Economic analysis of gender-based salary/wage discrimination

Yesterday, I went to a seminar at AUT on Testing theories of gender discrimination using linked employer-employee data, presented by Economist Isabelle Sin, based at Victoria University. The analysis explored the question of what proportion of salary difference can be attributed do a gendered bias. But it tackled the question in a way I hadn’t seen before. By exploring higher-level income data – based on PAYE information from IRD – and industry/employee data, Sin and co-authors aimed to calculate ‘productivity’ based on gender, and salary based on gender, and compare the two. Not being a quantitative analyst, some of the detailed analytic tables were hard to follow , but the gist of the analysis is that across a very large sample of private-sector, for-profit companies, women’s productivity can be calculated as 86% compared to a 100% for men, but salary is 74% compared to 100% for men (taking part-time/full-time status and certain other factors into account). This average discrepancy – effectively a 12% gender-pay-gap – varied also by age categories, and a finer grained analysis revealed the gap differs considerably by industry category.
What wasn’t defined clearly was what productivity is – this may be a well-defined concept that holds up well in economic analysis, but I struggle with the concept, in lots of ways. I don’t like way it evokes a very linear and literal conceptual model of workforce contribution. Furthermore, as noted in the questions, the lack of a gendered-analysis of industries themselves meant that those which came up as most/least disriminatory didn’t necessarily map ‘common-sense’: for instance, libraries, which tend to have a female-dominated workforce, was included in the industry-band which appeared most-discriminatory (though this may in part reflect issues of gender-distribution across that workforce). Farming in many forms was the least discriminatory, which may at least partly reflect the pay-data used… It was based on PAYE data, meaning contractors were excluded.
Overall, the analysis was interesting – it provided a very similar number to other estimates of the gender-based pay gap. But it also highlighted the value for detailed or micro-focused, and specifically gendered, analyses. Furthermore, as other research indicates that race/ethnicity also impacts salary, including here in NZ, I believe that needs to always be kept present in analysis of gender-based pay gaps.