0.7076

**, * denote statistical significance at the 1%, 5% levels, respectively

(26)

26 | P a g e

5.2 Model 2: Fixed effect model with a linear year effect

(27)

27 | P a g e

Table 5.2 Descriptive regression relating characteristics of the collective bargaining system on spread of wages using fixed effects model with a linear year effect, 1980-2006

Spread of wages

Year 0.00787**

(0.00172)

Trade union density -0.0131**

(0.00343)

Collective bargaining coverage -0.00348

(0.00184) Centralization level 1 0.364** (0.0897) Centralization level 2 0.137** (0.0517) Centralization level 3 0.0605 (0.0377) Centralization level 4 0.0499 (0.0385) Year dummies No

Country dummies Yes

Number of observations 195

R-squared 0.59

**, * denote statistical significance at the 1%, 5% levels, respectively Source: my calculations,

with data from the OECD 2006, stata

5.3 Model 3: Fixed effects model with a linear year and level effects

(28)

28 | P a g e

Table 5.3 Descriptive regression relating characteristics of the collective bargaining system on spread of wages using fixed effects model with a linear year and level effect, 1980-2006

Spread of wages

Year 0.00881**

(0.00171)

(0.00343)

Collective bargaining coverage -0.00541**

(0.00171)

Centralization level -0.0436**

(0.0153)

Year dummies No

R-squared 0.57

**, * denote statistical significance at the 1%, 5% levels, respectively Source: my calculations, stata

5.4 Model 4: Fixed effects model with a country grouping

(29)

29 | P a g e

A test on the equality of the coefficients for Nordic and Western European countries shows they are significantly different (F(6,174) = 7.26, p < 0.01). This is contrary to hypothesis 4 which expected no significant differences between Nordic and Western European countries. The histogram of the residuals shows a normal distribution (see graph 6.0). As to the JB test for normality, based on the kurtosis and skewness we have no reason to believe that the residuals are not normally distributed (p> 0.5). The errors in this model seem normal. With comparison with the first two models, the outliers of the residuals in this model over time are less strong. Thus, there seems to be no relation between residuals and time (see graph 6.1), although there still is a correlation of 0.4971 of the residuals with their value in the previous time period. Residuals might also still be different across countries (see graph 6.2). An LM test shows that the country significantly explains the squared residuals (p < 0.01).

Table 5.4 Descriptive regression relating characteristics of the collective bargaining system on spread of wages using fixed effects model using country grouping, 1980-2006

Nordic Countries Spread of wages Western European Spread of wages

Year 0.00383 Year 0.00383

(0.00405)

Trade union density -0.0141**

(0.00499) Collective bargaining cov. 0.00571

(0.00316)

Collective bargaining cov.

-0.00838** (0.00226)

Centralization level 1 --- Centralization level 1 0.148

(0.0907) Centralization level 2 0.193* (0.08) Centralization level 2 0.00305 (0.0515) Centralization level 3 0.0626 (0.038) Centralization level 3 -0.0418 (0.0355) Centralization level 4 0.0237 (0.0386) Centralization level 4 ---Year dummies No

R squared 0.64

(30)

30 | P a g e

Discussion

As the results show there is not such a difference between linear year effect and a fixed year effect. Therefore linear year effect has an advantage since it uses fewer parameters and allows for easier predictions. Thus, it is the preferred model. According to the linear model union density has a significant negative effect on the wage dispersion. This means that, as hypothesized, an increase in the proportion of people that are member of a union have lowering effect on wage dispersion. In the third model union density and centralization of collective bargaining has a significant negative effect on wage dispersion. Bargaining coverage also show significance negative effect on wage dispersion. However in this model centralization is combined into one variable. That is, it forces a linear effect of centralization. So the difference between level 1 and 2 is the same as level 4 and 5 in terms of effect on wage dispersion. In the final model bargaining coverage has a significant negative effect on wage dispersion for western European countries. This is as expected by hypothesis 2. However, no effect was found for the Nordic countries. For Nordic countries, union density has a negative effect on wage dispersion, but bargaining coverage has no effect. A plausible explanation might be that union density indicates the bargaining power of the unions. That is a higher union density would mean that unions can get stricter agreements regarding wages. On the other hand low union density combined with high bargaining coverage means that the agreements affect more people but the agreements do not have a lot of influence on the wage restriction. Since bargaining coverage produce significant results only in the third model and partly in the fourth model it is difficult to draw conclusions. However since the preferred model is the linear model, it is possible to stress that bargaining coverage is not significant to the wage dispersion.

(31)

31 | P a g e

As was previously mentioned Nordic countries have higher levels of centralization of collective bargaining and union density. Therefore results confirm that indeed there are differences between Nordic countries and Western Europe regarding the wage dispersion contrary to what predicted in hypothesis 4. Unlike for Western European countries, bargaining coverage does not seem to affect wage dispersion for Nordic countries. This has to do with the plausible explanation that every employee is automatically covered by an agreement in Nordic countries.

As to possible limitations, the fixed effect model we used assumes that countries are the same except for their default wage dispersion. As we previously discussed Nordic countries turned out to be different from Western European countries. So this raises the issue whether this assumption is justified. For example even within Western Europe there is a lot of differences in centralization of collective bargaining. However lack of data prevents us for estimating a model for each country separately.

Conclusions

(32)

32 | P a g e

Since Nordic countries have a long history of social democratic regimes, high union density and high level of centralization of collective bargaining, they have proven to have the lowest wage inequality among sample countries. In the last 100 years Nordic countries government have played a significant role in collective bargaining and conflict resolution with the use of political rhetoric, neutral expert reports, wage control, and stabilization agreements. The U.K, on the contrary, has undergone reforms in the beginning of the 1980’s under the Thatcher administration, which limited the power of labor institutions and raised their costs. As a result the U.K has the highest wage inequality among sample countries. A further research is needed to further explore the relation of wage dispersion with collective bargaining while taking into account factors such as FDI, unemployment rates and local factors which may differ for different countries.

Appendix:

5.1 Model 1: Fixed effects model with a fixed year effect

Graph 5.0: Line graph of the year dummy coefficients over time.

Graph 5.1: Histogram of the residuals of the fixed year effects model.

(33)

33 | P a g e

Source: my calculations, STATA, with data from the OECD 2006.

Graph 5.2: Scatter plot of the residuals of the fixed year effect model over time

Source: my calculations, STATA, with data from the OECD 2006. Pvalue =.11470562 0 1 2 3 4 D e n s it y -.4 -.2 -. 2 -. 1 0 .1 .2 e h a t 1980 1990 year

Graph 5.2: Scatter plot of the residuals of the fixed year effect model over time

0 .2 .4

resids2

2000 2010

(34)

34 | P a g e

Graph 5.3: Scatter plot of the residuals of the fixed year effect model per country

Graph 5.3b: Scatter plot of the residuals against their value in the previous time period

5.2 Model 2: Fixed effect model with a linear year effect

Graph 5.4: Histogram of the residuals of the linear year effects model.

(35)

35 | P a g e

Pvalue = .04125903

Graph 5.5: scatter plot of the residuals of the linear year effects model over time

Graph 5.6a: Scatter plot of the residuals against their value in the previous time

-. 2 -. 1 0 .1 .2 e h a t 1980 1990 2000 year -. 2 -. 1 0 .1 .2 e h a t -.2 -.1 0 ehat2

.5: scatter plot of the residuals of the linear year effects model over time

.6a: Scatter plot of the residuals against their value in the previous time period

2000 2010

.1 .2

.5: scatter plot of the residuals of the linear year effects model over time

(36)

36 | P a g e

5.3 Model 3: Fixed effects model with a linear year and level effects

Graph 5.7: Histogram of the residuals of the linear time and levels specific model

Pvalue = .03252572

Graph 5.8: Scatter plot of the residuals of the linear time and levels model over time.

0 1 2 3 4 D e n s it y -.4 -.2 0 ehat -. 2 -. 1 0 .1 .2 e h a t2 1980 1990 year

.3 Model 3: Fixed effects model with a linear year and level effects

.7: Histogram of the residuals of the linear time and levels specific model

STATA, with data from the OECD 2006.

.8: Scatter plot of the residuals of the linear time and levels model over time.

.2 .4

2000 2010

year

.7: Histogram of the residuals of the linear time and levels specific model

(37)

37 | P a g e

Graph 5.9: Scatter plot of the residuals of the linear time and levels specific model per country.

Graph 5.9a: Scatter plot of the residuals of the linear time and levels specific model against their previous value.

(38)

38 | P a g e

5.4 Model 4: Fixed effects model with a country grouping

Graph 6.0: Histogram of the residuals of the country specific model.

Graph 6.1: Scatter plot of the residuals of the country specific model over time.

Graph 6.2: Scatter plot of the residuals of the country specific model per country.

(39)

39 | P a g e Pvalue = .8291

Graph 6.2a: Scatter plot of the residuals against their value in the previous time period

-. 2 -. 1 0 .1 .2 e h a t -.2 -.1 ehat2

Scatter plot of the residuals against their value in the previous time period

0 .1 .2

ehat2

(40)

40 | P a g e

References

• Adkins, Lee C., and R. Carter Hill. Using Stata For Principles of Econometrics. 3rd ed. John Wiley & Sons, 2008.

• Aidt, Toke, and Zafiris Tzannatos. "The cost and benefits of collective bargaining: A survey." Social protection Discussion paper series; The world Bank 0120 (2001).

• Aidt, Toke S., and Zafiris Tzannatos. "Collective Bargaining and Economic Performance- A Short Review of the Theory." Industrial Relations Journal 5 (2002): 23-105.

• Aidt, Toke S., and Zafiris Tzannatos. "Trade unions, collective bargaining and macroeconomic performance: a review." Industrial Relations Journal 39 (2008): 258-95.

• Calmfors, Lars. "Centralization of wage bargaining and macroeconomic performance- A survey." OECD Economic studies (1993).

• Dell'Aringa, Carlo, and Laura Pagani. "Collective Bargaining and Wage Dispersion in Europe." British journal of Industrial relations 45 (2007): 30-53.

• Driffill, John. "The Centralization of Wage Bargaining Revised: What Have we Learnt." JCMS 44 (2006): 731-56.

• Hill, R. Carter, William E. Griffiths, and Guay C. Lim. Principles of Econometrics. 3rd ed. John Wiley & Sons, 2007.

• Machin, Stephen. "The decline of labor market institutions and the rise in wage inequality in Britain." European Economic Review 41 (1997): 648-57.

• OECD. "Wage-setting Institutions and Outcomes." OECD employment outlook 92-64-10812 (2004): 127-77.

• Rueda, David, and Jonas Pontusson. "Wage Inequality and Varieties of Capitalism." World politics (2000): 350-83.

• Visser, Jelle. "Patterns and variations in European industrial relations." Amsterdam Institute for advanced labor studies 1 (2000): 11-57.

(41)

41 | P a g e

• Visser, Jelle. "Patterns and variations in European industrial relations revised." European commission: Industrial relations in Europe (2004).

• Elvander, Nils. "The Labor Market Regimes in the Nordic Countries: A comparative analysis." Scandinavian political studies (2002).

• Elvander, Nils. "Collective Bargaining and Income policy in the Nordic Countries: A comparative analysis." British Journal of Industrial Relations 12.3 (1974). Print.

• Austen-smith, David, Jeffery A. Frieden, Miriam A. Golden, Karl Ove Moene, and Adam Przeworski.

The political economy of inequality, unions, and social democracy. Cambridge. Print.

• Andrews, Martyn, and RobinNaylor. "Declining union density in the 1980s: what do panel data tell us." British journal of industrial relations (1994). Print.

• Elvander, Nils. "The labor market regimes in the Nordic countries: a comparative analysis."

Scandinavian political studies 25.2 (2002). Print.

• Gregory, Mark. "Reforming the labor market: an assesment of the U.K policies of the Thatcher era."

The Australian economic review 31.4 (1998). Print.

• Machin, Stephen. "The decline role of labor market institutions and the rise in wage inequality in Britain." European economic review (1997). Print.

• Blau, Francine D., and Lawrence M. Kahn. At Home and Abroad: U.S. Labor-Market Performance in International Perspective. New York: Russel sage foundation, 2002. Print.

• Fuess, Scott M. "Union bargaining power: a view from Japan." Department of Economics, university of Nebraska. 2001. Web.

(42)

Collective Bargaining Effects on Wage Dispersion in Western Europe \ A. Ben-Moyal