Erling Barth, Institute for Social Research, Oslo


In every OECD-country, the gender wage gap has decreased over the last three decades (OECD 2002). This is true for the “raw” wage gap, the difference in pay between the average man and the average women. Also the “adjusted” wage gap has decreased in most countries. The adjusted wage gap is the average difference in pay between men and women with the same length of education and labour market experience. The gender wage gap has decreased considerably, but women still earn, on average, 16 percent less than men per hour worked (OECD 2002). This paper discusses the scope for a continuation of the reduction of the gender pay gap in Europe, with a particular focus on the role of unions and collective bargaining.

To make predictions about the gender pay gap, it is necessary to identify the sources of the differences in pay between men and women. The sources of the gender wage gap are thus discussed in the next section. Next, some underlying trends are presented and briefly discussed in relation to the mechanisms behind the gender wage gap. These are educational upgrading, increasing female labour market participation, a possible declining role of unions and more flexible work organisations. Lastly, some evidence is presented in an attempt to quantify some of the mechanisms at work.

The sources of the gender wage gap

The discussion in this paper is limited to hourly pay. This means that gender difference in working hours is not the issue here. The raw gender wage gap is simply the difference between the average hourly wage of men and women. In figure 1, the darkest columns show the raw gender wage gap in Norway for the years of 1980 and 2000. The figures are estimated from the Level of living surveys of these two years. [4] The wage gap has been reduced from about 22 percent in 1980 to about 17 percent in 2000. The next column shows the adjusted gender wage gap between men and women with the same level and field of education, the same labour market experience, the same sector, and major industry.

(Illustrasjon mangler)

Source: Estimated on data from the Level of Living Surveys each year. Calculated from the coefficient for gender (full-time employees). The grey columns also include years of education, broad occupational category, 1-digit industry, sector, experience, experience square and part time status interacted with gender.

If one further includes controls for establishment or detailed occupation, the wage differential drops to about 7-9 percent (see Barth and Mastekaasa 1996). Further control for very fine grid of detailed job-titles reduces the wage differential to almost nothing (See e.g. Meyersson et al. 2001). What does this tell us? It tells us that the wage differential we observe between men and women of about the same qualifications, is due to an interaction of wage differences and gender segregation across occupations, establishments and job-titles. Men tend to work in better paying occupations, establishments and jobs; - women in low paying jobs.

Let us first discuss the sources of wage differences in general. Wages differ according to productivity. Employees with different levels of human capital are paid differently. But, this is not the whole story. Groups of workers also differ with respect to bargaining power. Bargaining power arises mainly from the ability to threaten the employer with losses if an agreement is not struck, and secondly, from the ability to sustain a conflict over time. Workers may also have different degrees of market power. Market power consists of three features, the attractiveness of alternatives, the availability of alternatives and the ability to credibly threaten to exercise the option of quitting to such and alternative. So, there are three main sources of wage differences:

  • Productivity Differences
  • Bargaining Power
  • Market Power

When we observe a systematic segregation of men and women across jobs characterised by differences in pay, the main question is the direction of the causality. Is it the case that wages tend to be smaller where there are many females, for some or all of the different reasons listed above? Or is it the other way around; that women tend to crowd where pay is low - or in other words, that men tend to crowd where pay is high. If wages tend to be smaller where there are many women, it must be the case that women are less productive, have less bargaining power or less market power. If the causality is such, on the other hand, that women tend to end up in low paying jobs, it has to be the case that women either do not get the good jobs or because they do not go for them, to the same extent as men.

The answer, as I see it, is that the causality runs both ways, and that the mechanisms are interrelated. Let me explain with a brief discussion of what I think is an interesting example; namely the issue of mobility and employee stability. Many authors have argued that the instability of women employment is harmful for them (e.g. Polaheck 1986). Because young women traditionally expect career interruptions when they have children, they may not qualify for the fast-track career jobs. Knowing this, many women may not go for the most demanding careers either. This also means that they invest less in human capital, which again leads to lower productivity and lower pay. Lack of job stability among women, thus leads to both segregation and lower pay.

But consider another form of instability, namely the kind of quits that leads to better paying jobs. Job-to-job moves are known to be very profitable for careers, perhaps especially in the beginning of the career (see e.g. Topel and Ward 1992). A tendency to leave for better paying jobs, increases ones market power in the establishment. If someone is reliable and stays on regardless of what she gets, the employer has an incentive to keep wages down. It turns out empirically, that men are more willing and eager to quit their job in order to obtain higher wages (Barth and Dale-Olsen 1999). In this case, the cost minimising employer keeps women’s pay low.

So, one kind of mobility, the one which is not wage induced, is thus harmful for female dominated occupations for two reasons, both career options and human capital investment become smaller. On the other hand, a low level of wage-induced mobility is also harmful for female dominated occupations; compared to men, women tend to be stuck in low paying jobs, and this in itself induces lower market power and further lower pay.

These are two examples of traditional gender differences that involve causation both ways, female dominated jobs pay less, and women have a tendency to end up in low paying jobs. And the mechanisms tend to enforce each other. There are several other mechanisms which operate at the same time, which is also why the gender pay gap is so pervasive, both over time and across geography. But, since these stories involve self-reinforcing mechanisms, we may also expect there to be both large variations in gender wage differentials as well as potential for rather fast changes and considerable effects of particular changes in underlying trends. Over the last quarter of the last century, the gender wage gap shrank considerably - particularly in the 80’s. But now, it seems that the change has levelled out many places. Let us take a closer look at some recent empirical results for Europe.

Figure 2 shows adjusted gender wage gaps in Europe, estimated in 1997.[5] I have controlled for level of education, part time status and experience. The difference between countries is considerable. Our Nordic countries are represented by Denmark, the winner of the contest, so to speak, and Norway which is not far behind (Sweden probably goes in between here, but I do not have data from the ECHP to support this. See Edin and Richardson 2002 for recent evidence for Sweden.). Finland has a considerably larger gender wage differential than the Scandinavian countries.

The European trend over the last part of the nineties, is clearly but slowly downward, from 14 percent in 1994 to 12 percent in 1997. Figure 3 displays the estimated European trend as well as the estimated trends for the Nordic countries included in our data set. The picture confirms a downward trend in Denmark and Norway, but not for Finland.

(Illustrasjon mangler)

Source Own calculations based on year-to-year estimations of the type presented in figure 2 based on the same data.

Where are we heading?

We do not know what the future brings, of course, but consider the following trends that appear to be quite pervasive in Europe at large:

  • Increasing levels of education in the labour market
  • Increasing female participation rates
  • Declining role of unions in wage setting
  • More flexible work organisations

The following offers a brief discussion of the consequences of these trends on the gender wage gap.

The Education boom

We are in the midst of an educational boom in Europe, and women are catching up on men with respect to skills. In a European project we are conducting for the Norwegian Centre for Gender Equality, financed by the EU-Commission, we have estimated wage gaps for different groups (see Barth, Røed and Torp 2002). It turns out that while the average wage differential, after control for the level of human capital, for blue collar workers in the food industry in 14 European countries is about 17 percent, the similar wage gap for science professionals is 8 percent. Skills upgrading of women tend to reduce the wage gap, even within skill group. This is a heavy trend that works against most of the mechanisms discussed above.

On the one hand, increased education increases the labour market attachment of women, since women with higher education have invested more in labour market skills. Furthermore, high-education jobs seem to be organised along less traditional gender lines than low-skilled jobs. On the other hand, it seems that the competition hardens at the very top of the hierarchy, and it is still difficult to dismiss the notion of a glass ceiling towards the very top.[6] All in all it, seems that the considerable skills upgrading going on, in particular among women, is working towards a closing of the gender gap.

Increasing female participation rates

The other obvious trend is increased labour market participation of women. At first thought, one would think that a country with a high female labour participation rate is a more gender equal society, and thus produce smaller wage gaps. But the opposite is true, and the reason is simple. As a country increases the labour market participation of women, it taps into the part of distribution of women who have the lowest labour market attachment. Those who enter the labour market first, are the ones who look most like men, in terms of labour market behaviour, while the last ones are the ones who would otherwise stayed home. The new waves of women enter the tertiary sector, and in many countries they enter public services. This is one reason why we find low wage gaps in Italy for instance, and perhaps the other way around for Finland.

Declining unions and decentralisation of bargaining

We first note that in line with the increased female participation rates, there has been a strong increase in the share of women within the unions. Figure 4 shows the average share of women in the main confederation in some European countries. The average share of women has increased from about 18 percent in the 50’s to well above 40 percent in 1995. This implies that the bargaining power of women has increased relative to men over this period, both from the increased bargaining strength towards employers, but also through a strengthened collective voice in the process of policy making within the large confederations.

(Illustrasjon mangler)

In recent years several observers note that it appears to be a decline in union density across many countries in Europe. However, looking at the evidence, Ebbinghaus and Visser (2002) find no consistent pattern of declining unions in Europe. They find that “trends in union membership and density did not converge. The cross-national variations in density rates have increased rather than decreased” (p. 62). In particular, we have not seen a decline in membership in the Nordic countries.

There are however, signs of fragmentation and there seem to be powerful movements towards more decentralisation of wage bargaining in many countries. Fragmentation and reduced union density many places, add to the co-ordination problems, of course, since free-rider problems become large at the national level.

How do these trends affect the gender wage differential?

It is well known that unions tend to compress wage differences in general, - in particular, this is the case for co-ordination in bargaining. Blau and Kahn (1996), accordingly, argue that the main effect of unions on the gender wage differential, is simply to compress the wage structure. Even without changing the relative position for men and women on the wage ladder, a reduction of the distance between the steps will reduce wage differentials. This is one consequence of declining union membership and bargaining coordination.

How the decline in unionism will affect the relative bargaining power of men and women is, however, not clear. Ebbinghaus and Visser (2000) find that the decline in unionism is mainly among men. This evidence is in line with the finding of increased female share among unionised workers, and works opposite to the wage compression effect: with declining bargaining power among men, women may do relatively better.

Other trends in the “new economy”

Finally, and related to both the decline of unions and skills upgrading in the European labour market, we have other important issues around the “new economy”: More autonomous workplaces, flexible organisations, decentralised and performance related pay. It is not clear how this should affect the gender wage gap. But to the extent that wage differences are increased, and the pattern of segregation prevails, it will tend to increase gender wage differentials. It is, however, not clear how, why or to what extent these new trends should add to segregation and the systematic sorting of men and women across jobs or tasks with different pay.

So all in all, we have a mix of trends. The main gravitational forces, pulling down the gender wage gap, are the skill-upgrading and increased bargaining influence of women, making women more like men in the labour market. The other factors are working against this effect, but as we saw, at least on average for the years 1994 to 1997, did not offset it completely. So let us see if we can sort out some of these effects empirically.

Some European-level evidence

Using the data from the 14 EU-countries over the four years, we obtain 51 observations of the gender wage differential in Europe. The wage differential is estimated from almost 180-thousand observations of individual wages in these countries, and includes control for human capital and part time affiliation.

Table 1 reports the results from a regression analysis of the 51 observations of year times country.[7] Firstly, in model 1 we find that the wage gap is declining at a rate of almost half percentage-point per year, however, not statistically significant. Secondly, we find that increasing the share of workers with tertiary education, (college and university level) by 10 percentage points, reduces the wage differential by 1 percentage point. Thirdly, we find that increasing female participation increase the wage differential considerably. In model 2, I include the share of part timers and a Duncan-index of occupational segregation. In line with the previous discussion, the gender wage differential is increasing in occupational segregation, and at least some of the effect of female participation is captured by higher occupational segregation.


Model 1
Model 2
Model 3
Model 4
-0.410 (0.415)
-0.572 (0.364)
-0.617Z (0.340)
-0.484 (0.297)
Tertiary education
-0.093* (0.045)
-0.080* (0.041)
-0.109* (0.052)
-0.036 (0.049)
Share of females
0.221z (0.118)
0.168 (0.117)
0.412* (0.139)
0.309* (0.124)
Share part time
-0.214* (0.095)
-0.026 (0.137)
0.203 (0.133)
Occupational Segregation
0.273* (0.089)
0.191 (0.124)
-0.114 (0.135)
Union density
-0.056* (0.026)
-0.047Z (0.026)
-0.040 (0.035)
0.044 (0.038)
Coll. coverage
0.145* (0.070)
0.208* (0.063)
Centr. Index
-2.178 (1.551)
-1.280 (1.364)
Residual wage dispersion
0.450* (0.120)
Adj. R-square



Lastly, we find that increasing union density by 10 percentage points, is associated with half a percentage point reduction in the gender pay differential. Model 3 introduces some country specific variables representing the prevalence of collective coverage, which may be very different from union density in some countries, as well as an index of co-ordination in bargaining. There is a strong, but not statistically significant, negative relationship between the wage differential and the level of co-ordination. The point estimate here indicates that going from a completely decentralised system to a fully co-ordinated one, is associated with a reduction of the wage differential by more than 6 percentage points.

I argued that the main influence of unions on the gender wage gap is through a reduction of wage differences in general. This is apparent from model 4. Here a measure of wage dispersion is inserted into the equation, after controlling for the variation in human capital. We find a strong association between general wage dispersion and the gender wage gap, of course. Furthermore, we find that the point estimate for centralisation drops by almost one half, and that the relationship between union density and the gender wage gap ceases to be significant, and even switches sign. This strongly indicates that the main role of unions and bargaining co-ordination is to reduce the overall wage dispersion, and not to improve women’s position per se.

Trends in the “new” economy, some cross-section evidence from Norway

We are then left with the question of how women fare in the “new economy”, with higher levels of autonomy, more individual contracts and performance pay. Figure 5 shows some preliminary results based on an analysis from Norway. The gender wage differential is simply estimated separately within the sample of private sector establishments where the manager has reported certain indicators of decentralisation of authority and pay.[8] The solid line is the average gender wage differential in the private sector this year. Using only employees in establishments that use individual contracts only, (about 13 percent of the private sector labour force work in such establishments), we find a slightly higher wage differential of about 13 percent. In establishments where the manager reports that the core occupation has large autonomy, the wage gap is above 13 percent. Note that these groups are in no way distinct, but to a large extent overlapping. In establishments that use performance pay, the same number is above 14 percent.

(Illustrasjon mangler)

There are thus significant, but rather small, differences between the more traditional workplaces, and the ones that many people view as belonging to the “new economy”. However, we should be careful in interpreting these results, since they do not control fully for potential differences in working hours between the genders represented in the different samples (see footnote 5). In line with the above observation, we find lower wage gaps on average in the group of establishments that have centralised bargained wages only,[9] and in general for the group of unionised establishments (which comprise above 93 percent of all employees in the private sector in Norway). This is probably a result of the general wage compressing effect discussed in the previous section.

Summing up

Gender wage differentials have been decreasing over the last two or three decades. The gender wage gap arises from a combination of wage differences and gender segregation. In order to assess the potential for a further reduction in the pay gap, it is necessary to understand both how wage differentials and the negative patterns of segregation arise.

Major forces determining wage differences in general, are supply and demand conditions, together with bargaining arrangements and the strength of unions. It seems that all in all, the increase in the supply of higher education in Europe has, together with coordinated bargaining and relatively strong unions, kept the overall wage dispersion at bay. In a study of the US labour market, Blau and Kahn found that women were “swimming upstream” in the labour market. Women’s position was steadily improved, but not without resistance from powerful currents. In the US, the main obstacle has been a general increase in wage dispersion. Apart from the UK, this has not been the case on a large scale in Europe.

A high level of union density and coordination in bargaining produces lower gender wage gap through their effect on overall wage dispersion. As shown by Ebbinghaus and Visser (2002), however, there are no consistent trends in union membership and density in Europe at present, and it is not clear in what direction union density is developing. Furthermore, it seems that what we have observed in terms of declining union membership, occurred for the most part among men. So there is no clear evidence indicating that the development in union membership is a major factor working against further reductions in the gender pay gap. However, it seems that there is an increased level of union fragmentation and a movement away from heavy coordination in bargaining. Together with some of the features of the “new” economy at the plant level, these trends may add to an increase in wage differences also between men and women.

Concurrently, there are factors affecting the level of gender segregation as well. The increased participation rate of women, in particular together with a “marketisation” of care, tends to increase segregation and thus the observed gender pay gap in the labour market. Of course, if household work is counted in, gender pay equality is improved by this trend, but since we measure the pay gap in the paid labour market only, the observed effect is to increase the gender pay gap. The educational boom works in three ways. First women are catching up in terms of human capital, thus giving a reduction in the raw wage gap, secondly, women’s investment in human capital increase their labour market attachment, affecting the adjusted wage gap as well. It is also my conjecture that increased education reduces the level of segregation, since many of the high-skilled jobs are less bound to traditional gender patterns of activity in the household, than are many low-skilled jobs. However, it is also the case that as women work their way towards the top of the organisations, they may meet harder competition, and eventually meet some sort of “glass ceiling”, preserving the present level of segregation at the top. It is not clear how more flexible organisational-, pay- and job-patterns affect segregation. On the one hand, more flexibility may by itself tend to reduce job-segregation. On the other hand, this type of working relations may induce new forms of sorting of men and women between jobs.

All in all, there are concurrently several trends with conflicting effects on gender differences in the labour market. These trends affect both wage differentials and gender segregation. Also developments regarding collective bargaining and wage setting regimes have partially offsetting effects on the development of the wage gap. It seems likely, however, that the gender wage gap will continue to decline also in the next decades.


Albrecht, J., A. Björklund and S. Vroman (2003): ”Is there a Glass Ceiling in Sweden?”, forthcoming in Journal of Labor Economics.

Barth, E. (1992): Lønnsforskjeller mellom kvinner og menn i Norge (Wage differences between men and women in Norway), Rapport 92:7. Oslo: Institute for social research.

Barth, E. and H. Dale-Olsen (1999): Monopsonistic Discrimination and the Gender Wage Gap, Working Paper No. 7197, National Bureau of Economic Research, Cambridge, Massachusetts.

Barth, E. and A. Mastekaasa (1996): “Decomposing the Male-Female Wage Gap: Within and Between Establishment Differences”, Labour, Vol.10 (2), 339-356.

Barth, E., M. Røed and H. Torp (2002): “Towards a Closing of the Gender Pay Gap, A comparative study of three occupations in six European countries”, The Norwegian Centre for Gender Equality, Oslo.

Blau, F. and L. Kahn (1996): “The Gender Earnings Gap: Learning from International Comparisons,” American Economic Review (May 1992), 533-538.

Datta Gupta, N., R. Oaxaca and N. Smith (2001): “Swimming Upstream, Floating Downstream: Trends in the US and Danish Gender Wage Gap”, CLS WP 01-06, Aarhus School of Business Administration.

Edin, P.-A. and K. Richardson (2002): “ Swimming with the Tide: Solidarity Wage Policy and The Gender Earnings Gap”, Scandinavian Journal of Economics, 104:49-67.

Ebbinghaus, B. and J. Visser (2000): Trade Unions in Western Europe since 1945. The Societies of Europe, Mannheim Centre for European Social Research.

Meyersson Milgrom, E.-M., V. Snartland and T. Petersen (2001): “Equal Pay for Equal Work? Evidence from Sweden, Norway and the U.S.”, Scandinavian Journal of Economics, Vol. 103.

OECD (2002): OECD Employment Outlook. Paris, July 2002.

Polachek, S. (1981): “Occupational Self Selection: A Human Capital Approach to Sex Differences in Occupational Structure”, The Review of Economics and Statistics, 63:60-69.

Topel, R. and M.P. Ward (1992): “Job Mobility and the Careers of Young Men”, Quarterly Journal of Economics, 107:441-79.

[4] The data is provided by Statistics Norway. When not specifically stated, the results reported here are based on my own calculations.[]

5 The data source is the European Community Household Panel. The Norwegian data are taken from the Level of Living Survey. When not specifically stated, the results reported here are based on my own calculations.[]

6 Both from Denmark (Datta Gupta et al. (2003)) and Sweden (Albrecht et al. (2003)) there are some recent studies pointing to higher wage differentials between men and women at the very top of the wage distribution. It is also the case that men is grossly overrepresented in top-jobs, but we are still a bit early to assess to what extent this is a result of educational differences 30 years ago or the labour market of today, simply because the women of the education boom have just recently began to compete for the top-jobs.[]

7 This analysis is discussed in Barth et al. (2002), who also provide a detailed comparative analysis of three occupations in six European countries.[]

8 We should be somewhat careful in interpreting these results, since the data in this survey does not include individual information on working hours, but use imputed values of this variable, which does not vary across the strata.[]

9 This observation is not in accordance with what is found in Barth (1992) on a different data set, and thus warrants further research.

Publisert 25. nov. 2010 13:52 - Sist endret 14. nov. 2013 13:39