Gender Ratios in Top PhD Programs in Economics
-
Upload
new-economic-school -
Category
Documents
-
view
230 -
download
0
description
Transcript of Gender Ratios in Top PhD Programs in Economics
Gender Ratios in Top PhD
programs in Economics
Galina Hale
Tali Regev
Opinions are our own and do not necessarily represent those of the Federal Reserve
Bank of San Francisco or Federal Reserve System
12/15/2012 Moscow
Our paper is motivated by the
importance of gender segregation
At least 45% of the gender wage gap is accounted by
segregation of women into lower paying occupations,
industries, establishments and jobs.
Women are underrepresented in high status
occupations, such as science and engineering
And economics
Do gender policies, such as affirmative action, have
desired effects?
12/15/2012 Moscow 2
Reasons for gender segregation
Employer discrimination
Dynamic effects of gender composition: occupations with larger female share tend to attract more females Political power: females advocate for females (D).
Learning by employers: females perform well, reducing the bias (D).
Mentoring and social environment: female entrants prefer to work around females (S).
We find the presence of both mechanisms in econ departments – higher share of female faculty leads to more women in graduating PhD class
At this stage we can’t identify the exact mechanism of the causal effect In our sample 11% of female students and only 4% of male students have
female advisors. 12/15/2012 Moscow 3
What do we already know: Gender segregation exists
Carrington and Trosky 1995, Petersen and Morgan 1995, Bayard, Hellerstein et al 2003.
There are gender differences in academic career paths of economists
Kahn 1993, Kahn 1995, McDowell, Singell et al 1999, Ginther and Kahn 2004, Lynch 2008.
There is evidence of gender discrimination: audit studies and sex-blind hiring
Neumark 1996, Goldin and Rouse 2000.
Female students tend to work with female faculty
Neumark and Gardecki 1998, Hilmer and Hilmer 2007, Hoffmann and Oreopoulos 2009, Blau, Currie et al 2010, Bettinger and Terry Long 2004, Zinovyeva and Bagues 2010.
12/15/2012 Moscow 4
We study gender composition of faculty and
grad students in top econ departments
There is a positive correlation between share of women on the faculty and share of women among PhD students
Some of it is attributable to unobservable gender bias
Some of it is attributable to path dependence that is causal: more women on faculty => more female PhD students graduate
Challenge in establishing causality: share of female faculty is endogenous
12/15/2012 Moscow 5
Plan of the talk
Describe the data
Present the correlation results (OLS) and gender bias
effects
Present evidence of causal effects of female faculty
share on share of women among PhD students (IVs)
12/15/2012 Moscow 6
Sample: ten top econ
departments 1983-2007 All ladder faculty:
gender and academic career, including tenure, publication history
and PhD institution.
Sources: Faculty lists, CV’s, Harzing’s Publish or Perish
Departments chosen by data availability (not gender related).
750 economists, 98 of these women.
All students awarded PhDs between 1983-2006:
Gender , institution (field, academic advisor)
Sources: NSF Survey of Earned Doctorates, Proquest,
Dissertation abstracts at institutions’ libraries.
Only those who ended up getting a PhD – not all admissions
Race composition of the econ department’s entering PhD class
Gender composition of entering PhD class at the University level
12/15/2012 Moscow 7
Female faculty share in top econ
departments is still very low
12/15/2012 Moscow 8
Increase in female share is uneven
across institutions
12/15/2012 Moscow 9
Inst 1
Inst 2
Inst 3
Inst 4
Inst 5
Inst 6
Inst 7
Inst 8
Inst 9
Inst 10
Share of women in entering PhD
class is higher
12/15/2012 Moscow 10
And also uneven across
institutions
12/15/2012 Moscow 11
Inst 1
Inst 2
Inst 3
Inst 4
Inst 5
Inst 6
Inst 7
Inst 8
Inst 9
Inst 10
In most institutions female shares are trending
up, but there is sufficient heterogeneity for
analysis
12/15/2012 Moscow 12
We measure correlation between
these shares using OLS regression
12/15/2012 Moscow 13
Female faculty share has a large
“effect” on female student share
12/15/2012 Moscow 14
There are time-invariant differences
across departments: Institution FEs
12/15/2012 Moscow 15
Inst 1
Inst 2
Inst 3
Inst 4
Inst 5
Inst 6
Inst 7
Inst 8
Inst 9
Inst 10
This correlation is robust
12/15/2012 Moscow 16
Strong evidence of positive
correlation – summary
Magnitude of coefficient is ~ 1:
1 pp increase in share faculty is associated with 1 pp higher share of students.
Doubling the 2000 faculty share from 0.1 to 0.2, is associated with an increase of student share from 0.2 to 0.3.
OLS results are robust to:
including faculty share in female friendly fields.
Non-linear effects of faculty share (found that effect is linear).
Differences in tenure of female faculty
Institution-specific trend.
12/15/2012 Moscow 17
What about causality? IV! Omitted variables might be responsible for both shares,
producing spurious OLS results
Need an instrument which changes the female faculty share but uncorrelated with student share in any other way.
Instrument: male exits.
Mechanically increase the female faculty share
female share = #females/ (#females+#males)
True for lagged exits as well if not replaced immediately
12/15/2012 Moscow 18
Find causal effect of female faculty share
of female student share
12/15/2012 Moscow 19
Testing the exclusion restriction:
exits of male faculty have no direct effect on
female student share
Share of female faculty does not predict male exits (no
reverse causality) – specification test
Limit to exogenous male exits – robustness test
Predict male exits with individual-level exogenous
variable – robustness test
12/15/2012 Moscow 20
Female faculty share does not predict
male exits
12/15/2012 Moscow 21
Are male exits exogenous? Male exits could be due to
retirement (older) – could be anticipated
transfers (young and old, move within sample) – could be
associated with gender policies
failure at tenure (young - non tenured, move out of sample) – most
likely exogenous and unanticipated
12/15/2012 Moscow 22
First stage with alternative instruments and
alternative sets of controls
12/15/2012 Moscow 23
Second stage largely unaffected
12/15/2012 Moscow 24
What if male exits are still endogenous to
gender policies? Instrument them!
Predict exits at individual level using age and
publications as instruments:
Young male exit due to poor chance at (or denied) tenure
Older male exit due to retirement
Two equations for “zero” stage
12/15/2012 Moscow 25
The result still holds
12/15/2012 Moscow 26
Conclusion
We find that higher share of female faculty leads to
more female graduate students
We are pretty sure of this being a causal effect
We cannot tell exactly which mechanism is driving it
We do find some evidence of a time-varying gender bias
We also find an effect of quality of female faculty
We also find that female students are more likely than male
students to be advised by female faculty
12/15/2012 Moscow 27
Thank you!!!
12/15/2012 Moscow