October 2016 newsletter – The gender earnings gap at the LSE

Oriana Bandeira was recently asked to examine the university’s own data to measure the gender gap at LSE. This a brief summary of her findings. 

Like most organisations, universities pay the average woman academic less than her male counterpart. Unlike most organisations, universities have access to excellent data and employees who have the skills to analyse the determinants of these gaps. Yet, statistical analyses of university data are surprisingly rare and so are evidence-based policies. The London School of Economics (LSE hereafter) took a first step towards change by asking Oriana Bandiera, one of its economics professors, to measure gender gaps among academics and professional service staff. This column summarises her findings for academics, both on earnings and promotions.

The raw difference in average earnings is in line with the national average across occupations: women academics are paid 16.5 per cent less than their male colleagues, slightly less than the national average of 18 per cent. Mean differences are of course difficult to interpret as they might be driven by differences in traits that determine productivity. The key innovation of Bandiera’s analysis is that she exploits the fact that LSE, as all research universities, collects information on its academics’ projected REF scores. These are a good proxy for research productivity, which, in turn is a key determinant of pay (at the LSE, 1 point improvement in REF scores is associated with 18 per cent higher earnings). The analysis can thus assess whether women are paid less because they are less productive or whether they receive lower pay for the same level of research productivity. The evidence favours the latter: Men and women have similar REF scores, and controlling for REF scores does not explain the gender gap. Differences in age and tenure (women tend to be younger) explain some of the gap, but controlling for all these factors together leaves an unexplained 11 per cent difference.

The rest of the analysis sheds light on the anatomy of this difference always controlling for age, tenure and REF scores. Three findings are of note.

First, the gap is larger at higher deciles, ranging from 0 at the first decile to 9 per cent at the median, to 30 per cent at the highest decile. This shows that the gap is driven by highly paid academics, who are disproportionately male.

Second, women tend to work in departments that pay less. For instance, women account for 50 per cent of faculty in Anthropology and less than 25 per cent in Finance and Economics. This differential sorting explains about half of the gap, but a large gap remains among in high paying departments. While the number of women is too low to estimate department specific gender gaps, pooling Economics with related quantitative departments (Accounting, Management and Finance) reveals a 19 per cent gap.

Third, differences in rank (Lecturer, Reader, Professor) account for 25 per cent of the earnings gap. In other words, women earn less both because they are less likely to belong to the higher ranks and because they are paid less at parity of rank. In line with the quantile estimates, the gap is largest among Professors.

The last set of findings suggests that the gap is partly due to differences in promotion. The second part of the report tackles this question directly by following the cohorts of men and women hired as lecturers between 1998 and 2002 from the start of their career until today. These cohorts were chosen as they were the only ones to guarantee a long enough observation period for Lecturers to have been promoted to Professors.

Looking at earnings along the entire career (from recruitment till 2015) reveals that the gap is zero for the first four years and it grows at a steady rate after that. This is matched by a difference in promotion probabilities to the rank of Reader and Professor. The differences are stark: after 10 years, 35 per cent of men, but only 20 per cent of women are estimated to be readers, but the probabilities start converging after year 11, so that after 17 years both men and women have a 50 per cent chance of having been promoted to reader. However as the Reader promotion gap starts to close, the Professor promotion gap widens. Men start being promoted earlier than women and, by year 15, the probability that a man has been promoted to Professor is more than twice as large: 24 per cent for men and 11 per cent for women. The promotion gap does not close in our sample period: 17 years after they were hired as Lecturers, women are still half as likely to have been promoted to Professors.

Taken together, the findings reveal substantial earnings and promotions gap at parity of age, tenure, and, most importantly, research productivity. What is left unanswered is why the gap exists. One possibility is that, due to social norms that place most of the burden of childcare on women, men are more mobile and hence have a higher outside option when bargaining for salary. While this, and other possible explanations, create demand for future research, the current analysis enables the LSE to use evidence to inform its gender policies and hopefully serves as an example for other universities equally concerned about gender inequality.