The university hired a third party vendor to review and analyze faculty salaries and perform an equitable-pay analysis for tenured and tenure-track faculty by race, ethnicity, and gender group. The study was conducted in accordance with industry-wide accepted statistical analysis methods. The purpose of the analysis is to identify whether systematic differences exist by race, ethnicity, or gender, and also highlight individual cases that require further investigation to determine whether action is required to achieve standards of pay equity.

Data Collection

The university provided key information about each tenured and tenure-track faculty member that was used in this analysis. The data include factors such as:

• Demographic group status (e.g., female or not)
• Current academic rank
• UO work experience
  • Years in current rank
  • Other years of UO service, or
  • Years in each academic rank (Assistant, Associate and Full)
• Distinguished professorship/endowed chair status
• Recurring teaching award status
• Potential years of previous work experience
• College/division/department
• Honors College
• Tenure status
• Highest level of education

Equal-Pay Analysis

Regression analysis was applied to the data set. Regression analysis is used in salary equity studies to determine whether specific groups of people have systematically lower (or higher) salaries than a baseline reference group. This same analysis can also be used to determine which individuals have salaries that are “very unusual” within the different groups described by a common discipline and rank, etc. These individuals, commonly referred to as “outliers,” may have unexpectedly high salaries, or unexpectedly low salaries, compared to the average for other people in the same discipline and rank, etc.

Analyzing Trends

Regression analysis permits the observed variations in salaries across people to be decomposed into components, so expected differences can be accounted for, and “unexplained” differences can be more easily identified. For example, given that salaries are expected to vary systematically by discipline and rank, regression analysis allows the analyst to control for these expected differences across groups of faculty so that it is easier to look for “unexplained” differences. The analysis process first controls for those factors that can be incorporated into the model and that are expected to impact salary, such as discipline, rank, department, etc., and then, additionally, analyzes the effects of indicators for gender or race/ethnicity. If these additional indicators are found to have systematic effects on average salaries among those finer groupings of employees, this evidence would suggest there may be systematic differences in salaries, solely due to gender or race/ethnic group, that show up on average, across all disciplines and ranks.

Evaluating Individual Results

The same regression analysis is used to evaluate individual salaries in comparison to the larger groupings. As outlined above, regression analysis can determine whether specific groups of people have lower (or higher) salaries than a baseline reference group. It can also be used to determine which individuals have notably different salaries  within the various  groupings described by a common discipline and rank, etc.  These “outliers” may have unexpectedly high salaries, or unexpectedly low salaries, among other people in the same discipline and rank, etc. 

Determining how different from a group average an individual salary needs to be before it is considered “remarkable” is a subjective question. Social scientists typically adopt a statistical convention that defines an outlier as a difference from the group mean that is big enough to happen only 5% of the time. With a focus on low-salary outliers, this translates into the lowest 2.5% of salaries within a group, or 1.96 standard deviations below the conditional mean.  In terms of relative differences, a salary is often considered to be remarkably low if it is less than 80% of the comparison group's mean.  A combination of these two criteria is often used. However, different thresholds can be adopted.

For the purposes of this study, the threshold used to statistically identify “negative outliers,” those earning lower-than-expected salaries, will likely be identified using a combination of these two measures: (a)  the number of error standard deviations below the group mean at which a given person's salary lies, and (b) the individual's salary as a percent of the comparison group's mean.

Important Considerations

Many of these within-group salary differences may be explained by legitimate, yet less-readily quantifiable, factors such as research productivity, administrative assignments, retention offers, and so on. Regression analysis, as a statistical approach, is not able to answer the question of what might be the complete set of “acceptable reasons why a given individual may have a lower-than-expected salary relative to others in the same group.” Many factors that legitimately affect salaries must be excluded from the regression models because they cannot readily be quantified for every faculty member and included in the overall model. 

Additionally, when outliers are defined and identified, regression analysis cannot reveal what might be the appropriate remedy for these salary differences. For this reason, this statistical approach serves as a guide for identifying outliers that warrant further investigation and helps focus attention on options to remedy these differences as deemed necessary.

Next Steps

The Faculty Salary Equity Study work group is in the process of reviewing the regression analysis results and setting the threshold that identifies “negative outliers,” namely those faculty members earning lower-than-expected salaries for the purpose of this study. Identifying “negative outliers” is critically important to fulfilling the requirements of this study. Therefore, the threshold the work group establishes will be more inclusive than the common convention of 1.96 standard deviations below expected salary and 80% of expected salary.

Once the threshold is set, members of the work group will begin to investigate the individual faculty compensation of identified “negative outliers” in the salary data. Members of the work group will be taking a closer look at the employment records and salary histories to determine whether legally permissible factors account for these large differences. Instances when an unexpectedly low salary cannot be attributed to legally permissible factors may lead to a recommendation for action to remedy the difference.

UPDATE (2/22/19)
The work group is conducting a case-by-case review of those faculty members identified as outliers. In order to complete the review, the work group is asking college and school leadership to provide answers to standardized questions regarding faculty compensation in their units.

Members of the work group will be taking a closer look at the employment records and salary histories to determine whether legally permissible factors account for these large differences. Instances when an unexpectedly low salary cannot be attributed to legally permissible factors may lead to a recommendation for action to remedy the difference.

After this information is received from the units, the work group will review it and, ultimately, make a recommendation to the provost regarding equity adjustments. The committee hopes to have a recommendation to the provost by the end of April.

UPDATE (5/20/19)
The Faculty Salary Equity Study Work Group has completed the data review and analysis required to determine if salary inequities related to gender, race, or ethnicity exist and provide recommendations for remedy as warranted. A report of preliminary recommendations was submitted to Jayanth Banavar, provost and senior vice president, in late May for his review. The provost is ultimately charged with making final determinations with respect to the, evaluating the data, and determining next steps, including equity adjustments.

A copy of the full report is available on the Faculty Salary Equity Report webpage.