An Inter-temporal Analysis of Gender Wage Differentials and Discrimination in Germany and the UK. By Michael Brookes,* Middlesex University Business School. Abstract Gender wage differentials and discrimination have been widely recorded, unfortunately there have been few attempts to analyse them over time and across countries. This paper contributes to the literature by providing a cross-country analysis of the UK and Germany, as well as identifying the causes of change over time. Using decomposition analysis of earnings functions the results indicate that in 1991 the gender wage differential was wider in the UK than in Germany, however by 1996/1997 this position had reversed. The majority of this reversal being caused by a more favourable employment mix for UK women within the public sector and large firms. JEL: J16, J31, J71. Keywords: Gender, Wage Differentials, Wage Discrimination. *Centre for Applied Research in Economics, Middlesex University Business School, London NW4 4BT email: m.brookes@mdx.ac.uk 1
1. Introduction In situations where gender wage differentials are caused by unequal opportunities resulting from discrimination, policy makers should be able to improve efficiency by implementing appropriate labour market policies. However if the differentials actually result from differing individual choices, any government intervention is likely to distort the market and increase the level of inefficiency. Consequently a clear understanding of gender wage differentials and discrimination is essential, especially for policy makers, who need to be able to establish the true cause of any differential. To date the most usual approach to this problem is to decompose the gender wage differential into its 'explained' and 'unexplained' portions using the method pioneered by Oaxaca (1973). This indicates the upper bound of any gender wage discrimination. There are numerous examples within the literature where this type of analysis has been carried out, however there are two clear areas where our understanding of this issue can be improved. Firstly there are very few studies that analyse changes in the differential over time, Greenhalgh (1980) and Harkness (1996), have attempted this for the UK. Secondly there are similarly few examples of cross-country comparisons being made, Blau & Kahn (1992) and Callan et al (1996) being two of the more comprehensive studies. Consequently there is lots of evidence relating to the size of the gender wage differential and the level of discrimination in specific countries at particular moments in time, the next section summarises the existing relevant evidence. However there is little understanding of how the differentials change over time and how different countries compare with one another. 2
All of the previous attempts at either cross-country or inter-temporal analysis have been flawed because of problems with differing variable definitions, or having to use different data sets for each country or time period. This study seeks to overcome these problems by using the Panel Comparability Project (PACO) data set as it provides harmonised and standardised data from existing single country longitudinal studies. This will allow for more meaningful comparisons to be made across countries, and time, since the problems outlined above have been overcome. Germany and the UK are the chosen countries since of the non-transitional economies in the project they have the most recent overlap of years. 2. Analysing Discrimination Table 1, below, summarises the findings from most of the relevant studies, the previous paragraphs indicate why it is difficult to make comparisons both across countries and across time. However it is fair to conclude that in most cases the gender wage differential is narrower in West Germany, and it has continued to be so in the unified Germany. The most likely causes of the narrower differential in Germany are differences in wage inequality, collective bargaining and part-time employment. Blau & Kahn (1992) identify that wage inequality plays a key role in determining the size of the gender wage differential. Greater levels of inequality lead to a larger wage gap because the higher returns from education and experience disproportionately favour the men. Gottschalk & Smeeding 1997 reveal that inequality has consistently 3
Table 1 Author(s) Dataset F/M Ratio Earnings Unexplained Diff.* UK Estimates Chiplin & Sloane 1976 Firm Survey 1976 0.86 Annual 62% Greenhalgh 1980 GHS 1971 0.85 Hourly, singles 24% GHS 1975 0.97 10% Zabalza & Tzannatos 1985 GHS 1975 0.62 Hourly 18% Miller 1987 GHS 1980 0.61 Hourly 38% Wright & Ermisch 1991 WES 1980 0.67 Hourly, marrieds Approx. 50% Harkness 1996 GHS 1974 0.66 Hourly, full-timers 86% BHPS 1992-93 0.8 92% Joshi & Paci 1998 NCDS 1991 0.71 Hourly 60% German Estimates Gerlach 1987 Postal Questionnaire 0.68 Hourly, marrieds Approx 85% 0.89 Hourly, singles Approx 85% Hubler 1991 Pooled ISSP 1984-86 0.72 Hourly, full-timers 38% Cross-country Studies Gunderson 1989 1960 Germ 0.65 Germ. all workers No Estimates UK 0.61 UK man. workers 1980 Germ 0.72 UK 0.79 Blau & Kahn 1992 Pooled ISSP 1985-89 Germ 0.69 Hourly, all workers No Estimates UK 0.63 Blanchflower Pooled ISSP 1985-87 No Hourly, all workers Germany 46% & Oswald 1989 Estimates UK 56% Callan et al 1996 LIS 1991 Germ 0.75 Hourly, manufacturing UK 0.71 No Estimates Black et al 1999 ISSP 1989 Germ 0.72 Hourly, marrieds Germany 95% UK 0.57 UK 80% Brookes et al 2001 PACO 1991-93 Germ 0.79 Hourly, all workers Germany 90% UK 0.75 UK 96% 4
been higher in the UK, which has then obviously impacted upon the relative size of the wage gap in the two countries. For similar reasons the level of coverage of collective bargaining will have an impact upon the gender differential, since the greater the extent of collective bargaining the less inequality there tends to be. Consequently, since Traxler 1996, puts the level of coverage of collective bargaining at 80% in Germany and 35% in the UK, the narrower German gap is to be expected. Finally, Ermisch & Wright (1993) indicate that for part-time workers their supply of labour function is less elastic, therefore at best their hourly wage will be equal to full-timers, but in most cases lower. In the UK 44% of all women employees are part-time, (OECD 1992), compared to 34% in Germany (Employment in Europe 1992), so once again the relative gender differentials are as expected. In terms of discrimination the relative shortage of cross-country comparisons make it difficult to identify the true position. However Black et al, (1999), suggest that the level of discrimination was lower in the UK prior to 1980 and the situation reversed after that date, although the reasons for this reversal are unclear. The most recent estimate of discrimination by Brookes et al (2001) supports this, with a slightly smaller unexplained differential in Germany. More interestingly the studies by Black et al (1999) and Brookes et al (2001) reveal that in both countries, as the gender wage differential has gradually narrowed over time, the proportion of this gap which may be attributed to discrimination has increased. 3. Methodology 5
The purpose of this paper is to establish 3 things, firstly the size of the gender wage differential and the extent of discrimination in each of the two countries, secondly how the differentials compare across the two countries and finally how the relative position of women has changed over time. In order to achieve the first objective earnings functions are estimated for men and women separately in each country using the most recent available data, 1997 for the UK and 1996 for Germany. The earnings functions are of the form; ' lnw = Z β + u, where lnw i is the natural log of the hourly wage, Z i is i i i the endowment of productive characteristics, β is the vector of estimated coefficients and u i the normally distributed error term. Following the method pioneered by Oaxaca (1973) the gender wage gap can be decomposed as follows; lnw ln W = ( Z Z ) β + Z ( β β ) m f m f m f m f or lnw ln W = ( Z Z ) β + Z ( β β ). The subscripts m and f are m f m f f m m f to represent males and females. The choice of method is dependent upon whether it is assumed that the male or female wage structure would be present in the absence of discrimination. The unexplained earnings differential (d) is then given by; d = exp( Z f ( β m β f )) 1 if the male wage structure is deemed to be the relevant one, and d = exp( Z m ( β m β f )) 1 if it is the female wage structure. The absence of discrimination is, however, unlikely to lead to the current male wage structure or current female structure prevailing, therefore the wage gap is further decomposed using the technique proposed by Cotton (1988). 6
lnw ln W = Z' ( β β*) + Z' ( β* β ) + ( Z Z )' β *. m f m m f f m f With β * being the representation of the estimated non-discriminatory wage structure, given by; β* = Ωβ + ( 1 Ω) β, with Ω being the proportion of the sample m f made up by men. The first term of the decomposition is the male overpayment, the second term is female underpayment and the final one is that which is explained by labour market characteristics. Interestingly this allows for different types of discrimination to be identified, Becker (1957) indicates that employer discrimination will be revealed by female underpayment, whilst employee discrimination will be typified by male overpayment. The second objective is achieved by applying the method pioneered by Blau and Kahn (1992); accordingly the cross-country wage differential can be de-composed as follows: The wage equation for male worker i in country j is; lnw = Z β + σ θ. ij ij j j ij Where lnw ij is the log of the hourly wage for worker i in country j. Z ij is the vector of explanatory variables. β j is the vector of estimated male coefficients in country j. σ j is the residual standard deviation of male wages in j. θ ij is a standardised residual, with mean zero and variance 1. The male-female wage gap for country j is; 7
D lnw lnw = δz β + σ δθ j mj fj j j j j The f and m subscripts refer to male and female averages, the δ prefix signifies the average male-female difference for the immediately following variable. The pay gap difference between two countries j and k can then be de-composed as follows; D D = ( δz δz ) β + δz ( β β ) + ( δθ δθ ) σ + δθ ( σ σ ) j k j k k j j k j k k j j k This reveals that the pay gap difference between two countries is the sum of four terms. The first term being the contribution of inter-country differences in productive characteristics. The second the impact of male-female price differentials for productivity characteristics in each country. The third compares the relative positions of women when their wage residuals are ranked in the distribution of male wage residuals; this reflects differences in unmeasured characteristics. Finally the fourth term reveals inter-country differences in residual inequality, in effect it is the price of the unobserved characteristics from term 3. Since the mean male wage residual equals zero, δθ can be estimated for each country by estimating female wages with the male coefficients imposed on their wage function. Then the resultant residuals are used to estimate the average female position in that country s distribution of male residuals. (Blau 1996). The final aim of identifying changes over time is addressed by repeating the above processes for 1991, the earliest year where data is available for both countries. 4. Results 8
Table 2 Country Year Gender No. Obs. Mean Wage Gender No. Obs. Mean Wage Wage Gap Germ. 91 Male 2728 18.39DM Fem. 1937 13.90DM 32.30% Germ. 96 Male 2234 24.10DM Fem. 1875 18.45DM 30.60% UK 91 Male 2191 5.81 Fem. 2174 4.18 39% UK 97 Male 1738 7.38 Fem. 1725 5.67 30.10% Table 2, above, displays the wage gap expressed as a proportion of the mean female wage in both countries for the years considered. Initially the wage gap was narrower in Germany, 32.3% compared with 39%, however over the period there was a slight narrowing to 30.6% in Germany and a much larger narrowing to 30.1% in the UK. This suggests that by 1996/7 women in the UK were in a marginally more favourable position. Table 3 Germany 1996 Male Female Variable Mean Coefficient Stan. Err. Mean Coefficient Stan. Err. Constant 1 2.3897** 0.0507 1 2.0500** 0.0678 Exp 20.19 0.0227** 0.0032 19.51 0.0350** 0.0039 Exp2 524.3-0.0005** 0.0001 496.7-0.0007** 0.0001 Educ 13.73 0.0403** 0.0031 13.52 0.0479** 0.0043 PT 0.019 0.0485 0.0575 0.275-0.0851** 0.0245 Public 0.222 0.0363* 0.0196 0.37 0.1101** 0.0235 Large 0.536 0.1675** 0.0185 0.455 0.1186** 0.0243 Small 0.177-0.1499** 0.0238 0.261-0.1132** 0.0277 Marr 0.715 0.0410** 0.0193 0.651-0.0525** 0.0237 East Germ 0.284-0.2561** 0.0183 0.378-0.2712** 0.0230 Agric 0.016-0.2308** 0.0635 0.014-0.1801** 0.0846 Serv 0.079-0.0940** 0.0297 0.18-0.0811** 0.0290 Fin 0.043 0.0237 0.0391 0.077 0.1224** 0.0384 Dep. Var. Ln Wage Ln Wage Mean 3.182 2.915 St. Dev. 0.4399 0.492 Obs. 2234 1875 R-squ. 0.303 0.235 RSS 301.176 347.49 Log-Like -931.6-1080.23 9
D-W Stat. 1.89 1.99 ** and * represents significance at the 5% and 10% levels respectively. The earnings function estimates for German men and women in 1996 are reported in Table 3 above. Making comparisons across the genders the results are somewhat atypical in that women actually enjoy greater returns from education, both Black et al (1999) and Brookes et al (2001) report the opposite in earlier years, indicating that women s returns have improved. They also have greater returns from experience, which concurs with the same two studies, (Black et al 1999 and Brookes et al 2001), perhaps suggesting that the longer that women remain in the labour market the fewer the concerns relating to continued labour force attachment. In this case the major driving forces behind the wage gap are, firstly, the negative wage effect for the 27.5% of women who are part-timers. Ermisch and Wright (1993) find that the lower elasticity of supply for part-time workers means that their hourly wage can only at best equal comparable full-timers. Secondly, the differential returns from employment in a large firm, with the 53.6% of men affected enjoying better returns than the 45.5% of women, which could suggest that men are more successful in capturing economic rent from their employers. However given the incidence of centralised bargaining in the German labour market, it is more likely that men are more successful at gaining promotions, Miller (1987) and Brown et al (1980) find the same for the UK and USA respectively. And finally, the opposite impact that marriage has upon male and female wages. The 71.5% of men in the sample who are married receive positive and significant returns, whereas the 65.1% of women have significant but negative returns. Both Waldfogel (1995) and Dolton and Makepeace 10
(1987) find the same, suggesting that the employers read marriage as a different signal of motivation and labour force attachment from men and women. Table 4 UK 1997 Male Female Variable Mean Coefficient Stan. Err. Mean Coefficient Stan. Err. Constant 1 0.9004** 0.0627 1 0.7970** 0.0683 Exp 16.65 0.0454** 0.0035 17.53 0.0339** 0.0038 Exp2 396.5-0.0009** 0.0001 430.5-0.0007** 0.0001 Educ 14.95 0.0424** 0.0036 14.32 0.0458** 0.0037 Tenure 4.859 0.0073** 0.0020 4.44 0.0128** 0.0025 Kids 0.781-0.0044 0.0122 0.777-0.0521** 0.0135 Public 0.211 0.0586** 0.0284 0.405 0.2054** 0.0293 PT 0.035 0.0296 0.0614 0.355-0.1324** 0.0263 Marr 0.602 0.1028** 0.0271 0.601 0.0469* 0.0255 Small 0.403-0.1282** 0.0248 0.518-0.1608** 0.0257 Large 0.205 0.0863** 0.0292 0.183-0.0248 0.0328 Agric 0.012-0.2288** 0.0998 0.004 0.1617 0.1743 Min 0.01 0.2618** 0.1096 0.003 0.2857 0.1880 Serv 0.152-0.2028** 0.0328 0.197-0.1110** 0.0340 Trans 0.081-0.0484 0.0404 0.034 0.1333** 0.0636 Fin 0.145 0.2317** 0.0328 0.151 0.2235** 0.0360 Dep. Var. Ln Wage Ln Wage Mean 1.999 1.736 St. Dev. 0.546 0.539 Obs. 1738 1725 R-squ. 0.345 0.293 RSS 339.3 353.6 Log-Like -1046.5-1080.9 D-W Stat. 1.94 1.88 ** and * represents significance at the 5% and 10% levels respectively. Table 4, above, displays the regression results for men and women in the UK in 1997. In this case the results are slightly more typical in that men receive higher returns for their experience, which again concurs with Brookes et al (2001), however once again it is the women who enjoy greater returns from education. As with the German case it is the behaviour of the part-time, marriage and large firm variables that have the most significant impact upon the wage gap. There is again the negative wage effect for the 11
part-time women. There are positive returns from marriage for both genders, although the male returns are clearly higher, so UK employers do not appear to view continued labour force attachment from married women in the same fashion as their German counterparts. However the negative returns for women with children indicate that UK employers tend to view the presence of children as the crucial indicator. Finally, men have significant and positive returns from large firm employment compared to the insignificant returns for women, again suggesting that men are either, more able to capture economic rent, or achieve greater success in promotions. Table 5 Wage Gap Decompositions Country Year Wage Gap Male Over. Fem. Under. Explained Germ. 91 32.30% 33.40% 44.40% 22.20% Germ. 96 30.60% 34.70% 43.20% 22.10% UK 91 39% 37.70% 39.50% 22.80% UK 97 30.10% 37.80% 45.90% 16.30% The decomposition results are reproduced in Table 5 above, the 1991 regression results are not reported but are available on request. There is evidence of both employee discrimination, through male overpayment, and employer discrimination, through female underpayment, in all cases. For Germany the 3 separate elements of the decomposition remain remarkably stable over the period. Despite the slight narrowing of the wage gap the proportions designated to male overpayment, female underpayment and differences in characteristics are virtually identical. However for the UK with the larger narrowing of the differential there is a clear switching of shares between the explained term and female underpayment. If the sum of the underpayment and overpayment terms is interpreted as the upper bound of discrimination, the discrimination term was 77.8% of the wage gap in Germany in 12
1991 this had risen slightly to 77.9% by 1996. For the UK the original position was 77.2% which had risen to 83.7% by 1997. Assigning values to these reveals that on average German women were paid up to 3.49DM per hour less as a result of discrimination in 1991, this rose to 4.4DM per hour by 1997. For the UK the loss of income started at 1.26 per hour and rose to 1.43 by 1997. Table 6 Cross-country Decompositions Year Term 1 Term 2 Term 3 Term 4 Total 1996/7-2.1-1.9-2.2 5.7-0.5 1991 2.1-1.7-1.9 8.2 6.7 Table 6, above, summarises the results from the Blau & Kahn (1992) cross-country decompositions from the perspective of German women. By the end of the period studied, 1996/7 the German gender wage gap was 1/2 a percentage point wider than in the UK, this slight relative disadvantage was actually the sum of a number of opposing factors. German women were 2.1 percentage points worse off as a result of their relative endowment of labour market characteristics, they were also 1.9 percentage points worse off due to the returns from these characteristics. They also suffered to the extent of 2.2 percentage points due to their endowment of unobserved characteristics, but were 5.7 points better off from the prices paid to these characteristics. The most revealing element of this analysis is the comparison of each of the terms at the start of the period and at the end. Clearly the most significant change identified by this paper is the relatively rapid narrowing of the UK wage gap, in 1991 they were 6.7 13
percentage points worse off than their German counterparts. By the end of the period they were actually 1/2 a percentage point better off and closer inspection of the decompositions reveals the causes of this. The majority of this 7.2 percentage point improvement is made up from changes in term1, these indicate that UK women improved their position by 4.2 percentage points as a result of changes in their relative endowment of labour market characteristics. Almost all of the 4.2 percentage point improvement by UK women in this area is due to two factors, a relative improvement in the numbers of women employed in large firms and in the public sector. Both countries showed employment in these areas having a significant and positive impact upon the individual wage. In 1991 22.6% of German males and 35.5% of German females were employed in the public sector, by 1996 this had changed little with men at 22.2% and women at 37.0%. This contrasts with the UK position of 22.5% male and 35.1% female in 1991 which became 21.1% male and 40.5% female by 1997, a clear improvement in the relative endowment of labour market characteristics for UK women. Similarly with those employed in firms with more than 500 employees there was a comparable trend that again favoured women in the UK. 5. Conclusion Briefly summarising the major findings from this paper are as follows. Firstly the gender wage gap was initially narrower in Germany. Secondly there was a narrowing of the gap in both countries over the period considered, the narrowing was at a much faster rate in the UK so that by the end of the period the differential was marginally 14
narrower in the UK. Thirdly, as with Brookes et al (2001) and Black et al (1999), in both countries as the wage gap has narrowed the proportion assigned to 'discrimination' or the unexplained term has increased, with both male overpayment and female underpayment being clearly present in all cases. Lastly, the majority of the improvement for women in the UK was due to more women finding employment in the public sector and with large firms. It is clear from these results, particularly for the UK, that women are able to narrow the wage gap by obtaining the right labour market characteristics, be that more education, employment in large firms, more experience or employment in the public sector. It also clear that women actually enjoy more favourable returns from certain characteristics, i.e. education and experience. However it does appear that as the wage gap narrows a stubborn rump of discrimination still remains, and in both countries there are a small number of factors behind this. The discrimination almost entirely results from the treatment of women employed by large firms, married women and, for the UK, women with children. It is most likely that within large firms superior performance by men in achieving promotion is the key, consequently policies to address this would be beneficial. Furthermore the treatment of married women as well as women with children suggests that employers form different expectations about the future behaviour of these women when measured against comparable men. This may result from the choices women have to make in response to the expense and inflexibility of the 15
available child care. Therefore, policies which enable greater numbers of women to remain in the labour force may well be capable of changing this mindset. References Becker, G. S. 1957 The Economics of Discrimination. The University of Chicago Press. Black, B. Trainor, M. and Spencer, J.E. 1999 Wage Protection Systems, Segregation and Gender Pay Inequalities: West Germany, the Netherlands and Great Britain. Cambridge Journal of Economics Vol. 23, No. 4 1999. Blanchflower, D.G. and Oswald, A.J. 1989 International Patterns of Work. In Jowell, R., Witherspoon, S. and Brook, L. (eds), British Social Attitudes Special International Report, London, Gower. Blau, F. 1996 Where are we in the Economics of Gender? The Gender Pay Gap. National Bureau of Economic Research, Working Paper No. 5664. Blau, F. and Kahn, L. 1992 The Gender Earnings Gap: Learning From International Comparisons. American Economic Review May 1992. Brookes, M., Hinks, T. and Watson, D. 2001 Comparisons in Gender Wage Differentials and Discrimination Between Germany and the United Kingdom. Review of Labour Economics and Industrial Relations, vol. 15, No. 2, June 2001. 16
Brown, R., Moon, M. and Zoloth, B. 1980 Incorporating Occupational Attainment in Studies of Male-Female Earnings Differentials. Journal of Human Resources Vol.15 No. 1. Callan, T. et al 1996 Gender Wage Differentials: New Cross- Country Evidence. Luxembourg Income Study. Working Paper Series No. 134. Chiplin, B. and Sloane, P.J. 1976 Personal Characteristics and Sex Differentials in Professional Employment. Economic Journal vol. 86 1976. Cotton, J. 1988 On the Decomposition of Wage Differentials. Review of Economics and Statistics, vol. 70, pp 236-243. Dolton, P.J. & Makepeace, G.H. 1987 Marital Status, Child Rearing and Earnings Differentials in the Graduate Labour Market. Economic Journal 97 Dec. 1987. Employment in Europe 1992. Commission of the European Communities. Ermisch, J.F. and Wright, R.E. 1993 Wage Offers and Full-time and Part-time Employment by British Women. Journal of Human Resources Vol.28 No. 1 Winter 1993. Gerlach, K. 1987 A Note on Male-Female Wage Differences in Germany. Journal of Human Resources Vol.22 No. 4 1987. 17
Greenhalgh, C. 1980 Male-Female Wage Differentials in Great Britain: Is Marriage an Equal Opportunity? Economic Journal vol. 90 1980. Gottschalk, P. and Smeeding, T.M. 1997 Cross National Comparisons of Earnings and Income Inequality. Journal of Economic Literature vol. XXXV, No. 2. Gunderson, M 1989 Male-Female Wage Differentials and Policy Responses. Journal of Economic Literature Vol. XXVII March 1989. Harkness, S. 1996 The Gender Earnings Gap: Evidence From the UK. Fiscal Studies Vol. 17 No.2 1996. Hubler, O. 1991 Sex Wage Discrimination and Sex Dependent Earnings Determinants. Jahrbucher fur Nationalokonomie und Statistik Vol. 208. Joshi, H. and Paci, P. 1998 British Cohort Studies. Unequal Pay for Women and Men: Evidence From the The MIT Press. Miller, P.W. 1987 The Wage Effect of the Occupational Segregation of Women in Britain. Economic Journal 97 Dec. 1987. Oaxaca, R. 1973 Male-Female Wage Differentials in Urban Labor Markets. International Economic Review, vol. 14, No. 3, October 1973. 18
Traxler, F. 1996 Collective Bargaining and Industrial Change: A Case of Disorganisation? A Comparative Analysis of 18 OECD Countries. European Sociological Review Vol. 12. Waldfogel, J. 1995 The Price of Motherhood: Family Status and Women s Pay in a Young British Cohort. Oxford Economic Papers 47 1995. Wright, R.E. and Ermisch, J.F. 1991 Gender Discrimination in the British Labour Market: A Reassessment. Economic Journal Vol. 101 May 1991. Zabalza, A. and Tzannatos, Z. 1985 The Effects of Britain s Antidiscriminatory Legislation on Relative Pay and Employment. Economic Journal 95 Sept. 1985. 19