Pay Equity Study

Revised 2015-2016


In 2015, a joint Administration-Academic Senate Committee redesigned our annual campus pay equity study of ladder rank faculty salaries.

The analyses presented in this report focus on regression models that go beyond the annual residual analysis conducted in the past (1997-2014) and include evaluation of rate of progression through the ranks. Data were examined at the whole campus level, and for 14 Schools/Units. School of Medicine faculty were examined separately in this study due to the differences in compensation associated with participation in the Health Science Compensation plan.

Analysis of salary data from July 1, 2015 (after all salary adjustments had been applied) indicate no evidence of systemic disparity in pay associated with gender and/or ethnicity at the campus level when experience, discipline, and rank are included in the model. However, there is further work to do to understand the issues around the 1) low percentage of women and minority faculty at the higher ranks and steps across campus, and 2) differences in the rate of progression through the ranks and salary disparities by gender/ethnicity in some units.

Methodology: Multiple Linear Regression Model

A series of regressions were used to examine potential correlations between gender/ethnicity variables and salary.  This approach provides a broad view of faculty employment and pay structure by the demographic variables and by experience, discipline, and rank.

  • Demographic factors enter the equation as indicator variables for Women, Asian, and Underrepresented Minorities (URM).
  • Experience variables include Years Since Degree, Years of Service, and Decade of Hire. Years Since Degree is the number of years passed from the year the highest degree was earned to the present.  Years of Service is the number of years passed since the individual became a Ladder Rank faculty member.  Decade of Hire consists of four binary categorical variables to account for the decade the individual became senate faculty:  2007 to 2016, 1997 to 2006, 1987 to 1996, or prior to 1996.
  • Discipline is accounted for by adding an indicator variable for each school. The discipline variable accounts for internal demand and a market ratio derived using AAUDE salary data for UCI’s peer institutions is used to account for external demand by field.
  • Rank includes Current Rank and Step, Initial Rank and Step at time of hire, and Progress Rate as predictor variables.

Progress Rate measures number of years the faculty member is ahead or behind normal progression through the ranks. Normative time to achieve each rank is determined by computing the number of years it would take to move from the initial rank to the current rank and step, if the individual is progressing at the university’s established normal rate.  If an individual was promoted to their specific rank/step in the normative time, then rate of progression is 0. If they took longer than normative time, rate of progression is expressed as a negative number (years). If they took less than normative time then rate of progression is expressed as a positive number (years). The Progression Matrix shows normative time table and sample calculations.

In order to evaluate whether biases exist within progression through the ranks, several box and scatter plots by gender, ethnicity, rank, and school were generated to visualize and investigate the data.  Progression rate differences by demographic groups were also tested with t-tests, ANOVA, and Bonferroni statistical methods.  Lastly, a series of regression models were run to quantify progression rate differences that may exist by gender or ethnicity.

There is a possibility that one or more of the explanatory factors in the salary regression models are correlated; we therefore evaluated the effect of multicollinearity in our models. For the whole campus data there was little evidence of collinearity and therefore all variables were included in the regression equations. However, in a small number of Schools/units there was evidence of collinearity and in those cases data is presented with and without removal of one or more of the variables from the regression analysis.


Comments are closed