英语论文网

have generated a proxy variable, also known as a Mincerian variable (Mincer, 1974). It is assumed that a proxy for actual work experience is (Oaxaca, 1973, p.697):

Occupation should not be controlled for in such decompositions, as preconceptions for respective occupations can cause occupational segregation and hence be an instrument through which labour market discrimination manifests itself (Gunderson, 2006). This could also be said about firm size and industry, nonetheless discriminatory behaviour is less likely in these scopes and hence they are suitable controls. Human capital characteristics gained outside the labour force (e.g. health state, marital status) are controlled for, as they do not contain information about discriminatory behaviour of employers. Again, we are only concerned about discriminatory behaviour in the labour market (i.e. hiring decisions and gender pay inequality). Since the econometric model is completed with a wide range of productivity characteristics, the gender wage gap in the UK for year 2003 is decomposed in the following section.

In the three-fold decomposition (Table X) the first part reports the average increase in female weekly wages if they had the same characteristics as males (Jann, 2008). The increase of 0.523 reported in the table implies that about 86% of the wage differential is attributable to differences in endowments. The second component gives a numerical value for the increase in women’s wages if male coefficients were applied to female characteristics. That is, 16.1% of the pay gap is explained by differences in coefficients. The third figure refers to the interaction term that measures the simultaneous effect of differences in endowments and coefficients which is 7.9% in our estimated model.

In the two-fold decomposition panel the gender pay gap is split into two parts, as mentioned in section Y, explained and unexplained. To derive conventional interpretations I have also converted the decomposition into an exponential form (Table W). Adjusting women’s human capital levels to the levels of males would require an increase of 63.4% in female weekly earnings. However, more than 12% of the differential remains unexplained, i.e. attributable to being female. The main drawback of the two-fold decomposition is that compared to the previous one, we observe that less of the pay gap is now attributable to endowments (0.49 as compared to 0.52). This suggests that some fraction of the explained differential was transferred into the unexplained part which has negative effects for the interpretation of results. Therefore, the three-fold regression decomposition offers a more desirable output.

It is important to look at individual effects on the gender wage differential to truly understand what factors determine the mean outcome. In table B we have a detailed estimation output from which further interpretations can be brought into the picture. It is essential to note that the positive values of the coefficients in the output table reflect the benefit of males and the negative value refers to a positive income contribution for females.

Age is in a substantial element in the explanation of gender pay gap and it is thus unsurprising that the age coefficients show a positive relationship between wage and person’s age, and negative between the square of age. This implies that both male and female earnings increase with time, howeve