SEARCH

SEARCH BY CITATION

Abstract

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

This paper examines differences between women’s and men’s wages in 18 selected OECD countries in the period 1970 to 2005. The study is based on 12 manufacturing sector- and skill-specific sets of panel data on the gender wage gap. We apply a system generalised method of moments (GMM) estimator to the extended version of the conditional gender wage gap convergence equation, controlling for sector concentration and industry-specific measures of openness using a difference-in-difference approach: trade-affected concentrated sectors versus trade-affected competitive sectors. The results indicate that: (i) an increase in sector concentration is associated with wage gap growth; (ii) both import and export penetration are associated with a reduction of the high-skill gender wage gap growth in concentrated industries; (iii) there is evidence of a widening impact of trade on the medium and low-skill occupational gender wage gap growth in less competitive industries; (iv) institutional regulations of the labour market have an impact on the development of the gender wage gap: for highly-skilled labour an increase in labour market regulation raises the growth of the gender wage gap, while for medium- and low-skilled workers, it lowers it.


1. Introduction

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

‘A crucial question concerns the sharing of potential gains from globalisation, between rich and poor countries, and between different groups within a country’ Sen (2004, p. 18).

Gender discrimination means unequal treatment based solely on a person’s sex. Discrimination can occur in a variety of contexts. For example, the Global Gender Gap Index calculated by the World Economic Forum takes into account economic, political, educational and health-based criteria.1 According to the International Labour Organization (ILO), discrimination in the labour market, when it comes to employment and occupation, is defined as ‘treating people differently and less favourably because of certain characteristics, such as their sex [...] irrespective of their merit or the requirements of the job’ (ILO, 2003, p. 15). One of the main ‘material’ results of such differential treatment is the gender wage gap: a difference in remuneration paid to female and male workers for the same job that is not justified by any characteristic other than sex. There is a common consensus that, although there is a decreasing trend, wage gender discrimination still exists, even in the most developed countries (World Economic Forum, 2010) despite national and international laws and regulations against gender discrimination.2 Hence, both the determinants and consequences of the existing persistence in the gender wage gap are prominent in political and economic discussion (to name a few recent studies, Weichselbaumer and Winter-Ebmer, 2005; Arulampalam et al., 2007; Neumayer and De Soysa, 2007; Bardasi and Gornick, 2008; Binder et al., 2010; Gradin et al., 2010; Seguino, 2011; Schober and Winter-Ebmer, 2011). The theme has gained major importance in recent years within the debate on the impact of globalisation (mainly through trade liberalisation) on the position of women in the labour market. The main question is whether trade policy reduces or reinforces gender inequalities (for a detailed literature review on gender discrimination and trade, see Van Staveren et al., 2007; Fontana, 2009).3

Discussion on the impact of trade on the gender wage gap needs to make a clear distinction between an impact on overall gender wage differentials (raw wage gap) and on the residual wage gap (understood as the gender wage gap that remains after controlling for differences in factors such as qualifications, experience and skills).4 For example, in the classic Hecksher-Ohlin-Samuelson (H-O-S) theory, trade increases the relative demand for intensively used factors of production and raises their relative remuneration. If there is a greater proportion of women among less-skilled workers, then trade expansion will positively impact on their relative demand and wages in countries specialised in low-skill abundant goods (i.e. developing countries). This will affect the raw gender wage gap but not necessarily the residual gender wage gap if skills are controlled for.

The theoretical background to wage discrimination consists of at least two main countervailing streams: one neoclassical (Becker, 1971) and the other non-neoclassical (Williams, 1987). The first argues that under the assumption of a competitive market, discrimination, being costly to employers, will not persist. Consequently, as increasing trade (e.g. import penetration) results in greater competition within the domestic market, the male/female wage gap should narrow. At the same time, as jobs shift to export-oriented industries, these sectors should absorb increasing numbers of women, and the shift in demand for female labour should increase women’s wages in relation to men’s (Wood, 1991).

On the other hand, the non-neoclassical approach assumes that increasing foreign competition may weaken the position of female workers by lowering their bargaining power. Women will be over-represented in sectors characterised as being competitive due to low labour costs, and to maintain their competitiveness, there will be further downward pressure on females’ wages in the industries affected by increasing trade, causing an increase in the unadjusted wage gap. Additionally, this effect can be worsened by a shift in the demand for specific skill categories. For example, if technological progress favours specific occupations (so-called ‘skill-biased technological change’), this will result in rising demand for high-skilled workers and, under the assumption that women, on average, tend to possess fewer skills than male workers, their situation will worsen. Trade could have a similar effect, as a widening difference between the female and male payroll could be simply due to a widening skilled-unskilled wage gap (Wood, 1995). The basic argument that technological change is expected to widen the gender wage gap is connected with the masculinisation of technical jobs and the potential exclusion of women from training for more technologically sophisticated jobs (Kongar, 2006). Again, as in the case of the H-O-S theory, trade connected with skill-biased technological change can increase the skilled–unskilled wage gap and therefore the unadjusted wage gap, but not necessarily the residual wage gap.5

However, thorough analysis suggests that the relationship between the gender wage gap and industry/economic openness is still unresolved and the results in the empirical literature are not without ambiguity (Anderson, 2005; Van Staveren et al., 2007).

Black and Brainerd (2004) and Menon and Van Der Meulen Rodgers (2009) show that the impact of trade depends on the structure of the market – concentrated versus non-concentrated. If a market is non-concentrated (competitive), then there is no room for discrimination; an increase in trade affects concentrated markets, and to escape foreign competition firms may either reduce discrimination, to lower costs, or only decrease the wages of already discriminated-against workers.

Still, the ‘gender wage gap – trade nexus’ is, to a large extent, an empirical issue that in spite of having been studied has not been fully characterised. Our research is motivated by the fact that most previous studies have considered one, or a limited number, of countries and have had a micro-level character (Weichselbaumer and Winter-Ebmer, 2005). In this context, the Indian economy has been analysed by, among others, Chamarbagwala (2006) and Menon and Van Der Meulen Rodgers (2009), while the Mexican economy has been studied by Artecona and Cunningham (2002) and Hazarika and Otero (2004), the United States by Black and Brainerd (2004), China by Shen and Deng (2008) and Germany by Antonczyk et al. (2010).

Few studies have looked at the impact of trade on the gender wage gap in an international setting. In particular, Gupta (2002) examines a cross-section of 21 developing countries to identify the negative effects of export-orientation on male/female non-agricultural wage ratios, after controlling for alternative factors that might raise female wages. Similarly, Oostendorp (2009), using ILO data for more than 80 economies, shows that, in general, more trade is associated with a lower gender wage gap. However, Oostendorp (2009) underlines the differing impact of trade on workforces with different skills; trade narrows the occupational gender wage gap for low-skilled occupations, both in poorer and richer countries, and for high-skill occupations in richer countries only. A widening effect of trade on the gender wage gap was confirmed by Berik et al. (2004) for Taiwan and the Republic of Korea. Finally, Weichselbaumer et al. (2007) uses meta-analysis to arrive at the conclusion that there is a strong negative correlation between market competitiveness (also measured by trade flow) and the gender wage gap.

Taking into consideration the limitations of the previous analyses (limited country and time coverage), we propose an empirical study on the effects of domestic and foreign competition on male/female wage differentials for 12 manufacturing sectors in 18 OECD countries, 14 of which are in Europe, for the period between 1970 and 2005. We do not limit ourselves to country-specific or sector-specific evidence, but draw our conclusions by using a database constructed for the purpose of this study and match, at the disaggregated industry level, wage statistics with measures of domestic and foreign competition. We use industry statistics (EU KLEMS, 2008) on female and male wages at relatively disaggregated industry levels (NACE, 2 digit sectors) that distinguish between wages paid to different groups of workers classified according to skill level: high, medium and low. To the best of our knowledge, this database has not been previously used in analyses of gender wage differentials.

The remainder of this paper is structured as follows: in Section 2, we present the theoretical basis of our investigation, together with formal modelling of the impact of domestic concentration and trade on the gender wage gap. In Section 3, we describe trends in the share of employment and wage ratio across different industries and in terms of skill specifications. In Section 4, we estimate a dynamic model of the gender wage gap, revealing the effects of domestic and foreign competition on the evolution of wage differential using a difference-in-difference approach: trade-affected concentrated sectors versus trade-affected competitive sectors, together with numerous robustness checks. Finally, Section 5 concludes.

Our main findings are the following: (i) an increase in sector concentration is associated with wage gap growth; (ii) both import and export penetration are associated with a reduction of the high-skill gender wage gap growth in concentrated industries; (iii) there is evidence of a widening impact of trade on the medium- and low-skill occupational gender wage gap growth in less competitive industries; (iv) institutional regulations of the labour market have an impact on the development of the gender wage gap: for highly skilled labour, an increase in labour market regulation raises the growth of the gender wage gap, while for medium- and low-skilled workers, it lowers it.

The results have straightforward policy implications, especially with regard to the education and skill-upgrading of female workers.

2. Theoretical Model

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

We present a model that shows the impact of trade on the gender wage gap, largely following the notation and exposition of Menon and Van Der Meulen Rodgers (2009), which, in turn, is based on Borjas and Ramey (1995). It is assumed that the domestic economy consists of two sectors: a competitive sector (sector 0) and an uncompetitive sector (sector 1), the latter being the concentrated one. Two consumption goods are produced: x by the competitive sector and y by the concentrated sector. The analysis below is restricted only to the concentrated sector where wage discrimination exists. An inverse demand curve represents the total demand for good y by domestic consumers:

  • image(1)

The inverse demand curve perceived by each domestic firm i is given by:

  • image(2)

where yi is the output of firm i, and yj is the output of each other domestic firm and m is net imports (positive if imports are higher than exports, negative otherwise). Firm i’s rent is simply the difference between total revenues and labour costs:

  • image(3)

where L1i indicates the number of employees of firm i in the concentrated sector and w1 is the wage in the concentrated sector. There are two groups of employees: Lm– males and Lf– females. We assume that discrimination present in the concentrated sector causes male and female wages to differ. The discrimination coefficient (d) is defined as the proportional difference between male and female wage rates (Becker, 1971):

  • image(4)

where inline image and inline image denote, respectively, male and female wages in the concentrated sector.

In the absence of discrimination, the equilibrium wage rate of men would equal that of women (e.g. in the competitive sector: inline image). If men and women are imperfect substitutes, they may receive different wage rates even in the absence of discrimination. A more general definition of the discrimination coefficient is that it is equal to the difference between male and female wages with and without discrimination:

  • image(5)

The optimal level of production yi is the result of maximising equation (3)6

  • image(6)

and the equilibrium rent for firm i equals:

  • image(7)

To obtain equilibrium wages for workers in the concentrated sector, we assume that the rents obtained by the workforce in that sector equal the part of the equilibrium rents (i.e. inline image):

  • image(8)

Coming to the difference in wages paid to female and male workers, we further assume that the equilibrium wage described in equation (8) is the weighted average of male and female wages:

  • image(9)

where inline image is the ratio of male workers to the total number of workers in the concentrated sector. Putting equation (4) into equation (9) and rearranging for the equilibrium wage of females, we obtain

  • image(10)

while for males, we obtain

  • image(11)

Further, we follow the proposition of Menon and Van Der Meulen Rodgers (2009) and model the relationship between the discrimination coefficient and net imports:

  • image(12)

but contrary to their interpretation, we do not a priori impose the sign of the above relationship (whether α1 is positive or negative will be tested in the empirical section).

This model has a number of implications. First, an increase in net imports (or a drop in exports) decreases rents (equation 7) and consequently equilibrium wages for workers in the concentrated sector (equation 8). The impact of imports on rents and wages is more pronounced in less competitive sectors (with lower n in equations 7 and 8). Second, because the rents of domestic firms fall as imports increase, the impact on differences between male and female wages depends on the reactions of firms affected by lower rents. If they are forced to cut costs (including the cost of maintaining discrimination), in order for them to remain competitive, the wage inequality between males and females should fall (then α1 < 0), alternatively a firm can also cut costs by hiring more females than males. In the alternative interpretation of the model, trade-affected firms can maintain higher male wages at the expense of female wages, with the latter dropping more and the gender wage gap increasing and/or less profitable firms (previously non-discriminating) exiting the market (then α1 > 0). Reduced employment in a trade-impacted concentrated sector might force women to move to the lower-paying competitive sector (crowding out effect), and overall, the female wage falls relative to that of men (Borjas and Ramey, 1995). Finally, according to the model, imports and exports have opposite impacts on equilibrium wages in the concentrated sector. However, this opposite effect is not unambiguous, as exporting firms might also be forced to reduce costly discrimination to be competitive when entering foreign markets (Berik et al., 2004).

From a theoretical point of view, the effect of an increase in trade on gender wage differentials is not explicit and therefore needs to be tested empirically.

3. Empirical Setting

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

a. Data and Measurement Issues

Specifically for the purposes of this study, we construct a database matching labour market data (labour compensation and hours worked) and trade statistics at the level of single industries. We draw on industry-specific data that have recently been made available by the EU KLEMS Growth and Productivity Accounts database.7 EU KLEMS report data at the lowest level of aggregation for 71 industries grouped into 20 categories, according to the European NACE revision 1 classification, but labour input files are available only at more aggregated levels for the restricted sample of countries and time covered in this study. We use information on labour compensation and hours worked to calculate the wages per hour of different categories of workers: high-, medium- and low-skilled labour. Skills are defined here on the basis of educational attainment. High-skilled corresponds to tertiary education, medium skilled to upper secondary education and low-skilled to primary education.

The EU KLEMS database contains information on the share of total hours worked and the share of total labour compensation for both male and female workers, grouped by skill and age. First, we construct the share of labour compensation and hours worked for gender and skill categories of all ages by simply adding the values for persons aged 15 to 29, persons aged 30 to 49 and persons aged 50 and over. Then, we construct the female-sector-specific wages per hour:

  • image(13)

where i denotes sector, j country, t time and S the skill category of workers (high-, medium- or low-skilled). LABij,t is the value of total labour compensation in a given sector and country, and inline image is the share of labour compensation obtained by females with given skills (e.g. the labour compensation of high-skilled, medium-skilled and low-skilled females of all ages). Hij,t refers to the total number of hours worked by all persons engaged in a given sector i, and inline image to the share of hours worked by females of all ages and with given skills (e.g. hours worked by high-skilled, medium-skilled and low-skilled females of all ages).

Analogous calculations are carried out to obtain male wages (inline image), this time using the share of labour compensation and hours worked by male workers in given skill categories.

The rest of the sector-specific data, especially on imports, exports, gross output and value added, comes from the OECD STAN (released in 2009), which are classified according to ISIC Rev. 3. Table A1 in the Appendix shows the accordance between the ISIC Rev. 3 and NACE classifications.

Unfortunately, data on labour compensation and hours worked are not available at the most disaggregated level and for the whole sample of countries, which influences the final composition of our panel data set. In the end, our unbalanced panel is composed of 18 OECD countries (14 European) and 12 manufacturing sectors categories for the years 1970 to 2005. A list of sectors is provided in Table A1 and a list of countries in Table A2 in the Appendix.

b. Trends in Employment and Wages

Our panel data set is very detailed as far as country, industry and time span are concerned and a number of specific analyses can be performed on it. We restrict ourselves here to analyses of labour markets: trends in employment, skill shares and – what is our main concern – female and male wage shares and gender wage ratio dynamics.

The descriptive evidence from Table 1 on hours worked by women shows that female participation in the total economy rose from 32.5 per cent in 1970 to 41.3 per cent in 2005. There are some feminised industries where more women than men work: financial intermediation (52.8 per cent of hours worked by women), education (67 per cent) and, the most feminised, health and social work, with 75 per cent of hours worked by women. The most masculine industries are the following: construction (8 per cent of hours worked by women), mining and quarrying (17.3 per cent) and electricity, gas and water supply (19 per cent). Between 1970 and 2005, the biggest increases in the female share of work hours were experienced in public administration and defence, where the figure rose from 28.5 to 46.6 per cent; the sale, maintenance and repair of motor vehicles (29 to 41.9 per cent) and health and social work (62.9 to 75.6 per cent). On the other hand, two industries – textiles and food products – saw a noticeable rise in the share of male employees, although they still have the largest share of female workers among manufacturing sectors.

Table 1. Hours Worked by Females in % (All Countries Pooled Together)
NACEDescription of Sectors197019902005
  1. Source: Own calculations based on data from EU KLEMS 2008.

15t16C15T16 Food products, beverages and tobacco49.539.342.1
17t19C17T19 Textiles, textile products, leather and footwear62.751.147.7
20C20 Wood and products of wood and cork25.520.824.4
21t22C21T22 Pulp, paper, paper products, printing and publishing30.625.027.8
24C24 Chemicals and chemical products29.924.527.8
25C25 Rubber and plastics products31.624.327.7
26C26 Other non-metallic mineral products22.420.724.4
27t28C27T28 Basic metals and fabricated metal products21.019.723.8
29C29 Machinery and equipment, n.e.c.26.919.922.8
30t33C30T33 Electrical and optical equipment33.626.926.3
34t35C34T35 Transport equipment28.117.821.3
36t37C36T37 Manufacturing n.e.c. and recycling40.929.434.3
50C50 Sale, maintenance and repair of motor vehicles and motorcycles – retail sale of automotive fuel29.033.341.9
51C51 Wholesale, trade and commission excl. motor vehicles34.938.844.4
52C52 Retail trade excl. motor vehicles – repair of household goods42.744.749.0
60t63C60T63 Transport and storage15.518.821.4
64C64 Post and telecommunications33.628.337.7
70C70 Real estate activities42.341.442.5
71t74C71T74 Renting of mach. and equip – other business activities39.739.940.8
AtBC01T05 Agriculture, Hunting, Forestry and Fishing18.726.329.8
CC10T14 Mining and Quarrying13.612.617.3
DC15T37 Manufacturing37.528.428.7
EC40T41 Electricity gas and, water supply15.214.919.8
FC45 Construction7.88.68.6
GC50T55 Wholesale and retail trade – restaurants and hotels38.641.445.5
IC60T64 Transport, storage and communications19.921.124.8
JC65T67 Financial intermediation45.146.352.8
KC70T74 Real estate, renting and business activities40.740.041.0
LC75 Public admin. and defence – compulsory social security28.535.746.6
MC80 Education59.855.667.0
NC85 Health and social work62.969.475.6
TotCTotal total32.537.141.3

There are, of course, cross-country differences. Table A3 in the Appendix provides a detailed description of the percentage of women in total employment, by main economic activity in the countries analysed. In the majority of countries for which data are available, women are employed mainly in services: health, education, community, social and personal services. The maximum women’s share of any sector is recorded in Finland, where almost 90 per cent of employees in the health sector are women, while the minimum is found in Spain’s construction sector, where women constitute only 5 per cent of workers.

It is very important to note the female and male shares of employment in specific skill categories. Table A4 in the Appendix shows female and male shares of employment in high-skilled, low-skilled and medium-skilled workforces by economic activity in 1970 and 2005. In 2005, skilled labour constituted 20.53 per cent of all employees, low-skilled 18.37 per cent and medium-skilled 61.1 per cent. Unsurprisingly, the biggest share of skilled labour was in education, while the largest share of low-skilled labour was in manufacturing (food and textiles). Among skilled workers, females and males are not equally represented. In health, education, community and social services, there are larger shares of skilled females than males, while in the other sectors analysed the opposite is true: there are larger shares of high-skilled males. The difference is especially pronounced in the case of electrical equipment and the renting of machinery and equipment. Other business activities, including finance, insurance, real estate and business services, also favour males, by up to 10 percentage points. In the case of low-skilled labour, the biggest difference between female and male shares of employment exists in private households with employed persons (low-skilled females 14.89 per cent, low-skilled males 6.48 per cent) and construction (low-skilled females 1.48 per cent, low-skilled males 19.29 per cent).

Of course, the distribution of hours worked by different skill categories has changed over time. Generally, the high-skilled workforce has increased in all industries; for industry overall, it rose from 9.7 per cent in 1970 to 20.54 per cent in 2005, while the low-skilled workforce decreased. However, this increase was not equally distributed between women and men. The greatest increase in the female share of employment in skilled labour categories occurred in education and financial intermediation, with rises of more than 10 percentage points. The share of high-skilled males rose by more in the machinery, electrical and optical equipment and transportation equipment industries. The share of hours worked by low-skilled labour, both for females and males, decreased in all the sectors analysed, while the greatest drop for women was in private households with employed persons (in 1970, the share of low-skilled females was 49.93 per cent; in 2005, 14.89 per cent) and the largest decrease for low-skilled men was in mining and quarrying (from 45.56 per cent in 1970 to 2.81 per cent in 2005).

Finally, Figure 1 presents the male/female wage ratio, which measures the hourly gender wage gap. At all skill levels, women earn significantly less than men. The greatest differences are for low-skilled workers, where on average in 2005 women’s earnings represented around 70 per cent of men’s earnings (the inverse of the male/female ratio) and the smallest for medium-skilled workforces, where women’s wages constituted 78 per cent of men’s. In the time period analysed, a decreasing trend in the male-to-female earnings ratio is observed. The greatest drop was for low-skilled categories, which, however, started from the highest point in 1970, and so the figure is still significant.

image

Figure 1.   Trends in Male/Female Wages – All Industries Total (All Countries Pooled) Source: Own compilation based on the data from EU KLEMS 2008.

Download figure to PowerPoint

4. Empirical Analyses of the Trade – Gender Wage Gap Nexus

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

a. Empirical Strategy

We define the gender wage gap as the log-wage differential between males (M) and females (F) possessing comparable skills:

  • image(14)

where as before i refers to sector, j to country, S to skill classification (high-skilled wage-whs, medium-skilled wage-wms and low-skilled wage-wls) and t to time period. The shortcoming of our database is its macroeconomic nature. We do not possess information about education, qualification, experience, etc. that would make it possible to calculate the residual gender wage gap, that is, the gender wage gap that remains after controlling for differences in those factors. Because of that we are forced to assume that female and male workers have similar abilities in the three skill categories (low, medium and high), and, consequently, we treat wage differentials from equation (14) as a proxy for the residual wage gap. A similar approach has been taken in other macroeconomic studies (e.g. Oostendorp, 2009; Dominguez-Villalobos and Brown-Grossman, 2010).

To quantitatively address the interplay between changes in the gender wage gap and trade factors, we fit the following dynamic model in a difference-in-difference setting:

  • image(15)

where inline image is the first difference in the gender wage gap defined in equation (14) and Trend indicates time trend.8Conij,t measures sector concentration and is proxied by the price cost margin (PCM), measured, as in Aghion et al. (2008), by the difference between value added (VA) and labour compensation (LAB_COMP), as a proportion of gross output (GO): inline image. The PCM shows the Lerner index of pricing power. The index is in the range (0,1); the higher the index, the higher the pricing power and the lower the competition pressure. When analysing the relationship between the gender wage gap and trade flows, we need to distinguish between import and export flows, which can have different effects on the host economy. Consequently, increasing openness to trade is measured by changes in the trade penetration ratio, which alternatively denotes imports (ΔImpij,t), exports (ΔExpij,t) or total trade (ΔTradeij,t) equalling imports plus exports, to sector output. Finally, the model includes an interaction term between concentration and changes in trade (inline image) to check whether the impact of trade is stronger in concentrated sectors than in competitive ones as indicated by the theoretical model. We also include time invariant skill/sector/country-specific effect (inline image), which can capture the difference between country and industry settings for a given skill category. Our model is dynamic including lagged wage-dependent variable allows for persistence in wage differentials. The estimated coefficient of the lagged wage differential is an indicator of the conditional convergence; if it is negative, a gender wage gap approaches its steady state, which is in turn determined by the trend, industry concentration and trade as in equation (15).9

A summary of statistics for all the variables is presented in Table A5 in the Appendix. Additionally, in Table A6, we show the pairwise correlation between independent variables.

b. Results

Due to the availability of trade data, we restrict our empirical analysis to only 12 manufacturing sectors.

It must be pointed out that some endogeneity problems may arise with our model. First, there can be endogeneity due to the inclusion of the lagged wage gap as an independent variable. In this case, either pooled ordinary least squares (OLS) or fixed effects will be biased.10 Additionally, we have to take into account the possibility of trade being influenced by the gender wage gap (not vice versa): for example, a high gender wage gap that is due to low female wages may induce ‘female-led’ exports (Rodrik, 2000; Busse and Spielmann, 2006). To ensure statistical accuracy, we use the generalised method of moments (GMM) technique from the framework developed by Arellano and Bond (1991), where the endogenous variable is instrumented by its lags.

An important feature of our analysis is the distinction between the wages of different skill groups of workers. Thus, we estimate equation (15) for the wages of high-skilled (wHS), medium-skilled (wMS) and low-skilled (wLS) workers in distinct sectors. The results are presented in Table 2. Trade variables are measured alternatively by import, export or aggregated trade (import plus export) penetration, and additionally, we simultaneously include import and export penetration in one regression to check the theoretical model, which implies that imports and export might have opposite effects.11

Table 2. The Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14)
 S = High S = Medium SkilledS = Low Skilled
 (1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1% level.

  5. (iv) Results are reported for two-step GMM (generalised method of moments) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano-Bond test are the p-values. The last two rows of the table report the p-values of the tests of coefficients’ equality.

  7. Source: Author’s own.

inline image 0.1572*** [0.0003] 0.1410*** [0.0006] 0.1421*** [0.0005] 0.1389*** [0.0007] 0.0406*** [0.0004] 0.0378*** [0.0004] 0.0368*** [0.0005] 0.0410*** [0.0007] 0.0427*** [0.0003] 0.0453*** [0.0004]−0.0447*** [0.0003]−0.0413*** [0.0004]
Trend −0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]
Conij,t 0.0326*** [0.0015]0.0264*** [0.0011]0.0292*** [0.0015]0.0286*** [0.0024]0.0071*** [0.0006]0.0075*** [0.0007]0.0042*** [0.0005]0.0048*** [0.0010]0.0176*** [0.0005]0.0138*** [0.0005]0.0150*** [0.0003]0.0164*** [0.0009]
ΔImpij,t 0.0501*** [0.0014]  0.0322*** [0.0020]−0.0154*** [0.0007]  0.0024 [0.0013]−0.0044*** [0.0006]  0.0070 [0.0022]
ConΔImpij,t −0.1178*** [0.0138]  −0.0639*** [0.0197]0.3496*** [0.0078]  0.1675*** [0.0120]0.1572*** [0.0070]  0.0730*** [0.0196]
ΔExpij,t  0.0830*** [0.0007] 0.0828*** [0.0028] −0.0413*** [0.0014] −0.0409*** [0.0013] −0.0177*** [0.0013] −0.0174*** [0.0028]
ConΔExpij,t  −0.3586*** [0.0099] −0.4366*** [0.0287] 0.4188*** [0.0134] 0.3407*** [0.0151] 0.1802*** [0.0122] 0.1103*** [0.0283]
ΔTradeij,t   0.0541*** [0.0011]   −0.0062*** [0.0005]   0.0025 [0.0004] 
ConΔTradeij,t   −0.2332*** [0.0130]   0.1611*** [0.0069]   0.0316*** [0.0042] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.0990.0950.0890.0910.1800.2070.2350.2210.2280.2200.2650.188
N 6,3576,4886,4636,2366,4226,5486,5286,2966,4226,5486,5286,296
ΔImpij,t = (−1) ΔExpij,t   0.000   0.000   0.000
ConΔImpj,t = (−1) ConΔExpij,t   0.000   0.000   0.000

In all specifications, the negative and statistically significant coefficient in front of lagged wage inequality is obtained in accordance with the conditional convergence hypothesis. Further, the decrease in the growth rate of male/female wage gap for all categories of skills (high, medium and low) is confirmed by the negative and statistically significant parameter trend. However, the magnitude of trend is not high, indicating that the drop in wage differentials is slow (as in Figure 1). In all models, we find a positive and statistically significant impact of sector concentration on the growth of the gender wage gap. The PCM coefficient is positive and statistical significant for all specifications, showing that a rise in the concentration of sectors (a drop in their competitiveness) is associated with higher growth in differentials between female and male wages.

The main coefficient of interest is the interaction term between concentration and the change in trade variables. For high-skilled workers, the estimated coefficient is negative – the growth of gender wage gap is reduced relatively more by trade in concentrated sectors than in competitive industries that were also affected by trade. The sign of the coefficient does not depend on the direction of trade – whether imports, exports or total trade are taken into consideration. In competitive sectors, the opposite result is obtained: increases in both import and export penetration and total trade penetration are associated with greater growth in the wage gap. These two effects may seem contradictory, but they are not and can be explained. The impact of trade openness on the gender wage gap for concentrated industries is negative, in accordance with the neo-classical tenet, and Becker’s (1971) economics of discrimination, that concentrated industries affected by trade lessen discrimination to maintain competitiveness, while the positive parameter for competitive industries is consistent with the concept that trade hurts skilled women and favours skilled men, for example, through masculinisation of technical jobs resulting in higher discrimination (Kongar, 2006). Further, the analogous results of a positive relationship between import competition and growth in the residual wage inequality in competitive industries were attributed to the unobservable skill differences between men and women (Black and Brainerd, 2004).

As far as medium- and low-skilled workers are concerned, we obtain the opposite results. We obtain a positive coefficient estimate on the interaction term for concentration and change in imports and exports (when the respective trade variables are estimated in separate regressions). This indicates that greater import and export penetration in more concentrated industries is associated with a higher growth of wage gap between men and women. This result can be due to the women’s relatively weak bargaining power, especially in low-skilled jobs, which prevent them from negotiation for higher wages when firms are forced to compete with foreign competition (Berik et al., 2004; Menon and Van Der Meulen Rodgers, 2009). Additionally, Kongar (2006) underlines that discriminatory employers will be penalised only if the labour market is tight (i.e. there is no unemployment). Further, the estimated coefficient on the change in trade variables is negative both for imports and exports (columns 5, 6 and 9, 10, respectively). This result shows that the rate of growth of gender wage gap decreased in trade-affected competitive industries, compared to industries that experienced lower levels of international trade. However, in the case of low-skilled industries, the coefficient is not statistically significant when total trade is employed (column 11).

When we include both imports and exports in one regression, the results indicate some loss of precision for the import coefficients (columns 4, 8 and 12). Where appropriate, we test the proposition that imports and exports have opposite effects. In all cases, we are able to reject with 99 per cent confidence the hypothesis that the estimated effects of imports and exports are equal but opposite in sign (see the last two rows of the Table 2).

Equation (15) may capture the effects of skill-biased technological change on the gender wage gap, rather than a change in discrimination due to trade openness. We augment regression equation (15) by including industry and time varying measure of the change in skill intensity inline image (defined as the share of hours worked by persons with higher education, both women and men, of the total hours worked).12Table 3 presents the results. For high- and medium-skilled workers (columns 1 and 2), the positive coefficient of skill intensity confirms that changes in employment that favour more educated workers support skilled male workers. The basic argument for technological change being expected to widen the gender wage gap is connected with the masculinisation of technical jobs and the potential exclusion of women from training for such technologically sophisticated jobs (Kongar, 2006). However, negative coefficients for low-skilled workers indicate that the shift in demand for high-skilled workers favours the low-skilled female workforce (e.g. as discrimination is lessened to maintain competitiveness). When the skill intensity variable is added, the trade parameters and interaction terms are still significant and of the same sign.

Table 3. The Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14), Additional Variable Change in the Skill Intensity Δ Skillij,t
 S = High SkilledS = Medium SkilledS = Low Skilled
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1% level.

  5. (iv) Results are reported for two-step GMM (generalised method of moments) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano-Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.1015*** [0.0006]−0.0882*** [0.0004]−0.0930*** [0.0006]−0.0821*** [0.0006]−0.0327*** [0.0003]−0.0296*** [0.0003]−0.0274*** [0.0005]−0.0332*** [0.0005]−0.0396*** [0.0002]−0.0400*** [0.0003]−0.0405*** [0.0003]−0.0368*** [0.0006]
Conij,t 0.0506*** [0.0024]0.0626*** [0.0019]0.0504*** [0.0028]0.0504*** [0.0027]0.0207*** [0.0008]0.0182*** [0.0012]0.0178*** [0.0007]0.0112*** [0.0017]0.0153*** [0.0007]0.0070*** [0.0008]0.0070*** [0.0010]0.0141*** [0.0008]
Trend −0.0006*** [0.0000]−0.0006*** [0.0000]−0.0006*** [0.0000]−0.0005*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]
ΔSkillij,t 0.0018*** [0.0000]0.0015*** [0.0000]0.0017*** [0.0000]0.0017*** [0.0000]0.0009*** [0.0000]0.0008*** [0.0000]0.0008*** [0.0000]0.0012*** [0.0000]−0.0018*** [0.0000]−0.0020*** [0.0000]−0.0019*** [0.0000]−0.0017*** [0.0000]
ΔImpij,t 0.0729*** [0.0007]  −0.0001 [0.0024]−0.0298*** [0.0006]  −0.0158*** [0.0020]−0.0107*** [0.0009]  −0.0006 [0.0023]
ConΔImpij,t −0.3571*** [0.0110]  0.2306 [0.0261]0.5545*** [0.0085]  0.3740*** [0.0249]0.2434*** [0.0109]  0.1377*** [0.0246]
ΔExpij,t  0.1449*** [0.0011] 0.1264*** [0.0027] −0.0463*** [0.0012] −0.0309*** [0.0021] −0.0188*** [0.0011] −0.0157*** [0.0033]
ConΔExpij,t  −0.9090*** [0.0141] −0.8301*** [0.0273] 0.5983*** [0.0109] 0.3051*** [0.0250] 0.2460*** [0.0115] 0.1412*** [0.0376]
ΔTradeij,t   0.0533*** [0.0008]   −0.0081*** [0.0005]   −0.0060*** [0.0005] 
ConΔTradeij,t   −0.2262*** [0.0098]   0.2229*** [0.0071]   0.1333*** [0.0037] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.1000.0960.0930.0960.1970.2150.2490.2450.2970.2940.3450.241
N 4,9354,9784,9494,9024,9434,9864,9574,9104,9434,9864,9574,910

Finally, we have to take into consideration the possibility that there may well be other factors that have an impact on the gender wage gap, but have been omitted from our analysis thus far. We argue that the trend variable picks up some of them, but labour market institutions can have an exceptional influence on gender wage discrimination (Oostendorp, 2009).

As a measure of the role of institutions in wage-setting mechanisms, we utilise indicators of employment protection from the OECD (ΔInstj,t), which, however, are sector-invariant.13 The indicators are on a scale from 0 (least restrictions) to 6 (most restrictions) and measure the procedures and costs involved in dismissing individuals or groups of workers. The estimations obtained via augmented regression are presented in Table 4. For high-skilled workers, an increase in the regulatory score raises the growth of the gender wage gap (coefficient not always significant), while for medium- and low-skilled workers, it lowers it. What is more important, after adding an institutional wage-setting variable, the trade coefficients and interaction terms remain similar.

Table 4. The Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14), Additional Variable Change in Wage Institution Setting ΔInstj,t
 S = High SkilledS = Medium SkilledS = Low Skilled
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1, and *10% levels.

  5. (iv) Results are reported for two-step generalized method of moments (GMM) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano-Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.0972*** [0.0005]−0.0849*** [0.0006]−0.0941*** [0.0010]−0.0725*** [0.0009]−0.0679*** [0.0010]−0.0636*** [0.0005]−0.0683*** [0.0006]−0.0569*** [0.0008]−0.0654*** [0.0004]−0.0693*** [0.0005]−0.0702*** [0.0005]−0.0606*** [0.0005]
Conij,t 0.0430*** [0.0024]0.0452*** [0.0021]0.0423*** [0.0026]0.0407*** [0.0022]0.0020* [0.0011]0.0045*** [0.0013]0.0042* [0.0026]0.0047* [0.0024]0.0110*** [0.0014]0.0062*** [0.0018]0.0096*** [0.0015]0.0076*** [0.0020]
Trend −0.0003*** [0.0000]−0.0002*** [0.0000]−0.0003*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]
ΔInstj,t 0.0001 [0.0001]0.0007* [0.0001]0.0008* [0.0002]0.0004* [0.0002]−0.0052*** [0.0002]−0.0038*** [0.0002]−0.0046*** [0.0003]−0.0036*** [0.0002]−0.0047*** [0.0001]−0.0032*** [0.0001]−0.0047*** [0.0002]−0.0037*** [0.0003]
ΔImpij,t 0.0868*** [0.0007]  0.0217*** [0.0028]−0.0534*** [0.0010]  −0.0145*** [0.0016]−0.0074*** [0.0009]  −0.0128*** [0.0022]
ConΔImpij,t −0.5549*** [0.0097]  −0.0478 [0.0331]0.6107*** [0.0130]  0.1901*** [0.0176]0.1623*** [0.0134]  0.2328*** [0.0285]
ΔExpij,t  0.1236*** [0.0008] 0.1062*** [0.0028] −0.0516*** [0.0010] −0.0419*** [0.0020] −0.0237*** [0.0016] −0.0041 [0.0026]
ConΔExpij,t  −0.8170*** [0.0120] −0.7122*** 0.6173*** [0.0124] 0.5026*** 0.2419*** [0.0149] −0.0143
ΔTradeij,t   0.0624*** [0.0011]   −0.0143*** [0.0007]   0.0012 [0.0011] 
ConΔTradeij,t   −0.3750*** [0.0118][0.0346]  0.2282*** [0.0070][0.0207]  0.0138 [0.0097][0.0334]
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.0970.0950.0930.0940.2840.2690.2490.2730.5660.5700.5590.581
N 3,6953,7013,6943,6943,6953,7013,6943,6943,6953,7013,6943,694

Finally, we include a measure of union density which, similar to the previous indicator of labour institutions, is sector-invariant and obtained from the OECD database.14 According to the literature, the union might not be gender neutral, with men being more highly unionised and unionisation rates higher in concentrated industries. In this case, a union acting in favour of its members should increase the gender wage gap, especially in concentrated industries (Black and Brainerd, 2004; Kongar, 2006). However, the results (Table 5) confirm this positive relation only for high-skilled workers, while for medium- and low-skilled ones, a rise in union density lowers the growth of the gender wage gap.

Table 5. The Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14), Additional Variable Change in Union Density Unionj,t
 S = High SkilledS = Medium SkilledS = Low Skilled
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at *** and **5% levels.

  5. (iv) Results are reported for two-step generalized method of moments (GMM) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano-Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.1492*** [0.0007]−0.1344*** [0.0004]−0.1373*** [0.0005]−0.1309*** [0.0005]−0.0347*** [0.0003]−0.0336*** [0.0003]−0.0304*** [0.0004]−0.0373*** [0.0005]−0.0375*** [0.0003]−0.0395*** [0.0002]−0.0392*** [0.0003]−0.0365*** [0.0003]
Conij,t 0.0291*** [0.0024]0.0343*** [0.0016]0.0251*** [0.0018]0.0251*** [0.0026]0.0114*** [0.0008]0.0105*** [0.0009]0.0095*** [0.0010]0.0033** [0.0014]0.0201*** [0.0008]0.0139*** [0.0004]0.0154*** [0.0006]0.0178*** [0.0008]
Trend −0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0007*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0003*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0002*** [0.0000]−0.0001*** [0.0000]
ΔUnionj,t 0.0009*** [0.0000]0.0006*** [0.0000]0.0007*** [0.0000]0.0008*** [0.0000]−0.0005*** [0.0000]−0.0005*** [0.0000]−0.0004*** [0.0000]−0.0006*** [0.0000]−0.0010*** [0.0000]−0.0012*** [0.0000]−0.0011*** [0.0000]−0.0008*** [0.0000]
ΔImpij,t 0.0737*** [0.0009]  0.0471*** [0.0048]−0.0188*** [0.0007]  −0.0012 [0.0013]0.012 [0.0006]  0.0052 [0.0016]
ConΔImpij,t −0.3419*** [0.0114]  −0.2044*** [0.0537]0.3553*** [0.0070]  0.1712*** [0.0143]0.0179** [0.0071]  0.1056*** [0.0161]
ΔExpij,t  0.0926*** [0.0010] 0.0745*** [0.0051] −0.0623*** [0.0010] −0.0588*** [0.0015] −0.0041*** [0.0013] −0.0074*** [0.0018]
ConΔExpij,t  −0.4041*** [0.0124] −0.3134*** [0.0577] 0.6677*** [0.0111] 0.5495*** [0.0153] 0.0533*** [0.0140] 0.0001 [0.0200]
ΔTradeij,t   0.0460*** [0.0016]   −0.0102*** [0.0006]   0.0060*** [0.0006] 
ConΔTradeij,t   −0.1320*** [0.0170]   0.1913*** [0.0057]   0.0030 [0.0066] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.1140.1100.1050.1050.2200.2510.2870.2720.2520.2570.2950.211
N 5,6755,7175,6945,6355,7375,7745,7565,6925,7375,7745,7565,692

The estimates with additional variables representing the wage-setting mechanisms confirm that labour institutions may be important determinants of the gender pay gap especially for low-skilled workers as shown by other researchers (Blau and Kahn, 2003; Bertola et al., 2007).

c. Robustness Checks

We perform a number of robustness checks to ensure the stability of our conclusions. To save space, the detailed results referred to in this section are available from the author upon request.

First, instead of imposing a common time trend for all countries, we check whether the results still hold if country-specific time trends are included. Alternatively, time dummies are incorporated instead of the time trend (compare the estimates presented in columns 1, 2 and 3 of Table A7 in the Appendix). The change in the measure of the time variable does not change our main conclusions. It is also interesting to notice that all the coefficients representing country-specific time trends are negative and highly statistically significant, indicating that in each of the countries covered by our sample, the growth in wage differentials decreased in the period analysed.15

One factor that may affect the results is the time period over which the regressions are estimated. We estimate the regression for different starting dates: 1970, 1980 and 1990. In Table A8 in the Appendix, we report results that replicate those of Table 2 to confirm the robustness of our findings for the 1980 to 2005 and 1990 to 2005 subsamples. While these shortened periods of time reduce the number of observations, our estimates are very similar to the previous ones. The only difference is the looseness of statistical significance for the interaction term in case of imports for the 1980 to 2005 subperiod.

Additionally, we check whether the estimations are sensitive to the inclusion of outliers.16 Again, the results concerning the impact of sector trade penetration and sectoral concentration remain stable with respect to the benchmark results.

Next, we perform an analogous analysis, further limiting the country sample, first by removing Japan and Korea and then by performing the estimates only with European countries (Table A9 in the Appendix). When the sample of countries is limited to only the European ones, for each of the skill categories the results hold.

In addition, we estimate the regression without the textiles, textile products, leather and footwear sector, checking whether our results are driven by the high percentage of female employment in this industry (see Table 1). According to Kucera and Milberg (2000), the gender bias in employment induced by trade in most cases disappears completely when this industry is excluded from the analysis. In our case, we did not find confirmation of their results. Therefore, we conclude that our results are not driven by the textiles, textile products, leather and footwear sector.

Finally, we consider an alternative measure of sector concentration. Following Aghion et al. (2008), we construct estimates of the PCM, this time as the difference between output and both wage and capital costs as a proportion of output. This alternative measure of mark-up confirms a positive relationship between the price-cost margin and the growth of the gender wage gap, and the trade coefficients remain comparable to those reported previously.

5. Conclusion

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix

This study contributes new evidence on the link between industry concentration, trade and gender wage gap. We overcome the limitations of previous studies (mainly of limited country and time coverage) by using an original and vast set of sectoral data from 18 countries for the period 1970 to 2005. We use industry statistics (EU KLEMS, 2008) on female and male wages at relatively disaggregated industry levels (NACE, 2 digit sectors) that distinguish between wages paid to different groups of workers classified according to skill level: high, medium and low. The conjunction of this data set with a consistent estimation methodology (GMM) enables the study to represent an important extension of the existing literature.

An important feature of our analysis is its distinction between the wage gaps of different skill groups of workers: thus, we separately estimate a dynamic model for the wages of high-skilled, medium-skilled and low-skilled workers. The difference-in-difference approach utilised makes it possible to compare changes in the skill-specific gender wage gap in concentrated versus competitive manufacturing industries, with the latter being used as a control for changes in the gender wage gap not related to international competitive pressure.

Overall, there are three main conclusions. First, an increase in trade penetration has different effects for concentrated versus competitive industries. Second, the impact is different for high-, medium- and low-skilled workers. Finally, there is no difference concerning the direction of trade: whether trade openness is measured by import or export trade penetration.

Our estimations indicate that, in general, the growth of gender wage gap decreases for all categories of skills and all countries during the period analysed. This is confirmed by the conditional convergence hypothesis and the negative trend observed. Further, as predicted by theory and the model proposed, the growth of the gender wage gap increased more in concentrated sectors (less competitive). As for the link between trade openness and wage discrimination, we find that rises in both import and export penetration are associated with a lower growth of the high-skilled gender wage gap in concentrated industries, but the opposite is true for medium- and low-skilled workers: trade raises the growth of the medium- and low-skilled sectoral gender wage gap for less-competitive trade-affected industries. For competitive industries, trade penetration raises the growth of high-skilled male/female wage differentials, while for medium- and low-skilled, more trade causes a decrease.

Additionally, we confirm the role of skill-biased technological change as a source of wage gap growth in manufacturing sectors. We also check the importance of wage-setting institutions, measured by employment protection and union density. Their impact is confirmed, but the impact depends on the specific skill category: for highly skilled labour an increase in labour market regulation and an increase in the unionisation rate raises the growth of the gender wage gap, while for medium- and low-skilled workers, it lowers it.

We address the robustness of our findings in several ways. We alter our measure of sector concentration and perform estimation on different time periods, different subsamples of countries and sectors. None of these changes influence the results in a considerable way, and the main conclusions hold.

The results have straightforward policy implications, especially with regard to the education and skill upgrading of female workers. Our study has attempted to assess trade-linked gender impacts from the perspective of the sectoral level. More research is required to examine additional important factors that can have an impact on the gender wage gap in the age of globalisation. For example, it would also be very interesting to confront patterns of male/female wage ratios with the intensity of outsourcing and/or trade in intermediate goods.

Footnotes
  • 1

    Most of the gender-related indices, such as the Global Gender Gap Index calculated by the World Economic Forum or the Gender-Related Development Index (GDI) developed by the United Nations, employ the female to male wage ratio in their estimates, among other criteria.

  • 2

    Equal pay for women for their work of equal value is the objective of the ILO conventions: Equal Remuneration Convention (No. 100) from 1951, The Discrimination (Employment and Occupation) Convention (No. 111) from 1958 and Workers with Family Responsibilities Convention, 1981 (No. 156).

  • 3

    Interest in the gender effects of trade policies is also expressed by a growing number of national and international organisations. For example, initiatives by The Gender Expert Group on Trade (GEGT) established by the UK’s Department of Trade and Industry, the projects by the Commonwealth Secretariat and the ILO. Specifically focusing on research concerning the nexus between economic openness and gender discrimination, The Gender and Trade Initiative (GATI) was launched by the Society for Conflict Analysis and Resolution (SOFCAR), a New Delhi based research and advocacy organisation (http://www.igtn.org/home/) and the International Gender and Trade Network (http://www.igtn.org/home/).

  • 4

    The residual wage gap (unexplained part of the gender wage gap) is commonly estimated by Blinder-Oaxaca decomposition (Blinder, 1973; Oaxaca, 1973). We thank an anonymous referee for pointing this out.

  • 5

    We thank an anonymous referee for pointing this out.

  • 6

    For detailed steps to obtain equation (6) from equation (3), see Borjas and Ramey (1995).

  • 7

    We use the EU KLEMS March 2008 release, since the November 2009 release does not provide labour input files. Detailed information on sources and methods can be found in Timmer et al. (2007). The database is available at http://www.euklems.net. The database has been so far used to analyse skilled-unskilled patterns of wages in the EU (e.g. Lo Turco and Parteka 2011; Parteka 2011), but to the best of our knowledge, this database has not been previously used in analysis of gender wage differentials.

  • 8

    Alternatively, the model includes country-specific time trends or time dummies – see the robustness section.

  • 9

    The concept is similar to conditional beta convergence from growth theories like Barro and Sala-i-Martin (1992). Note that the estimated coefficient of the lagged gender wage gap indicates conditional not absolute convergence, and the gender wage gap might actually be increasing if the steady state gender wage gap is growing. We thank an anonymous referee for pointing this out.

  • 10

    For a formal exposition of bias, see for example, Arellano and Bond (1991) or Bond (2002).

  • 11

    We are aware of the multicollinearity problem that can arise due to the inclusion of both imports and exports in one regression (indeed, the pairwise correlation between export and import penetration equals 0.6, as shown in Table A6 in the Appendix). However, inclusion of both variables is supported by the fact that the theoretical model otherwise creates a misspecification bias. To combine the theoretical and econometrical requirements, we present four different specifications: imports, exports, total trade, and both imports and exports. We thank an anonymous referee for pointing this out.

  • 12

    Due to data unavailability, the calculation of skill intensity is not possible for all the years and countries in our sample. The estimations involving skill intensity are performed for restricted sets of countries and time covered. For AUS, AUT, BEL, DNK, ESP, FIN, NLD, the data start at 1980, for GER at 1991 and for CZE, HUN, POL, SVK, SVN at 1995.

  • 13

    Again, due to restricted data availability on employment protection index, the estimations were performed on the sub-sample of observations. For AUS, AUT, BEL, DNK, ESP, FIN, GER, ITA, JPN, NLD, UK, the data start at 1985, for KOR at 1990, for CZE, HUN, POL, SVK, USA at 1993, no data for SVN.

  • 14

    The union density data for ESP starts at 1981, for POL at 1990, and for CZE, HUN, SVK at 1995, no data for SVN.

  • 15

    Due to space constraints, the estimates of the coefficients for the country-specific time trends are not presented here, but are available from the author upon request as well as results for the low- and medium-skilled workforce.

  • 16

    Outliers were detected using the method of Hadi (1994), which identifies multiple outliers in a multivariate data sample.

References

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix
  • Aghion, P., M. Brauny and J. W. Fedderkez (2008), ‘Competition and Productivity Growth in South Africa’, Economics of Transition, 16, 4 74168.
  • Anderson, E. (2005), ‘Openness and Inequality in Developing Countries: A Review of Theory and Recent Evidence’, World Development, 33, 7, 104563.
  • Antonczyk, D., B. Fitzenberger and K. Sommerfeld (2010), ‘Rising Wage Inequality, the Decline of Collective Bargaining, and the Gender Wage Gap’, Labour Economics, 17, 5, 83547.
  • Arellano, M. and S. R. Bond (1991), ‘Some Tests of Specification for Panel Data: Monte Carlo Evidence and an Application to Employment Equations’, Review of Economic Studies, 58, 2, 27797.
  • Artecona, R. and W. Cunningham (2002), ‘Effects of Trade Liberalization on the Gender Wage Gap in Mexico’, Policy Research Report on Gender and Development Working Paper No. 21 (Washington, DC: World Bank).
  • Arulampalam, W., A. L. Booth and M. L. Bryan (2007), ‘Is There a Glass Ceiling over Europe? Exploring the Gender Pay Gap Across the Wages Distribution’, Industrial and Labor Relations Review, 60, 2, 16386.
  • Bardasi, E. and J. C. Gornick (2008), ‘Working for Less? Women’s Part-time Wage Penalties Across Countries’, Feminist Economics, 14, 1, 3772.
  • Barro, R. J. and X. Sala-i-Martin (1992), ‘Convergence’, Journal of Political Economy, 100, 2, 22351.
  • Becker, G. S. (1971), The Economics of Discrimination (Chicago, IL: University of Chicago Press).
  • Berik, G., Y. Van Der Meulen Rodgers and J. Zveglich (2004), ‘International Trade and Gender Wage Discrimination: Evidence from East Asia’, Review of Development Economics, 8, 2, 23754.
  • Bertola, G., F. Blau and L. Kahn (2007), ‘Labor Market Institutions and Demographic Employment Patterns’, Journal of Population Economics, 20, 4, 83367.
  • Binder, M., K. Krause, J. Chermak, J. Thacher and J. Gilroy (2010), ‘Same Work, Different Pay? Evidence from a US Public University’, Feminist Economics, 16, 4, 10535.
  • Black, S. and E. Brainerd (2004), ‘Importing Equality? The Impact of Globalization on Gender Discrimination’, Industrial and Labor Relations Review, 57, 4, 54059.
  • Blau, F. D. and M. Kahn (2003), ‘Understanding International Differences in the Gender Pay Gap’, Journal of Labor Economics, 21, 1, 10644.
  • Blinder, A. S. (1973), ‘Wage Discrimination: Reduced Form and Structural Variables’, Journal of Human Resources, 8, 43655.
  • Bond, S. (2002), ‘Dynamic Panel Data Models: A Guide to Micro Data Methods and Practice’, Cemmap Working Paper CWP09/02 (London: The Institute for Fiscal Studies, Department of Economics, UCL).
  • Borjas, G. and V. Ramey (1995), ‘Foreign Competition, Market Power, and Wage Inequality’, Quarterly Journal of Economics, 110, 4, 1075110.
  • Busse, M. and C. H. Spielmann (2006), ‘Gender Inequality and Trade’, Review of International Economics, 14, 3, 36279.
  • Chamarbagwala, R. (2006), ‘Economic Liberalization and Wage Inequality in India’, World Development, 34, 12, 19972015.
  • Dominguez-Villalobos, L. and F. Brown-Grossman (2010), ‘Trade Liberalization and Gender Wage Inequality in Mexico’, Feminist Economics, 16, 4, 5379.
  • EU KLEMS (2008), ‘EU KLEMS Growth and Productivity Accounts’, March. Available at: http://www.euklems.net (accessed 6 June 2011).
  • Fontana, M. (2009), ‘The Gender Effects of Trade Liberalization in Developing Countries: A Review of the Literature’, in M. Bussolo and R. De Hoyos (eds.), Gender Aspects of the Trade and Poverty Nexus: A Micro-macro Approach (Basingstoke: Palgrave Macmillan), 2550.
  • Gradin, C., C. Del Rio and O. Canto (2010), ‘Gender Wage Discrimination and Poverty in the EU’, Feminist Economics, 16, 2, 73109.
  • Gupta, N. D. (2002), ‘Gender, Pay and Development: A Cross-country Analysis’, Labour and Management in Development, 3, 2, 119.
  • Hadi, A. S. (1994), ‘A Modification of a Method for the Detection of Outliers in Multivariate Samples’, Journal of the Royal Statistical Society, 56, 3936.
  • Hazarika, G. and R. Otero (2004), ‘Foreign Trade and the Gender Earnings Differential in Urban Mexico’, Journal of Economic Integration, 19, 2, 35373.
  • ILO (2003), Time for Equality at Work. International Labour Conference 91st Session 2003, Report I (B) (Geneva: International Labour Office).
  • Kongar, E. (2006), ‘Importing Equality or Exporting Jobs?: Competition and Gender Wage and Employment Differentials in U.S. Manufacturing’, Working Paper No. 436 (New York: The Levy Economics Institute of Bard College).
  • Kucera, D. and W. Milberg (2000), ‘Gender Segregation and Gender Bias in Manufacturing Trade Expansion: Revisiting the “Wood Asymmetry”’, World Development, 28, 7, 1191210.
  • Lo Turco, A. and A. Parteka (2011), ‘The Demand for Skills and Labour Costs in Partner Countries: Evidence from the Enlarged EU’, Economics of Transition, 19, 3, 61137.
  • Menon, N. and Y. Van Der Meulen Rodgers (2009), ‘International Trade and the Gender Wage Gap: New Evidence from India’s Manufacturing Sector’, World Development, 37, 5, 96581.
  • Neumayer, A. and I. De Soysa (2007), Globalisation, Women’s Economic Rights and Forced Labour’, The World Economy, 30, 1, 151035.
  • Oaxaca, R. (1973), Male-Female Wage Differentials in Urban Labor Markets’, International Economic Review, 14, 693709.
  • Oostendorp, R. H. (2009), ‘Globalization and the Gender Wage Gap’, World Bank Economic Review, 23, 1, 14161.
  • Parteka, A. (2011), ‘Skilled-unskilled Wage Gap and Returns to Education in the Enlarged EU – Sectoral Analysis Based on the Experience of EU-15 Countries and Five New Member States’, Available at: http://iweb.cerge-ei.cz/pdf/gdn/RRCX_66_paper_01.pdf accessed 16 July 2011.
  • Rodrik, D. (2000), ‘Institutions for High-quality Growth: What They are and How to Acquire Them’, Studies in Comparative International Development, 35, 3, 331.
  • Schober, T. and R. Winter-Ebmer (2011), ‘Gender Wage Inequality and Economic Growth: Is There Really a Puzzle?–A Comment’, World Development, 39, 8, 147684.
  • Seguino, S. (2011), ‘Gender Inequality and Economic Growth: A Reply to Schober and Winter-Ebmer’, World Development, 39, 8, 14857.
  • Sen, A. (2004), ‘How to Judge Globalism?’, in F. J. Lechner and J. Boli (eds.), The Globalisation Reader (Oxford: Blackwell Publishing Ltd.), 1622.
  • Shen, J. and X. Deng (2008), ‘Gender Wage Inequality in the Transitional Chinese Economy: A Critical Review of Post-reform Research’, Journal of Organisational Transformation and Social Change, 5, 10927.
  • Timmer, M., M. O’Mahony and B. Van Ark (2007), The EU KLEMS Growth and Productivity Accounts: An Overview (Birmingham, AL: accessed 16 July 2011 University of Birmingham).
  • Van Staveren, I., D. Elson, C. Grown and N. Çağatay (eds.) (2007). Feminist Economics of Trade (London: Routledge).
  • Weichselbaumer, D. and R. Winter-Ebmer (2005), ‘A Meta-Analysis of the International Gender Wage Gap’, Journal of Economic Surveys, 19, 3, 479511.
  • Weichselbaumer, D., R. Winter-Ebmer and M. Zweimüller (2007), ‘Market Orientation and Gender Wage Gaps: An International Study’, Economics Working Papers 2007–12, (Linz: Johannes Kepler University).
  • Williams, R. (1987), ‘Capital, Competition, and Discrimination: A Reconsideration of Racial Earnings Inequality’, Review of Radical Political Economics, 19, 2, 115.
  • Wood, A. (1991), ‘North-South Trade and Female Labour in Manufacturing: An Asymmetry’, Journal of Development Studies, 27, 2, 16889.
  • Wood, A. (1995), ‘How Trade Hurt Unskilled Workers’, The Journal of Economic Perspectives, 9, 3, 5780.
  • World Economic Forum (2010), The Global Gender Gap Report (Geneva: World Economic Forum).

Appendix

  1. Top of page
  2. Abstract
  3. 1. Introduction
  4. 2. Theoretical Model
  5. 3. Empirical Setting
  6. 4. Empirical Analyses of the Trade – Gender Wage Gap Nexus
  7. 5. Conclusion
  8. References
  9. Appendix
Table A1. List of Sectors
NACEISIC Rev.3Description of Sectors
  1. Source: Own based on the correspondence table from United Nations Statistics Division.

AtB0105C01T05 Agriculture, Hunting, Forestry and Fishing
C1014C10T14 Mining and Quarrying
D1537C15T37 Manufacturing
15t161516C15T16 Food products, beverages and tobacco
17t191719C17T19 Textiles, textile products, leather and footwear
202000C20 Wood and products of wood and cork
21t222122C21T22 Pulp, paper, paper products, printing and publishing
242400C24 Chemicals and chemical products
252500C25 Rubber and plastics products
262600C26 Other non-metallic mineral products
27t282728C27T28 Basic metals and fabricated metal products
292900C29 Machinery and equipment, n.e.c.
30t333033C30T33 Electrical and optical equipment
34t353435C34T35 Transport equipment
36t373637C36T37 Manufacturing n.e.c. and recycling
E4041C40T41 Electricity gas and, water supply
F4500C45 Construction
G5055C50T55 Wholesale and retail trade – restaurants and hotels
505000C50 Sale, maintenance and repair of motor vehicles and motorcycles – retail sale of automotive fuel
515100C51 Wholesale, trade and commission excl. motor vehicles
525200C52 Retail trade excl. motor vehicles – repair of household goods
I6064C60T64 Transport, storage and communications
60t636063C60T63 Transport and storage
646400C64 Post and telecommunications
J6567C65T67 Financial intermediation
K7074C70T74 Real estate, renting and business activities
707000C70 Real estate activities
71t747174C71T74 Renting of mach. and equip. – other business activities
L7500C75 Public admin. and defence – compulsory social security
M8000C80 Education
N8500C85 Health and social work
Tot0199CTotal total
Table A2. List of Countries
Lp.Country ISOCountry Name
  1. Source: Own.

1AUSAustralia
2AUTAustria
3BELBelgium
4CZECzech Republic
5DNKDenmark
6ESPSpain
7FINFinland
8GERGermany
9HUNHungary
10ITAItaly
11JPNJapan
12KORKorea
13NLDNetherlands
14POLPoland
15SVKSlovakia
16SVNSlovenia
17UKUnited Kingdom
18USAUnited States of America
Table A3. Share of Hours Worked by Women by Academic Activity in 2005 (in %)
CountryAtBCDEFGIJKLMN
  1. Source: Own calculations based on data from EU KLEMS (2008).

AUS292525251142424444686868
AUT481526151148224538316668
BEL30192319748244941506875
CZE31173817853296344457680
DNK19133122840265142526083
ESP27112520547224550386574
FIN28213021857266749556489
GER272628261249295652586478
HUN25253925654256845507878
ITA43123120739224935487560
JPN361426151138134426204877
KOR17827141432252728606060
NLD19211821738193835334871
POL4317na177nana68na497881
SVK2516na167nana65na507782
SVN46183818756246345497684
UK252228221051245242487279
USA17112724836255841366575
Mean30172920945255341476776
Min1781814532132726204860
Max482639261457426852687889
Table A4. Female and Male Shares in Employment of High-skilled, Low-skilled and Medium-skilled Workforce, by Economic Activity
Sectors19702005
High SkilledLow SkilledMedium SkilledHigh SkilledLow SkilledMedium Skilled
FemaleMaleFemaleMaleFemaleMaleFemaleMaleFemaleMaleFemaleMale
  1. Source: Own calculations based on data from EU KLEMS (2008).

15t160.532.6636.3722.2412.5825.643.195.9111.3113.0227.5838.99
17t190.481.9341.3219.3020.8616.113.235.2212.9212.1531.5934.89
200.703.2117.0337.747.8033.514.468.375.2315.8914.6351.66
21t221.024.1817.7432.0211.7933.245.179.845.2613.3917.3948.96
240.947.7418.3129.5610.6532.815.7811.775.0712.8016.9447.63
250.664.2119.6730.4111.2433.814.079.235.8713.9917.8049.04
260.584.1613.4239.088.4034.363.918.825.1315.5815.3651.20
27t280.604.4613.0637.147.3737.373.858.795.0914.8914.8552.54
290.414.4117.4228.859.0739.842.5911.685.1511.6815.0353.88
30t330.505.9119.1725.3613.9735.103.2113.615.919.7317.2050.34
34t350.414.7220.5228.817.1538.382.5011.845.2011.8013.5955.06
36t370.453.9428.1828.3912.2526.782.857.678.1714.1823.2943.84
501.006.3517.4924.6810.5339.964.387.328.3110.0629.1840.76
511.128.0418.0422.6015.7534.465.029.578.298.8031.1337.18
521.346.8820.1722.7221.1627.735.027.719.599.0434.4334.20
60t630.603.148.7941.606.0839.783.236.284.1516.5914.0655.70
640.783.5315.3928.0117.4434.866.0711.646.1810.9025.4839.74
702.5814.3015.8913.0223.8730.3410.6220.457.387.0324.4630.06
71t742.8218.5815.8015.3321.1126.3611.9023.987.266.1021.6429.11
AtB0.654.8511.3549.426.7227.002.197.3410.5222.2317.0540.67
C0.695.436.0345.566.9235.363.3110.802.8115.1811.2056.71
D0.604.0124.5827.8412.3530.623.639.977.2114.0917.9047.20
E0.7410.425.6029.318.8345.104.3614.252.729.7512.7356.19
F0.297.864.3338.843.2345.451.548.801.4819.295.5963.29
G1.237.1419.1922.9518.1631.314.908.489.6710.0030.9236.03
I0.643.2410.3338.288.9338.583.897.445.0416.8615.8250.95
J3.7113.9212.956.9528.4733.9914.5719.374.453.3833.8124.44
K2.8418.1915.8414.3621.9826.7911.4123.587.946.8321.6728.58
L5.6617.398.6013.9114.2740.1715.3016.774.544.6926.7931.91
M20.8124.4214.405.2624.6310.4733.7119.665.161.8528.1611.45
N10.3917.6616.445.3336.0714.1018.0311.208.152.0949.4211.11
Tot2.297.4315.9629.9314.3030.108.4212.117.6410.7225.2635.85
Table A5. Summary Statistics
VariableObsMeanSDMinMax
  1. Source: Author’s own.

inline image 7495−0.0010.048−0.4390.963
inline image 7560−0.0030.026−0.3660.311
inline image 7560−0.0030.031−0.4310.366
inline image 77110.3860.201−1.8931.086
inline image 77760.3230.210−0.9991.002
inline image 77760.3450.214−0.3101.359
Conij,t 69130.1120.0740.0000.599
Impij,t 70320.4320.4230.0003.927
Expij,t 74180.3900.3680.0003.958
Tradeij,t 73020.7720.6510.0004.964
Skillij,t 52928.9397.1540.19245.297
Instj,t 39121.9570.9050.2103.800
Unionj,t 610837.98818.8448.34280.766
Table A6. Pairwise Correlation Between Independent Variables
  Impij,t Expij,t Tradeij,t Conij,t Skillij,t Instj,t Unionj,t
  1. Source: Author’s own.

Impij,t 1.00      
Expij,t 0.641.00     
Tradeij,t 0.890.821.00    
Conij,t −0.05−0.08−0.141.00   
Skillij,t −0.12−0.12−0.130.241.00  
Instj,t 0.030.170.10−0.10−0.251.00 
Unionj,t 0.190.230.23−0.10−0.170.111.00
Table A7. Robustness Check – the Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14), for High-skilled Workforce, Different Measures of Time Variable
 Common TrendCountry-specific TrendsTime Dummies
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1% level.

  5. (iv) Results are reported for two-step generalized method of moments (GMM) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano–Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.1572*** [0.0003]−0.1410*** [0.0006]−0.1421*** [0.0005]−0.1389*** [0.0007]−0.1939*** [0.0011]−0.1756*** [0.0012]−0.1762*** [0.0011]−0.1864*** [0.0010]−0.1544*** [0.0011]−0.1358*** [0.0009]−0.1402*** [0.0019]−0.1348*** [0.0026]
Trend−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]        
Conij,t 0.0326*** [0.0015]0.0264*** [0.0011]0.0292*** [0.0015]0.0286*** [0.0024]0.0067*** [0.0025]0.0040 [0.0028]0.0074*** [0.0027]0.0062 [0.0040]0.0333*** [0.0047]0.0262*** [0.0036]0.0291*** [0.0037]0.0275*** [0.0058]
ΔImpij,t 0.0501*** [0.0014]  0.0322*** [0.0020]0.0744*** [0.0026]  0.0378*** [0.0044]0.0542*** [0.0045]  0.0415*** [0.0071]
ConΔImpij,t −0.1178*** [0.0138]  −0.0639*** [0.0197]−0.4070*** [0.0253]  −0.1895*** [0.0446]−0.1572*** [0.0411]  −0.1420* [0.0749]
ΔExpij,t  0.0830*** [0.0007] 0.0828*** [0.0028] 0.1240*** [0.0032] 0.1222*** [0.0043] 0.0897*** [0.0049] 0.0734*** [0.0087]
ConΔExpij,t  −0.3586*** [0.0099] −0.4366*** [0.0287] −0.7797*** [0.0357] −0.8277*** −0.4583*** [0.0563] −0.3546*** [0.0918]
ΔTradeij,t   0.0541*** [0.0011]   0.0616*** [0.0021]   0.0595*** [0.0042] 
ConΔTradeij,t   −0.2332*** [0.0130]   −0.3410*** [0.0256]   −0.2809*** [0.0423] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.0990.0950.0890.0910.1000.0960.0910.0930.0990.0940.0880.090
N 6,3576,4886,4636,2366,3576,4886,4636,2366,3576,4886,4636,236
Table A8. Robustness Check – the Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14) for High-skilled Workforce, Initial Year: 1970, 1980 and 1990
 1970 – Initial Year1980 – Initial Year1990 – Initial Year
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1% level.

  5. (iv) Results are reported for two-step GMM (generalised method of moments) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano–Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.1572*** [0.0003]−0.1410*** [0.0006]−0.1421*** [0.0005]−0.1389*** [0.0007]−0.1157*** [0.0007]−0.1045*** [0.0005]−0.1069*** [0.0008]−0.0887*** [0.0010]−0.1061*** [0.0009]−0.0901*** [0.0007]−0.1011*** [0.0014]−0.0819*** [0.0013]
Trend −0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0008*** [0.0000]−0.0006*** [0.0000]−0.0005*** [0.0000]−0.0006*** [0.0000]−0.0005*** [0.0000]−0.0008*** [0.0000]−0.0007*** [0.0000]−0.0007*** [0.0000]−0.0007*** [0.0000]
Conij,t 0.0326*** [0.0015]0.0264*** [0.0011]0.0292*** [0.0015]0.0286*** [0.0024]0.0188*** [0.0014]0.0160*** [0.0013]0.0232*** [0.0013]0.0159*** [0.0018]0.0486*** [0.0023]0.0448*** [0.0021]0.0512*** [0.0027]0.0429*** [0.0028]
ΔImpij,t 0.0501*** [0.0014]  0.0322*** [0.0020]0.0327*** [0.0009]  −0.0017 [0.0023]0.0989*** [0.0007]  0.0149*** [0.0025]
ConΔImpij,t −0.1178*** [0.0138]  −0.0639*** [0.0197]0.0399 [0.0113]  0.2174 [0.0254]−0.5956*** [0.0091]  0.0917 [0.0253]
ΔExpij,t  0.0830*** [0.0007] 0.0828*** [0.0028] 0.1192*** [0.0009] 0.1207*** [0.0035] 0.1442*** [0.0013] 0.1348*** [0.0017]
ConΔExpij,t  −0.3586*** [0.0099] −0.4366*** [0.0287] −0.6916*** [0.0157] −0.7938*** [0.0405] −1.0792*** [0.0197] −1.0789*** [0.0216]
ΔTradeij,t   0.0541*** [0.0011]   0.0539*** [0.0017]   0.0875*** [0.0015] 
ConΔTradeij,t   −0.2332*** [0.0130]   −0.2428*** [0.0191]   −0.6149*** [0.0179] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.0990.0950.0890.0910.1110.1050.1020.1050.0900.0890.0860.090
N 6,3576,4886,4636,2365,0215,1375,0984,9773,3743,3963,3853,370
Table A9. Robustness Check – the Impact of Trade and Concentration on Gender Wage Gap: Dependent Variable Δ Gender Wage Gap as in Equation (14)– European Countries
 S = High SkilledS = Medium SkilledS = Low Skilled
(1)(2)(3)(4)(5)(6)(7)(8)(9)(10)(11)(12)
  1. Notes:

  2. (i) All computations made using XTABOND2 for StataSE 9.0.

  3. (ii) Constant not reported.

  4. (iii) Standard errors in parentheses. Statistically significant at ***1% level.

  5. (iv) Results are reported for two-step GMM (generalised method of moments) estimator, the lagged value of wage gap inline image and trade variables: ΔImpij,t, ΔExpij,t and ΔTradeij,t treated as endogenous and instrumented by their lags, country-sector effects included, clustered standard errors.

  6. (v) The figures reported for Arellano–Bond test are the p-values.

  7. Source: Author’s own.

inline image −0.1256*** [0.0005]−0.1119*** [0.0005]−0.1143*** [0.0007]−0.1149*** [0.0009]−0.0322*** [0.0003]−0.0291*** [0.0003]−0.0315*** [0.0005]−0.0283*** [0.0005]−0.0405*** [0.0002]−0.0403*** [0.0001]−0.0407*** [0.0002]−0.0390*** [0.0005]
Conij,t 0.0169*** [0.0009]0.0130*** [0.0013]0.0177*** [0.0014]0.0175*** [0.0021]0.0143*** [0.0007]0.0135*** [0.0003]0.0105*** [0.0007]0.0139*** [0.0008]0.0109*** [0.0008]0.0101*** [0.0008]0.0115*** [0.0006]0.0075*** [0.0010]
Trend −0.0007*** [0.0000]−0.0007*** [0.0000]−0.0007*** [0.0000]−0.0007*** [0.0000]−0.0002*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]−0.0001*** [0.0000]
ΔImpij,t 0.0835*** [0.0008]  0.0718*** [0.0018]−0.0182*** [0.0008]  −0.0052*** [0.0011]−0.0050*** [0.0006]  0.0079 [0.0020]
ConΔImpij,t −0.3079*** [0.0110]  −0.3291*** [0.0211]0.4231*** [0.0122]  0.3158*** [0.0166]0.1895*** [0.0059]  0.1058*** [0.0198]
ΔExpij,t  0.0856*** [0.0013] 0.0534*** [0.0025] −0.0393*** [0.0012] −0.0249*** [0.0018] −0.0162*** [0.0011] −0.0186*** [0.0011]
ConΔExpij,t  −0.3225*** [0.0177] −0.1755*** [0.0334] 0.3782*** [0.0125] 0.1201*** [0.0175] 0.1395*** [0.0121] 0.0884*** [0.0161]
ΔTradeij,t   0.0539*** [0.0014]   −0.0099*** [0.0008]   0.0059 [0.0005] 
ConΔTradeij,t   −0.1786*** [0.0156]   0.2168*** [0.0098]   0.0185*** [0.0054] 
ar2p0.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.0000.000
ar2p0.1820.1840.1760.1780.3600.4210.3560.3970.6210.6370.6220.629
N 4,8024,9014,8924,7134,8674,9614,9574,7734,8674,9614,9574,773