英语论文网

of men in the workforce, there is a 1.3 percent increase in the wage rate (Olsen and Walby, 2004). Therefore, we can observe evident correlation between number of males and the average wage rate for a given sample in the labour force.

The findings of Olsen’s and Walby’s (2004) simulation reveal that the second half of the gender pay gap is accounted for by two other important elements. 8 percent of the differential is the result of the fact that older women had poorer educational attainment as a result of limited access to education and different social patterns decades ago. Other 38 percent are thus explained as an outcome of discriminatory behaviour in the labour market (Olsen and Walby, 2004). However, such conclusions should be drawn with caution. 38 percent is attributable to being a female but not necessarily direct discrimination. To state otherwise, more than a third of the gender wage gap simply could not be explained by human capital or other observable factors.

Other studies from the post-1980 period (that also use regression analysis) yield similar results. In 1989 Blanchflower and Oswald carried out a simple wage regression analysis where discrimination on the base of sex was estimated by a dummy variable using data from seven countries for 1985-87. It was found that British women were overwhelmingly disadvantaged as compared to other countries, being underpaid 56 percent relative to males. Paci and Joshi (1996) insist that full-time female wage would be 17 percent higher if discrimination were absent, or that gender discrimination was 8 percent of the male wage in 1991. On the other hand, Harkness (1996) uses a more similar procedure to this paper and discovers a 20 percent estimate for gender discrimination as a fraction of 22 percent log-wage differential for the full-time employees. To summarise the evidence above, it can be said that there is a sufficient ground to believe that the gender pay gap is closing up. However, the issue of discrimination has not diminished. In fact, as research shows, it is becoming harder and harder to explain gender earnings inequality by differences in skills and work life patterns. Modern studies seem to be yielding higher discrimination estimates. Therefore, this paper shall look into more recent data and explore these trends.

A similar suggestion was made by Cotton (1988), however, instead of using the averages, the parameters are weighed by the group sizes. Whereas Neumark (1998) argues that a pooled regression should be run over two groups to obtain the coefficients later used as a measure of β*. However, both Jann (2008) and Fortin (2006) criticise such approach as it may incorrectly transfer some of the unexplained fraction of the differential into the explained component. In order to avoid such misrepresentation, Jann (2008) recommends always including a group indicator in the pooled model as an additional covariate.

An important downside of the methodology presented in Oaxaca (1973) is that the discrimination and productivity elements exhibit significant amplitude depending on the wage structure (male or female) being used. Hinks et al. (2001) recognises this weakness and uses a competitive wage structure to analyse the impact of discrimination. Some studies (Olsen and Walby, 2004, and others) also criticize the conventional Oaxaca-Blinder approach for using separate regressions for males and females, which inevitably imp