Pathways and Pitfalls in a Salary Equity Analysis of University Faculty and Staff Lauren Young Office of Institutional Analysis University at Buffalo Beyond Overworked and Underpaid National Data ID: 1028515
Download Presentation The PPT/PDF document "How Do You Know When You’re Underpaid?" is the property of its rightful owner. Permission is granted to download and print the materials on this web site for personal, non-commercial use only, and to display it on your personal computer provided you do not modify the materials and that you retain all copyright notices contained in the materials. By downloading content from our website, you accept the terms of this agreement.
1. How Do You Know When You’re Underpaid? Pathways and Pitfalls in a Salary Equity Analysis of University Faculty and StaffLauren YoungOffice of Institutional AnalysisUniversity at Buffalo
2. Beyond Overworked and Underpaid – National Data*Title IX prohibits gender discrimination in 1972
3. Beyond Overworked and UnderpaidWomen made up 39% of full time faculty and 48% of part time faculty in 2004IPEDS shows underrepresented minorities making up only ~13% of faculty of known ethnicity in 2010Issues exacerbated in STEM disciplinesCosts include threats to faculty and staff morale and inability to foster a culture of diversity Cumulative costs of inequity to faculty and staff (and students?) over lifetimes
4. Identifying Alternative Predictors of Salary InequityTraining and experienceHighest degree earnedTime since degreeTime in titleProductivity and meritNumber of publicationsGrant dollars awardedRank
5. More Alternative Predictors of Salary InequityQuality of workCitationsAwardsDisciplineMarket factorsAverage pay (CUPA)Past/present administrative responsibilitiesOthers?
6. OFCCP Minimum Set of Salary PredictorsTime at institutionTime in titleAgePart time vs. full timeSalaried vs. hourlySalary grade or rankLocationSimilarly situated employment group (SSEG)Gender Race
7. Three Equity Analysis MethodsSingle regression analysis of total populationSingle regression of majority group with analysis of discrepancies in minority group (residual analysis)Comparison of separate regression analyses of majority and minority groups (Oaxaca decomposition)All analyses have common goal of describing unexplained wage gap
8. Single Regression with Total PopulationSimple and efficientRelatively easy to communicate to non-statisticiansMethod assumes that other predictors of salary have same impact on men and women; if not, inequity can be maskedSSEGs typically analyzed using separate models, assuming minimum headcounts
9. Hierarchical Blocked Regression Explains the “Over and Above” + GENDERor+ UNDERREPRESENTEDMINORITY STATUSRankFTEHighest DegreeYears in TitleDiscipline Market FactorPublicationsCitations
10. Hierarchical Blocked Regression Explains the “Over and Above” – Sample DataPredictorChange in Salary (B)Significance(p)Change in Explained Salary (R2)Block 156.3%Full Professor------Associate Professor (vs. Full)-20,500.000Assistant Professor (vs. Full)-35,300.000FTE5,800.254Years in Position200.065Discipline Market Factor ($K)800.000Block 2Female-6,550.0009.5%Unexplained wage gap,men>womenStatistically significant predictors of salary (p<.05)Gender accounts for salary over and above other predictors
11. Majority Group Regression and Residual AnalysisRegression equation predicts salary among majority group (men, whites, etc.)Residuals, or discrepancies between minority’s actual salary and what majority regression predicts can indicate patterns of biasAverage residual = unexplained wage gapDoes not mask different pay rates for groupsNo simple test to identify statistical significance or differential effects of other single predictors
12. Majority Group Regression – Men Only
13. Majority Group Regression - Residual Analysis of Women
14. Two Regression Model – Different Effects for Two GroupsTwo regression model permits evaluation of different effects of experience on salary for men and women
15. Two Regression Model – Oaxaca DecompositionCreates statistical expression for differential prediction of salary for men and women Combines differences in prediction between majority and minority for more precise calculation of unexplained wage gap butTests for significance of differences between models and of unexplained wage gap are more complex and time consumingProcedure not intuitive or easy to explain to non-statisticians
16. How Significant is Significance?Statistical significance generally indicates probability that differences would be found in random sample butUnexplained wage gap is based on actual population, not random sampleSize of unexplained wage gaps (“practical significance”) can be more important than statistical significance at p<.05Consider size of population being analyzed and comprehensiveness of modelGood judgment is the key!
17. A Salary Equity Analysis of FacultyProactive approach, based on OFCCP audit method and minimum predictor listAttempted to include all with faculty appointmentsSingle regression, total populationSubstituted FTE for part time / full time dichotomyAdded discipline market factor (i.e., average of ladder faculty salaries at AAU peer institutions) to control for demand where appropriate Separate analyses for SSEGs: tenure track, non tenure track, visiting/adjunct, equal opportunity center, librarians
18. Problems with PredictorsStrong intercorrelations or multicollinearityAge, years at institution, years in title highly correlated and effectively measure the same thingCannot statistically disentangle the unique effects of eachFocus on effect of gender or race/ethnicity might argue for keeping all predictors as “background” controlsChose to keep only years in title and eliminate others as redundant because other predictors added little and made explanations less clearLength of appointment highly related to discipline12-month faculty appointments localized within health sciencesHigher salaries for 12-month appointmentsAdjusted 12-month salaries to “9-month level” (x .818)
19. Problems With PredictorsCategorical variables, low counts in some categoriesOnly one visiting/adjunct professor with full rank: contrasted assistant vs. combined full/associate ranksMost librarians are associate or senior assistant, few full or assistant librarians: combined full/associate as tenured vs. non-tenured othersSome SSEGS just too small Affirmative Action manually reviewed theseMissing predictorsProductivity: publications, grant dollarsQuality: citations, awards
20. Salary Equity Results for Tenure Track FacultyPredictorChange in Salary (B)Significance(p)Change in Explained Salary (R2)61.7%Full Professor------Associate Professor (vs. Full)-33,806.000Assistant Professor (vs. Full)-47,362.000FTE14,523.103Years in Position197.007Discipline Market Factor ($K)613.000Female-1,221.4260.0%Not close tosignificance!Significancefor appropriatepredictors ofsalary
21. Salary Equity Results for LibrariansPredictorChange in Salary (B)Significance(p)Change in Explained Salary (R2)65.6%Tenured Librarian18,646.000FTE18,767.060Years in Position451.004Female3,058.2330.9%Substantial unexplainedwage gapGender effect not approaching statisticalsignificance
22. A Salary Equity Analysis of StaffPredictors adapted from OFCCP minimum listAttempted to evaluate all with classified or professional staff appointmentsIncluded highest degree, which varies considerably more among staff than among faculty (a very large majority of the latter have PhDs)Include time at institution rather than time in title Still highly correlated with age, time in titleSlightly stronger relationship with salary, more intuitive predictor in this groupSSEGs based on job categories provided by affirmative action office
23. Job Classification IssuesComputer professionals include:Programmer/analysts Database administratorsTechniciansAll professional staff with similar types of responsibilitiesSenior staff assistants share one title and one pay gradeResponsibilities vary widelyProfessional pay grades cover a large rangeNo disciplinary market factor exists to correctCan only explain 21% of variation in salary for senior staff assistants (vs. 74% for computer professionals)Other job categories include broad titles similar to “senior staff assistant”
24. Salary Grade IssuesProfessional salary grades much broader than classified salary gradesNarrower classified salary grades may subsume most of variation in salaryMight inequities based on gender or race/ethnicity be harder to detect among classified staff?Salary grades may vary little (or not at all) in some job categoriesSalary grades often have linear relationship to salary in lower and middle levels but not at the highest levels
25. Confidence Intervals for Professional and Classified Staff Salaries Within Grades
26. Salary Equity Results for Computer ProfessionalsPredictorChange in Salary (B)Significance(p)Change in Explained Salary (R2)74.0%Salary Grade (Professional)18,090.000FTE11,098.210Years at Institution723.000Less than Bachelors Degree3,949.126Bachelors Degree------Graduate Degree1,013.517Minority1,342.5990.0%
27. Salary Equity Results for Administrative Support ProfessionalsPredictorChange in Salary (B)Significance(p)Change in Explained Salary (R2)23.9%FTE1,550.210Years at Institution444.000Less than Bachelors Degree-118.126Bachelors Degree------Graduate Degree-1,662.207Minority1,951.2740.4%
28. What Salary Equity Analysis Does Not DoDoes not account for bias in hiring and initial packagesDoes not address bias in promotion and tenureDoes not demonstrate why unexplained wage gaps or differential rewards may exist – only that they do or do not existDoes not address equity at the individual level
29. Questions?