Leaving Careers in IT: Differences in Retention by Gender and Minority
Status
Paula Stephan & Sharon LevinJanuary 2005
Acknowledgements
Supported by National Science Foundation: ELA 0089995; SEWP-NBER
Uses data from Sciences Resources Statistics, National Science Foundation
Focus
Considerable interest in recent years concerning low prevalence of women and underrepresented minorities in the IT workforce.
Initial focus motivated by concerns regarding equity
Interest augmented in 1990s because of key role IT sector played in economic expansion and concern that shortage of IT workers existed.
Size of IT Workforce Depends onPipeline In
Much discussion in 1990s concerned how pipeline could be expanded, making careers in IT more attractive and possible for women and minorities.
Case in point: Carnegie Mellon initiative, “unlocking the clubhouse door” which focused on recruiting and attracting women and minorities into IT programs at CM.
Size of IT Workforce Also Depends onPipeline Out
IT workforce is diminished when trained individuals leave either for
– Careers outside of IT or– Leave the labor force
IT workforce is diminished when “recruited” individuals leave.
Focus of this research is whether retention varies by gender and minority status.
Interest is on retention subsequent to working in occupation; not retention while in a degree program.
If those working in IT in ’93 had been retained in ’99 . . .
IT workforce would have had 40% more women 50% more underrepresented minorities 25% more men Conclude:
– IT workforce would have been bigger– More balanced by gender and
underrepresented minority status
Plan for Today’s Presentation
– Overview of data used– What we mean by IT trained– What we mean by IT occupations– Descriptive Data– Logit Analysis
Data
Drawn from SESTAT (college degree or higher, focus in S&E)
Integrated database built on three different NSF surveys
Years: 1993, 1995, 1997, 1999 National Survey of College Graduates National Survey Recent College Graduates Survey of Doctorate Recipients
NSCG
Sampling frame is college educated (BA or higher) 1990 Census
Surveyed in 1993 to determine if degree held in 1990 is in S&E or whether working in an S&E occupation in 1990
S&E identified sample followed in 1995, 1997, 1999
NSRCG
Sampling frame is individuals who earn bachelors or masters S&E degrees during the decade of 1990s
Refreshes NSCG but only adds those educated in U.S.
SDR
Sampling frame is individuals who earn Ph.D. degree in U.S. and indicate plan to stay in U.S.
Note: excludes individuals who earn Ph.D.s outside U.S.
Shortcomings of Data
Excludes scientists and engineers trained outside U.S. after 1990
Excludes college-trained individuals working in S&E after 1993 but not trained in S&E
Excludes associate degree holders Does not consider programming to be a
field of training in S&E or an occupation in S&E
Definition of IT Trained; IT Work
Follow lead of IT Data Project concerning definition of IT trained
Follow lead of IT Data Project and IT Workforce report for definition of IT work
Available on our web page:http://www.gsu.edu/~ecopes/itworkforce/index.htm
Definition of IT Trained: One or More Degree in…
Computer/information sciences Computer science Computer system analysts Information service and systems Other computer and information sciences Computer and systems engineers Electrical, electronics and communications
engineering if recipient also minored or did second major in area of computer or information sciences.
Definition of IT Occupations
Computer analyst Computer scientists except system analysts Information system scientists and analysts Other computer and information science occupations Other computer and information sciences Computer engineers; software engineers Computer engineers—hardware Computer programmers (Note:only programmers
picked up in SESTAT are those trained in an S&E field who work as a programmer or individuals not trained in S&E but working in an S&E occupation in 1993.)
Big Picture
Find about 1 million individuals (weighted data) working in IT in 1993 were in SESTAT in 1999.– 30% women; – 84% white – 9% Asian– 4% African American– 3% Hispanic & “Other”
Big Picture Continued
About 70% of those working in IT in 1993 were retained in 1999.
Retention rate higher for those trained: (80% vs 65%) Retention rate higher for men than women (73% vs.
66%) Retention rate higher for whites than African Americans
(70% vs. 66%) Retention rate higher for Asians (70%) than whites
(70%)
Table II. Weighted means for individuals employed in IT occupations in 1993 and in SESTAT in 1999.
All Females Males Whites Asians African Americans
Hispanic & Others
Ittrain93 0.387 0.366 0.395 0.384 0.571 0.454 0.387
retained 0.710 0.658 0.732 0.703 0.790 0.660 0.716
retained & IT trained
0.804 0.735 0.824 0.800 0.840 0.778a
retained & not IT trained 0.651 0.604 0.672 0.649 0.725 0.618a
work out of IT 0.232 0.247 0.225 0.236 0.155 0.316 0.237
no work 99 0.059 0.095 0.043 0.061 0.055 0.025 0.047
Unemployed 99
0.012
0.011
0.013 0.012
0.006 0.018 0.015
out of labor force lf99 0.046 0.084 0.031 0 0.049 0.049 0.007 0.032
n 1,058,989 314,564 744,425 8 885,600 97,688 44,914 30,786
% of sample 100% 29.7% 70.3% 8 83.6% 9.2% 4.2% 2.9%
Compared to Engineering
Retention in IT is higher (71% vs. 66%) Higher for women (66% vs. 52%) Higher for African Americans (66% vs. 54%) Conclude—as does Preston—that retention is
a major issue
Table III. Weighted means for individuals employed in engineering occupations in 1993 and in SESTAT in 1999.
All Females Males Whites Asians African Americans
Hispanic & Others
engtrain93 1.000 0.387 1.000 1.000 1.000 1.000 1.000 1.000
retained 0.659 0.521 0.673 0.656 0.705 0.544 0.694
wknoeng99 0.282 0.340 0.276 0.283 0.248 0.396 0.254
nowork99 0.059 0.139 0.051 0.061 0.048 0.060 0.052
unempl99
0.012
0.010
0.012 0.011
0.016 0.019 0.018
outlf99 0.047 0.129 0.039 0 0.048 0.032 0.041 0.035
n 1,159,923 108,188 1,051,735 8 977,662 111,061 32,665 38,535
% of sample 100.0% 9.3% 90.7% 8 84.3% 9.6% 2.8% 3.3%
What Do the IT trained do when they leave IT?
Top and mid-level managers (32.4%) Electrical and Electronic Engineering (9.2%) Accountants (7.2%) Other Management (6.4%) Other Administrative (4.0%) They also leave the labor force…especially
true of women (8% for women vs. 3% for men)
Retention Analysis
Look at those in IT occupation in 1993 (trained and untrained)
Determine IT workforce status in 1999– In IT– In another occupation– Not working (unemployed or out of labor force)
Estimate a multinomial logit model
Right hand side variables
Training variables Family status variables Change in family status variables Citizenship status and change in citizenship status Age Self employment Race/ethnicity Gender
Findings: Staying in IT vs. Moving to non-IT occupation
Positively related to whether IT is latest degree;
Negatively related to whether self-employed; had taken additional training in a non-IT field and African American.
Note: “female” is not significant
Findings: Working in IT vs. Not Working
Negatively related to being self employed and being female and, for women, whether one began parenting a child under six during the interval.
Findings: Working Not in IT vs. Not Working
Positively related to being African American Negatively related to being female and, for
women, beginning to parent a child under six during the interval.
Summarize
African Americans leave IT occupations for other occupations; do not leave the labor force or become unemployed.
Women leave IT occupations to leave the labor force or become unemployed, not to move into another occupation
Results consistent with Xie & Shauman: No evidence that marriage per se affects the retention of women IT workers; but the arrival of young children makes women less likely to remain in the labor force.
Do African American Women Respond the Same as White Women and/or African American Men?
Interact variable female and African American Find: African American women are
significantly more likely to remain in the labor force than are white females.
Cannot reject hypothesis that African American women are any more or less likely to leave IT for another job than African American men
Re-estimate, splitting the sample by training
Find that change in visa status is related to leaving IT for another occupation for the “non-trained.”
Suggests that IT occupations are used as an entrée to getting an H-1B visa.
Change in visa status does not affect probability of retention for those trained in IT.
Gender Effects
In both trained and un-trained samples, the “female” result holds
The “female-get children” result only holds for those without formal training.
African American results become more fragile—related to “thinness” of sample
Policy Implications
Policies directed towards retention will have differential outcomes depending upon group in question
Women would be likely to respond to initiatives that provide on-site child care.
African Americans more likely to respond to initiatives that make IT occupations more attractive relative to non-IT jobs.
Usual Caveats
Data “thin” for URM; especially when split by gender.
Data does not include certain groups working in IT.
Results may be clouded by strong labor market for IT workers in late 1990s.
Labor force patterns are fluid; some of those who have left will return