Definitions and Assumptions for the Analysis of the Salary Situation

There are more than 4,000 University employees coded as faculty in the UW payroll system. Taking out auxiliary faculty (clinical faculty, residents, visiting scientists, teaching associates), research faculty (including postdoctoral research associates), and part-time instructors, there are more than 2,700 teaching faculty employed by the University. For purposes of this study, however, we narrowed the universe even further to include the approximately 1,200 full-time Seattle campus faculty who are on 9 or 10-month appointments and are budgeted at least 50 percent on the instructional budget. Thus, only a minority of those officially classified as "faculty" is included in the following analysis.

Those who are excluded have special characteristics that make comparisons especially difficult. The pay schedule for faculty on 11 or 12-month appointments (largely in Health Sciences) is not simply a pro-rata adjustment of the monthly salaries for 9-month appointments, but may reflect an implicit trade-off of security for a lower monthly wage. Some faculty members in the Health Sciences also receive income in addition to their base university pay. Another group that we have excluded is faculty with less than 50 percent teaching responsibilities. We have also restricted the analysis to Seattle campus faculty.

For our initial analysis of temporal trends, salary is measured as monthly income in constant 1992 dollars. This means that additional summer salary (for teaching or research) and administrative supplements is excluded. The primary indicator of salary levels is the mean or the simple arithmetic average of all faculty in a population (college, division, or department). In making comparisons over time, two major problems must be considered: changes in composition and variance around the mean.

The primary objective of temporal analysis is to measure the improvement (or lack of improvement) in compensation; e.g., has the average faculty member experienced a rise in pay? Changes in the salaries of individuals over time are included in the aggregate, but so are changes in the composition of individuals. If a highly paid faculty member retires and is replaced by an assistant professor with a starting salary $40,000 lower than the retiring faculty member, then the overall mean may be depressed even if all continuing faculty members experienced a salary increase. (The opposite could also happen the appointment of a highly paid faculty member could boost the mean relative to the experiences of individuals.) To minimize the impact of composition, we analyze trends within ranks (professor, associate professor, and assistant professor) and for large aggregates (the Seattle campus, and major divisions of the College of Arts and Sciences).

The mean is simply the sum of all salaries divided by those receiving a salary. Changes in the mean do not represent the average individual in the distribution (this would be the median), nor do they reflect the experiences of those with higher or lower salaries. To provide some check on this question, we also examined trends for selected points in distribution: the 25th percentile, the 50th percentile (the median), and the 75th percentile. Although not perfect, the mean appears to capture most of the changes that would be reflected in other measures. In addition to the mean salary level, we present two other summary measures, the standard deviation and the coefficient of variation. The coefficient of variation (the standard deviation divided by the mean) is a first order indicator of inequality in a distribution.

Inequality is used here as a descriptive statistical term to indicate the average dispersion of salaries relative to the mean. These descriptive terms - equality and inequality - are sometimes conflated with equity and inequity that refer to whether variance is justified relative to external criteria. Different observers can look at the same data and come to different conclusions regarding equity depending on their evaluations of market forces and other criteria. Our focus here is simply on measures of inequality.