Allerton Presentation for New Administrators

1
Gender Equity Salary Studies The Good, the
Bad, and the Ugly
Presentation April 3, 2008 University of Illinois
at Chicago Carol Livingstone livngstn_at_uiuc.edu
2
Gender Equity Studies The Bad, the Better, and
the Ugly

3
Why should salaries be equitable?

• Fairness the right thing to do

• Retention of best faculty

• Its the law

4
What are our goals in studying salary equity?

• To identify and correct any systematic bias

• To identify and correct any individual salary
  errors

• To emphasize the institutional commitment to
  gender equity

5
Some BAD Ways to Study Gender Equity

• Anecdotal evidence

• Simple campus-wide averages

6
Simpsons Paradox (The Fallacy of the Averages)

The average salary of female faculty members at
one institution is 64 of the average male's
salary. Does this institution discriminate
against women?
7
Suppose the institution has just two colleges,
Engineering and Social Work

Fact Engineers are paid more than Social
Workers.
Fact Engineering is predominantly a male field,
and Social Work is predominately female.
8

94,000 60,500 64
Averages are misleading
9
A BETTER Way to Look at Gender Equity

Multiple regression analysis
Dependent variable constant
independent variable 1 coefficient 1
independent variable 2 coefficient 2
independent variable 3 coefficient 3
10
Using Multiple regression to look for systematic
discrimination

Include gender or race/ethnicity as an
independent variable. A coefficient statistically
different from zero implies a correlation between
gender and salary.
11
Using Multiple regression to look for individual
discrimination

• Exclude gender and race/ethnic code from
  independent variables.

• Find the regression equation.

• For each person, see what salary the regression
  equation predicts.

12
Assumptions of Multivariate Regression

• Factors are independent

• Each factor is linearly related to dependent
  variable

• Variables can be measured accurately

• Populations are sufficiently large

• All relevant factors are included

13
Urbanas History of Gender Equity Studies

• Chancellor commissioned first one in early 90s.
  Took a year to complete.

• Found some systematic bias, individual bias
  based on gender

• Resulted in many salary corrections

• Repeated many times since then results vary

14
BOT Gender Equity Report

• All three campuses were asked to submit a gender
  equity report in June, 2000

• Included a regression analysis of salaries,
  retention and promotion studies, comparisons with
  national benchmarks

15
Urbana Gender Equity Studies

Nine studies since 1990s (hmmm, 8
½) http//www.dmi.uiuc.edu/reg
16
Urbana Process

• Tenure-system faculty only
• On-going salary, no lump sums
• Much manual data collection/fixing
• Periodic revisions, especially with input
  from CSW

17
Urbana Independent Variables

• Rank
• Department
• Years from degree
• Having a Ph.D.
• Administrator flag
• Hired in as assistant professor
• Gender
• Race/ethnic group
• Years to reach associate professor
• Years to reach full professor

18
Urbana Regressions

• All faculty combined
• Assistant Professors
• Associate Professors
• New Assistant Professors
• Others - appendix

19
Regression Evaluation
R2 usually about 0.6-0.9

Model significant at the 0.0001 level
20
Significance of Gender term Regressions (2004)
Regression
Gender effect
R2
21
Coefficients from 2004
Dept factor ranged from 30,000 to 66,000
22
Actual Salaries as of Predicted (2006)

23
Other regressions run

• Using peer salaries instead of department dummy
  factor
• Using log(salary) instead of salary as dependent
  variable
• Added terms interacting gender with other
  variables significant but small interactions
  found with years to reach full professor number
  of other departments

24
Publication/Follow-up

• Report, general statistics, outcomes reported to
  Provost, Deans and posted on web
• Deans business managers get list of faculty
  with actual and predicted salaries
• Deans must fix or justify salaries 7 or more
  below prediction

25
The Ugly

• Claiming to have a precise answer

• Taking individual predictions as truth

• Selecting one regression (e.g. all faculty)
  result over another

• Confusing correlation with causality

26
The Ugly

Data wars! Adversarial attitudes from
administration or faculty are counterproductive.
27
Beyond Salary Equity Hiring

• Who is in the pool?
• Who applies?
• Who is on the hiring committee?
• Who is a finalist?
• Who gets an offer?
• What salary is offered?
• Who actually accepts?

28
Beyond Salary Equity Retention

• Promotions
• Teaching advising workload
• Committee assignments
• Salary increases, esp. matches
• Administrative appointments
• Sabbaticals
• Awards/Chairs
• Climate

29
Beyond Salary Equity Policy Analysis

Some data gathering is helpful, but dont get
bogged down in data. Spend your time thinking
about processes, policies, and decision points
30

Questions??