Excellence Through Diversity

Report of the Council on Faculty Diversity and Inclusion 2008

Faculty Compensation

Overview of Methods

The CFDI was asked to make recommendations on how best to ensure that salaries are given equitably to all faculty on the basis of merit. As a benchmark to inform our recommendations, we examined de-identified faculty salaries in all academic units. The salaries available to us were a snapshot taken for the 2007—2008 academic year. Salaries are far from static however, so this analysis is unlikely to reflect the current situation.

Two factors should determine the salary of a faculty member: their “merit” and the “market” for their expertise. Each faculty member undergoes an internal merit evaluation every year, in an effort to set individual salaries for the coming year on the basis of “merit criteria” as described in the Faculty Handbook (scholarly productivity, teaching expertise, service and so on), and more recently, on the basis of the “Faculty Expectations¹” intrinsic to each unit. The CFDI had no basis for evaluating the merit of any individual faculty member, but we began from the premise that the merit of the populations of male and female faculty members was expected to be equal, and we used statistical methods to control for additional variables that could be expected to affect salary, particularly those related to market.

Although we had no direct way to assess the “market value” of any individual faculty member’s expertise and accomplishments, we made use of several variables that are representative of the market. The market value for a faculty member is only known when that person tests the market by seeking salary offers from multiple institutions. Since those data are not available for most faculty members, we used the variables of “years at Boston University,” and “years since degree² ,” as well as “department” in our analysis, to represent effects of market. The rationale was that the faculty member had tested the market at the time of hire, and that differences in salaries in different disciplines would be captured by the “department” variable, allowing us to adjust for these differences and examine salaries of males and females independent of disciplinary differences in salaries. We performed an additional analysis of the salaries of faculty members who were hired as Associate or full Professors with tenure, to determine the effect that this circumstance had on salary, and whether there was an interaction with gender. These faculty usually had many years since degree, but few years at Boston University.

With the assistance of Prof. Timothy Heeren (Biostatistics, SPH) we performed two main types of analyses. The first was to examine simple scatterplots for each department of the deviation of individual faculty salaries from the departmental mean as a function of years since degree or at BU. Each faculty member was represented by a symbol that indicated gender and rank and, to preserve confidentiality, the departments themselves were not identified for the committee. Individuals with salaries that deviated substantially from the distribution of their rank for their department were flagged for follow-up with the appropriate Dean and the Provost. The second analysis was a multiple regression performed on de-identified faculty salaries in large units: CAS, CRC non-CAS, and the Medical Campus³. Salaries for both Tenured/TT and NTT faculty were investigated. This analysis allowed statistical adjustment of salaries for the variables described above, namely department, and years at BU or since degree⁴. On the Medical Campus type of degree (MD or DDS, PhD, MD/PhD, Other) was also used as a variable in the analysis, since degree type has a large influence on salaries for faculty in medical settings. This analysis allows comparison of the salaries of males and females at each rank before and after adjustment for these variables. It mitigates the uneven distribution of males and females among disciplines (i.e., departments) that have different salaries based on the market, and presents a truer picture of salary equity between males and females across large units than does a simple comparison of male and female salaries at the institution.

Results
The results of the multiple regression analysis were used in two ways. The salary residual (i.e., the deviation of that individual’s actual salary from their “expected salary” after adjustment for department and years since degree) for individual faculty members was provided to the Provost for follow-up with the appropriate Dean. It allowed identification of faculty with salaries that were exceptionally low or high, compared to what might be “expected,” given their department and years since degree or hire. However, this analysis of “outliers” is not sensitive to the possibility that there may be faculty with salaries that do not stand out from the “expected values” but who nonetheless are paid less than their “merit” would indicate (“inliers”). As stated above, since the committee had no data about merit, the working assumption was that the merit of the population of male and female faculty was the same. The committee compared the overall distribution of salaries by gender by placing the salaries into quartiles at each rank and simply asking whether the salaries of males and females were evenly distributed among the salary quartiles (“expected” value = 25%). This was done both before and after adjustment of salaries to account for department and years since degree or years at BU. A summary of the findings for tenure-track faculty with unmodified titles in CAS and other Charles River Campus non-CAS schools, and for faculty in MED⁵ after statistical adjustment for department and years since degree is shown in Table 6.

There were too few NTT faculty in CAS to perform the quartile analysis by rank adjusted for department. The analysis was therefore only carried out for the CRC schools outside of CAS. The analysis showed that 20% of the NTT female Assistant Professors had 2007 salaries in the bottom quartile and 20% had salaries in the top quartile,⁶ 20% of the adjusted salaries of female NTT Associate Professors were in the bottom quartile and 20% were in the top quartile, and 40% of the adjusted salaries of female NTT full Professors were in the bottom quartile and 20% were in the top quartile.

It appears from these data that overall, female Associate and full Professors are underrepresented in the top salary quartile, after adjustment to account for disciplinary differences in salary and the distribution of males and females among these disciplines. In some units and ranks, females are also underrepresented in the lowest salary quartile. It should be noted that there are several caveats to these data. 1) The overall numbers mask some differences in the distribution of salaries within each of the three big units—in some schools and divisions within schools the salaries of males and females sorted almost equally into the four quartiles, while in others there were large differences. 2) In many schools the number of female faculty at a given rank is small, so changes in the salary of a few faculty members can have a large effect on the salary distribution in a quartile analysis. 3) As noted elsewhere in this report, the systems in place to collect and track the types of data used for this analysis are inconsistent across the University, and data for some faculty may be missing, particularly for the Medical Campus. 4) The salary analysis included the salaries of department chairs, but not other faculty who also hold an administrative appointment. If there is an interaction between gender and those serving as chairs or other administrators, this could skew the analysis. 5) Most important, the analysis was based on 2007 salary data that did not reflect salary adjustments that have been a priority for the administration in the past two budget cycles. It will be critical to continue to monitor these data in the future.