1. Do Job
Networks
Disadvantage
Women?
BKM
Motivation
Experiment Do Job Networks Disadvantage Women?
Set-up
Main Result
Evidence from a recruitment experiment in
Theory
Network
Malawi
structure
Connections
Heterogeneous
Networks
Lori Beaman, Niall Keleher, and Jeremy Magruder
Social
Incentives
Screening Northwestern, IPA, and UC-Berkeley
Screening
Either Gender
versus Restricted
Conclusions
November, 2012
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
2. Do Job
Networks
Disadvantage
Women?
Motivation
BKM
Motivation • In Malawi, as in much of the world, women are
Experiment disadvantaged in labor markets
Set-up
Main Result
• underrepresented in the formal sector
Theory
• earn less
Network
structure
Connections
• There are a litany of possible explanations, e.g.
Heterogeneous
Networks • taste-based or statistical discrimination
Social • differences in baseline human capital
Incentives
• differences in preferences
Screening
Screening • differences in tenure/experience profiles
Either Gender
versus Restricted • and so on
Conclusions
• Current policy interventions focus on closing the gender
Bonus Slides
Comment 1 gap in educational attainment
Comment 2
Comment
Comment
3
4
• Question: will that be enough?
Comment 5
3. Do Job
Networks
Disadvantage
Women?
What about hiring processes?
BKM
• Much less research on whether the hiring process causes
Motivation (dis)advantages
Experiment
Set-up
• About half of jobs are found through networks
Main Result
• In developing countries, networks are key for risk sharing,
Theory
credit in addition to labor market access
Network
structure • Several advantages for employers
Connections
Heterogeneous
Networks • relatively costless way to circulate info
Social • (some) workers may have useful screening info about
Incentives
friends and relatives (Montgomery 1991, Beaman and
Screening
Screening
Magruder 2012)
Either Gender • tied contracts between reference and referral may solve
versus Restricted
Conclusions
moral hazard problems (Heath 2012)
Bonus Slides • But do they disadvantage groups?
Comment 1
Comment 2
Comment 3 • Calvo-Armengol and Jackson (2004): the use of networks
Comment
Comment
4
5
can lead to disadvantages between groups
• Are women one of these groups?
4. Do Job
Networks
Disadvantage
Women?
Women and Networks
BKM
• Priors are not so clear - potential advantages and
Motivation
disadvantages.
Experiment
Set-up • Could help women, if e.g. resume characteristics are scarce
Main Result
Theory
and hard-to-observe characteristics are more important
• Or, could leave women out: sociologists emphasize
Network
structure gender-homophily in networks
Connections
Heterogeneous
Networks
• necessary condition for Calvo-Armengol and Jackson
Social
(2004) mechanism
Incentives
• However: as a stylized fact from observational data,
Screening
Screening women are less likely to get networked jobs
Either Gender
versus Restricted
• In U.S. unemployed women are less likely to report using
Conclusions
Bonus Slides
their friends and relatives for help in search (27% of men
Comment 1 vs. 20% of women) (Ioannides and Loury,2004)
Comment 2
Comment 3
Comment 4
• Based on observational data - could be confounded by
Comment 5
differences in occupations, reporting choices, etc.
5. Do Job
Networks
Disadvantage
Women?
Why would networks leave women out?
BKM
• It may be more costly for firms to access female referrals
Motivation
• Men (or women) may not be connected to (high quality)
Experiment
Set-up women
Main Result
Theory • Sociology lit: women’s networks may be less organized
Network around work (e.g. Smith 2000)
structure
Connections • Or men (or women) may have those connections, but
Heterogeneous
Networks
prefer to refer men
Social
Incentives • If it is easier to get (high quality) male referrals than
Screening female referrals because of network characteristics, then
Screening
Either Gender cost-minimizing firms may end up hiring more men
versus Restricted
through referrals
Conclusions
Bonus Slides • Firms may get more out of using referral systems for male
Comment 1
Comment 2 hires
Comment 3
Comment
Comment
4
5
• References may be better able or more willing to screen
men
6. Do Job
Networks
Disadvantage
Women?
Our experiment
BKM
• We conducted a recruitment experiment as part of a hiring
Motivation drive for enumerators in Malawi
Experiment
Set-up
• Survey firm wanted to hire more women
Main Result
Theory
• Two waves: people encouraged to apply themselves and
Network people then asked to make a referral
structure
Connections • All applicants complete skills assessment
Heterogeneous
Networks
• Competitive job between genders:
Social
Incentives • 38% of people who apply themselves are women and
Screening perform similarly to men
Screening
Either Gender • One type of position, so differences in occupational sorting
versus Restricted
cannot affect results. Reporting clear, too.
Conclusions
Bonus Slides • Referral phase randomized whether applicants could refer
Comment 1
Comment 2 only men, only women, or anyone, and also terms of
Comment 3
Comment 4 contract
Comment 5
• Fixed finders fees or performance incentive
7. Do Job
Networks
Disadvantage
Women?
Preview
BKM • The use of referral systems disadvantages highly skilled
Motivation women
Experiment • Only 30% of referrals (versus 38% of applicants) are
Set-up
Main Result women when people have a choice
Theory • 2 reasons: men systematically refer men
Network • Women’s referrals (both men and women) are less likely to
structure
Connections
qualify
Heterogeneous
Networks • However, when we restrict gender choices, men and
Social
Incentives
women make references at the same rate under all
Screening
contracts regardless of which gender they must refer
Screening
Either Gender
• Men and women are connected to suitable men and women
versus Restricted
Conclusions
• We develop and test a model to find out which
Bonus Slides characteristics of networks lead to disadvantages
Comment 1
Comment 2
• Social incentives rather than productivity differences lead
Comment
Comment
3
4
to disadvantages
Comment 5 • Screening potential of networks is maximized when men
refer men
8. Do Job
Networks
Disadvantage
Women?
Outline of rest of talk
BKM
Motivation
Experiment
Set-up
Main Result 1 Describe experimental design
Theory
2 Test whether women are (dis)advantaged by referral
Network
structure systems
Connections
Heterogeneous
Networks
3 Discuss a model of optimal referral choices under different
Social network characteristics
Incentives
Screening
4 Test whether men and women are connected to suitable
Screening
Either Gender
women
versus Restricted
Conclusions
5 Test for gender differences in network characteristics
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
9. Do Job
Networks
Disadvantage
Women?
Our Experiment
BKM
Motivation
• IPA-Malawi regularly hires a large number of enumerators
Experiment for several projects
Set-up
Main Result • We posted fliers indicating a hiring drive at a number of
Theory visible places in Lilongwe and Blantyre
Network
structure • Applicants were instructed to appear at a local
Connections
Heterogeneous
employment center at a specific date and time, with a
Networks
Social
resume.
Incentives
• Upon arrival, applicants given an id card and resumes
Screening
Screening
collected
Either Gender
versus Restricted • Applicants completed a written test
Conclusions
• Several math problems, ravens matrices, English skills
Bonus Slides
Comment 1 assessment, job comprehension component, computer skills
Comment 2
Comment 3 assessment
Comment
Comment
4
5
• 2 similar versions of test to limit cheating
10. Do Job
Networks
Disadvantage
Women?
Our Experiment (2)
BKM
• Following, applicants completed a practical skills
Motivation assessment
Experiment
Set-up
• IPA enumerators act as survey respondents, applicants act
Main Result
as enumerators
Theory
• To test for hard-to-observe abilities, we made a number of
Network
structure incorrect answers to questions - i.e. inconsistent household
Connections
Heterogeneous
Networks
size, implausible values for household acreage
Social • Actors instructed to give the right answer if the applicants
Incentives
press them
Screening
Screening
• 2 versions of incorrect answers
Either Gender
versus Restricted
• We measure the number of traps that the applicants
Conclusions caught
Bonus Slides • Total score on all components averaged. Applicants
Comment 1
Comment
Comment
2
3
informed of qualification threshold.
Comment 4
Comment 5 • Qualified individuals called for enumerator positions as
positions open
11. Do Job
Networks
Disadvantage
Women?
CA men and women are
BKM competitive
Motivation
Experiment
Figure 1: CA Ability by Gender
.03
Set-up
Main Result
Theory
kernel density estimate
Network
structure
.02
Connections
Heterogeneous
Networks
Social
Incentives
.01
Screening
Screening
Either Gender
versus Restricted
Conclusions
0
Bonus Slides
Comment 1 20 40 60 80 100
Comment 2 CA's overall (corrected) score
Comment 3
Comment 4 Male CAs Female CAs
Comment 5
12. Do Job
Networks
Disadvantage
Women?
Experiment: Referral Rounds
BKM
Motivation
• Finally, applicants asked to make a referral
Experiment • Randomly assigned to one of following treatments:
Set-up
Main Result • Asked at random to make a referral who was male, a
Theory referral who was female, or a referral who could be male or
Network female
structure
Connections
Heterogeneous
• Cross-randomized the finder’s fee:
Networks
Social
• A fixed fee of either 1000 MWK or 1500 MWK ($6 or
Incentives $10).
Screening • A performance incentive of 500 MWK if their referral does
Screening
Either Gender not qualify or 1800 MWK if their referral does qualify
versus Restricted
Conclusions • All treatments fully blinded from the perspective of
Bonus Slides evaluators
Comment 1
Comment
Comment
2
3
• Referrals attend recruitment session 3 or 4 days later.
Comment 4
Comment 5 Complete same skills assessment.
13. Do Job
Networks
Disadvantage
Women?
Do Referral Systems disadvantage
BKM women?
Motivation
Table 1: Gender Distributions of CAs and Referrals
Experiment
Set-up (1) (2) (3) (4)
Main Result
Female Diff: p
Theory All CAs Male CAs
CAs value
Network A. CA Characteristics
structure
Connections
Fraction of CAs 100% 62% 38%
Heterogeneous CA is qualified 53% 56% 48% 0.047
Networks
N 767 480 287
Social
B. CA Characteristics: Made Referral, Either Gender Treatments
Incentives
Fraction of CAs 100% 61% 39%
Screening
CA is qualified 57% 62% 49% 0.061
Screening
Either Gender N 217 130 87
versus Restricted
C. Referral Characteristics: Either Gender Treatments
Conclusions Referral is Female 30% 23% 43% 0.002
Bonus Slides Referral is Qualified 49% 56% 38% 0.019
Comment 1 Referral is Qualified Male 34% 43% 22% 0.002
Comment 2
Comment 3
Referral is Qualified Female 14% 13% 17% 0.456
Comment 4 N 195 117 78
Comment 5
14. Do Job
Networks
Disadvantage
Women?
A simple model
BKM
Motivation
• Suppose conventional applicants (CAs) each know a
Experiment
Set-up collection of potential referrals, some men and women.
Main Result
Theory
• Each of these potential referrals has a social transfer they
Network will give the applicant
structure
Connections • Each also has an actual quality and an observed expected
Heterogeneous
Networks quality
Social
Incentives • Focus on individuals the CA might actually choose:
Screening
Screening
• For each perceived probability of qualifying, the person
Either Gender
versus Restricted
who maximizes social payments
Conclusions
• Therefore expected quality is decreasing in social payments
Bonus Slides • Observe referral choice under two contact types: fixed fee
Comment 1
Comment
Comment
2
3
and performance pay
Comment 4
Comment 5
15. Do Job
Networks
Disadvantage
Women?
Today, graphically
BKM sample similar networks
20
Motivation
Experiment
15
Set-up
Main Result
Theory
10
Network
structure
Social Payment
Connections
Heterogeneous
Networks 5
Social
Incentives
0
Screening
Screening
Either Gender
versus Restricted
-5
Conclusions
Bonus Slides
Comment 1
-10
Comment 2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Comment 3 Perceived Probability of Qualifying
Comment 4
Comment 5
Note: Diamonds: women, Circles: men
16. Do Job
Networks
Disadvantage
Women?
What we observe
BKM
Motivation
Experiment
Set-up • Whether someone chooses to make a referral
Main Result
Theory • For those who make a referral, we see 2 nodes in the
Network
structure
gender-specific network for each gender:
Connections
Heterogeneous
Networks • Characteristics of person who maximizes social incentives
Social • Characteristics of person who maximizes expected pay
Incentives
Screening
+social incentives under performance incentive contract
Screening
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
17. Do Job
Networks
Disadvantage
Women?
Are men and women connected?
BKM
• One reason women may be disadvantaged by referral
Motivation system is if (suitable) women are not integrated into men’s
Experiment
Set-up
networks
Main Result
• Men and women make a decision to make a referral if the
Theory
Network
expected payoffs are greater than 0
structure
Connections
• Under fixed fees, this means that they know a man or
Heterogeneous
Networks woman whose social payment is not too negative
Social • Under perf pay, this means that they know a man or
Incentives
woman whose total package of fixed fees + expected perf
Screening pay
Screening
Either Gender
versus Restricted • A stronger question: Are there men who only know
Conclusions
suitable women?
Bonus Slides
Comment 1 • Are men in “either” treatments more likely to return with
Comment 2
Comment 3 a referral than men in “male” treatments?
Comment 4
Comment 5
• [Later] does screening behavior look different in “either”
treatments versus restricted male referrals?
18. Do Job
Networks
Disadvantage
Women?
Are men less likely to know
BKM suitable women?
Motivation
Experiment Table 2: Probability of Making a Referral
Set-up
(1) (2) (3) (4)
Main Result
Theory Female Treatment ‐0.004 ‐0.055 ‐0.004 ‐0.042
Network (0.038) (0.054) (0.050) (0.074)
structure Either Gender Treatment 0.014 0.017 ‐0.052 ‐0.024
Connections (0.040) (0.055) (0.052) (0.071)
Heterogeneous Performance Pay ‐0.148 *** ‐0.113
Networks
(0.056) (0.080)
Social
Perf Pay * Female Treatment 0.004 ‐0.013
Incentives
(0.076) (0.111)
Screening Perf Pay * Either Treatment 0.152 * 0.086
Screening (0.079) (0.110)
Either Gender
versus Restricted
Observations 506 310 506 310
Conclusions
CA Gender Men Women Men Women
Bonus Slides Notes
Comment 1 1 The dependent variable is an indicator for whether the CA makes a referral.
Comment 2 2 All specifications include CA visit day dummies.
Comment 3
Comment 4
Comment 5
19. Do Job
Networks
Disadvantage
Women?
How would different network
BKM characteristics affect referral
Motivation choices
Experiment
Set-up
Main Result We identify four dimensions of heterogeneity:
Theory
Network 1 Maximal Social payment received: “Closest gender”
structure
Connections
Heterogeneous 2 Expected quality of closest person: “Quality”
Networks
Social
Incentives
3 Slope of social payment/expected quality tradeoff:
Screening “Network Shallowness”
Screening
Either Gender
versus Restricted 4 Variance of actual quality relative to expected quality:
Conclusions “Information”
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
20. Do Job
Networks
Disadvantage
Women?
Similar Networks
BKM sample similar networks
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
21. Do Job
Networks
Disadvantage
Women?
Closer men
BKM higher male social payments
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
22. Do Job
Networks
Disadvantage
Women?
Similar Networks
BKM sample similar networks
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
23. Do Job
Networks
Disadvantage
Women?
Higher quality men
BKM higher male quality
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
24. Do Job
Networks
Disadvantage
Women?
Similar Networks
BKM sample similar networks
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
25. Do Job
Networks
Disadvantage
Women?
Shallower women
BKM Shallower female network
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
26. Do Job
Networks
Disadvantage
Women?
Similar Networks
BKM sample similar networks
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
27. Do Job
Networks
Disadvantage
Women?
Worse information about women
BKM less information about women
20
Motivation
Experiment
Set-up 15
Main Result
Theory
Network 10
structure
Connections
Social Payment
Heterogeneous
Networks
5
Social
Incentives
Screening
Screening 0
Either Gender
versus Restricted
Conclusions
-5
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4 -10
Comment 5 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Perceived Probability of Qualifying
28. Do Job
Networks
Disadvantage
Women?
Model predictions
BKM
Motivation
Experiment 1 Under fixed fees: only differences in closeness affect which
Set-up
Main Result referral is chosen
Theory
Network
2 Higher quality increases returns under performance pay
structure
Connections • Quality (of person who gives highest social payment) is
Heterogeneous
Networks revealed under fixed fees
Social
Incentives
3 Worse info, more shallow networks can both lead to lower
Screening
Screening response to performance pay
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
29. Do Job
Networks
Disadvantage
Women?
Closest gender, quality & social
BKM incentives
Motivation
Experiment Men may know women, but would they share opportunities?
Set-up
Main Result
Theory
• Prediction 1 from the model: The person referred under
Network
fixed fees is the closest person in the network
structure
Connections
Heterogeneous
• If men are closer to men (or women), should see men
Networks
referred systematically under fixed fee - unrestricted
Social
Incentives
treatment
Screening
Screening • The restricted-gender fixed fee treatments also let us
Either Gender
versus Restricted observe the quality of the closest people in the network:
Conclusions
• If men’s networks of men are higher quality than men’s
Bonus Slides
Comment 1 networks of women, should see fixed fee restricted male
Comment
Comment
2
3
referrals being higher quality than fixed fee restricted
Comment
Comment
4
5
female
30. Do Job
Networks
Disadvantage
Women?
Characteristics of closest referrals
BKM
Motivation
Experiment
Set-up
Main Result C. Referral Characteristics: Either Gender Treatments
Theory Referral is Female 30% 23% 43% 0.002
Referral is Qualified 49% 56% 38% 0.019
Network
structure Referral is Qualified Male 34% 43% 22% 0.002
Connections Referral is Qualified Female 14% 13% 17% 0.456
Heterogeneous
Networks N 195 117 78
Social D. Referral Characteristics: Either Gender, Fixed Fee Treatments
Incentives
Referral is Female 32% 25% 43% 0.042
Screening Referral is Qualified 50% 60% 37% 0.012
Screening
Either Gender Referral is Qualified Male 34% 44% 20% 0.007
versus Restricted
Referral is Qualified Female 16% 16% 16% 0.983
Conclusions N 117 68 49
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
31. Do Job
Networks
Disadvantage
Women?
Do men refer similar male and
BKM female network members?
Motivation
Figure 2: Men's Fixed Fee Referrals
Experiment .03
Set-up
Main Result
Theory
Kernel density estimate
Network
structure
.02
Connections
Heterogeneous
Networks
Social
Incentives
.01
Screening
Screening
Either Gender
versus Restricted
0
Conclusions
20 40 60 80 100
Bonus Slides Referral's overall (corrected) score
Comment 1
Comment 2 Men referring men Men referring women
Comment 3
Comment 4
Comment 5
Note: figure compares men in restricted treatments only
32. Do Job
Networks
Disadvantage
Women?
What about women’s referrals?
BKM
Motivation
Figure 3: Women's Fixed Fee Referrals
Experiment .03
Set-up
Main Result
Kernel density estimate
Theory
.02
Network
structure
Connections
Heterogeneous
Networks
.01
Social
Incentives
Screening
Screening
Either Gender
0
versus Restricted
20 40 60 80 100
Conclusions
Referral's overall (corrected) score
Bonus Slides
Women referring men Women referring women
Comment 1
Comment 2
Comment 3
Comment
Comment
4
5
Note: figure compares women in restricted treatments only
33. Do Job
Networks
Disadvantage
Women?
Summary so far
BKM
Motivation • By design, we only observe clean evidence of differences in
Experiment social incentives for men or women who maximize social
Set-up
Main Result incentives (who are revealed through the fixed fee
Theory
treatments)
Network
structure • For those people:
Connections
Heterogeneous
Networks
• Men tend to maximize men’s incentives
Social • Low ability people tend to maximize women’s incentives
Incentives
• Closest women are low ability
Screening
Screening
• Closest men however are not systematically low ability
Either Gender
versus Restricted
• Can conclude: at least among socially closest people, men
Conclusions
and women have different social incentives
Bonus Slides
Comment
Comment
1
2
• Social incentives make it cheaper to (a) get male referrals
Comment
Comment
3
4
from men and (b) use men to get high quality referrals
Comment 5
34. Do Job
Networks
Disadvantage
Women?
Is Women’s Disadvantage
BKM Productive?
Motivation
Experiment
Set-up
Main Result • If employers encourage referral hires, they likely gain
Theory
something from their use
Network
structure • One thing which has been emphasized is screening (e.g.
Connections
Heterogeneous Montgomery (1990), Beaman and Magruder (2012))
Networks
Social • If employees see hard to observe characteristics, can
Incentives
improve outcomes for employer
Screening
Screening
Either Gender
• If men (women) are less able to screen women, it may lead
versus Restricted
to employers discouraging female referrals
Conclusions
Bonus Slides
• From the model: CAs will screen if and only if they have
Comment
Comment
1
2
good information, and networks are not too shallow
Comment 3
Comment 4
Comment 5
35. Do Job
Networks
Disadvantage less information about women
20
Women?
BKM
Motivation 15
Experiment
Set-up
Main Result
10
Theory
Social Payment
Network
structure
Connections 5
Heterogeneous
Networks
Social
Incentives 0
Screening
Screening
Either Gender
versus Restricted -5
Conclusions
Bonus Slides
Comment 1 -10
Comment 2 0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Comment 3 Perceived Probability of Qualifying
Comment 4
Comment 5
36. Do Job
Networks
Disadvantage
Women?
Low info and screening
BKM
Motivation • Low info makes the tradeoffs “steeper” - it becomes more
Experiment
Set-up
expensive and more infeasible to find very high quality
Main Result referrals
Theory
Network • Essentially, most referral probabilities of qualification
structure
Connections
pushed towards 1/2
Heterogeneous
Networks
• this increases the payoffs to referring someone who you
Social
Incentives think is relatively low ability under perf pay incentives
Screening
Screening
• and decreases the payoffs to referring someone who you
Either Gender
versus Restricted
think is relatively high ability under perf pay
Conclusions
• Empirically, if men (women) have lower ability to screen
Bonus Slides
Comment 1 women, should observe a smaller increase in performance
Comment 2
Comment
Comment
3
4
in response to perf pay
Comment 5
37. Do Job
Networks
Disadvantage
Women?
BKM
Table 4: Referral Performance
Motivation Referral Qualifies
(1) (2) (3) (4)
Experiment Female Referral Treatment ‐0.030 ‐0.190 ** 0.068 ‐0.181
Set-up (0.062) (0.083) (0.081) (0.113)
Main Result
Either Gender Treatment 0.071 ‐0.231 *** 0.227 *** ‐0.242 **
Theory (0.066) (0.082) (0.084) (0.107)
Network Performance Pay 0.267 *** 0.021
structure (0.093) (0.122)
Connections Perf Pay * Female Treatment ‐0.248 * ‐0.022
Heterogeneous (0.127) (0.171)
Networks
Perf Pay * Either Treatment ‐0.383 *** 0.032
Social (0.132) (0.169)
Incentives
Observations 390 227 390 227
Screening CA Gender Men Women Men Women
Screening Notes
Either Gender 1 The dependent variable is an indicator for the referral qualifying.
versus Restricted
2 All specifications include CA visit day dummies.
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
38. Do Job
Networks
Disadvantage
Women?
Screening Results
BKM
Motivation
• Men can screen men
Experiment
Set-up • Men cannot screen women (or, at least, won’t at these
Main Result
Theory
levels of incentives)
Network • Allowing the option to refer women eliminates the
structure
Connections screening premium
Heterogeneous
Networks
• Suggests that employers who want to maximize screening
Social
Incentives
may discourage men from making female referrals.
Screening • Some evidence that difference is info and not shallowness:
Screening
Either Gender
versus Restricted
men are more likely to make a low quality referral under
Conclusions perf pay-female treatments than under fixed fee-female
Bonus Slides • Women show less ability to screen men or women overall
Comment 1
Comment 2
Comment 3 • Some not quite sig evidence that they can screen women
Comment 4
Comment 5
39. Do Job
Networks
Disadvantage
Women?
Screening with choice of gender
BKM Why is there a lower screening premium when we allow either
Motivation
gender?
Experiment • Under performance pay, one maximizes the sum of social
Set-up
Main Result incentives and expected perf incentive
Theory
• This makes the theoretical effect ambiguous
Network
structure
Connections
• Considering either gender in general allows you to “buy”
Heterogeneous
Networks quality with giving up a lower amount of social incentives
Social
Incentives
• Increases both chance that you observe someone who has
Screening
a high chance of qualifying and gives you OK social
Screening incentives
Either Gender
versus Restricted
⇒ May increase performance premium
Conclusions
Bonus Slides • Also ↑ chance that you observe someone who has an OK
Comment
Comment
1
2
chance of qualifying but gives you great social incentives
Comment 3
Comment 4 ⇒ May decrease performance premium
Comment 5
• Happens, in particular, when info is bad about one gender
40. Do Job
Networks
Disadvantage
Women?
What exactly is being screened?
BKM
Motivation • Have much richer data than is being used here - detailed
Experiment assessments of different referral characteristics
Set-up
Main Result
• Men are screening in some ways across a broad category of
Theory
characteristics
Network
structure • Women are screening, too -
Connections
Heterogeneous
Networks • significantly, women screen women on language scores and
Social cognitive skills.
Incentives
• Women screen men on survey experience
Screening
Screening
Either Gender
• The former is probably more valuable as screening for
versus Restricted
Conclusions
employers. May be a role for encouraging female referrals
Bonus Slides
of women
Comment 1
Comment 2
• still, if employers use referrals for screening, biggest returns
Comment
Comment
3
4
are to get men to refer men
Comment 5
41. Do Job
Networks
Disadvantage
Women?
BKM
Motivation Table 5: Screening of Male CAs on Different Characteristics
Survey Tertiary Math Language Ravens Computer Practical Feedback
Experiment exp Education Score Score score Score Exam Score points
(1) (2) (3) (4) (5) (6) (7) (8)
Set-up
Main Result Female Referral ‐0.033 0.045 ‐0.017 ‐0.115 ‐0.092 0.062 1.033 3.003 ***
Treatment (0.069) (0.074) (0.142) (0.207) (0.194) (0.371) (0.661) (1.044)
Theory Either Gender 0.040 0.072 0.009 0.087 0.089 0.623 1.378 ** 1.856 *
Treatment (0.072) (0.077) (0.148) (0.215) (0.203) (0.387) (0.689) (1.089)
Network Performance Pay 0.080 0.067 0.134 ‐0.005 0.230 0.943 ** 0.496 1.883
structure (0.080) (0.085) (0.164) (0.238) (0.224) (0.428) (0.757) (1.197)
Connections Perf Pay * Female ‐0.075 0.025 ‐0.259 ‐0.027 ‐0.293 ‐0.915 ‐0.950 ‐2.443
Heterogeneous Treatment (0.108) (0.116) (0.223) (0.325) (0.305) (0.583) (1.026) (1.622)
Networks
Perf Pay * Either ‐0.165 ‐0.083 ‐0.065 ‐0.169 ‐0.367 ‐0.856 ‐1.768 * ‐3.371 **
Social Treatment (0.113) (0.121) (0.232) (0.338) (0.318) (0.607) (1.069) (1.696)
Incentives Observations 386 390 390 390 390 390 383 382
Notes
1 The dependent variable is an indicator for the referral qualifying.
Screening
2 All specifications include CA visit day dummies.
Screening
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
42. Do Job
Networks
Disadvantage
Women?
BKM
Motivation Table 6: Screening of Female CAs on Different Characteristics
Tertiary Math Language Ravens Computer Practical Feedback
Survey exp
Experiment Education Score Score score Score Exam Score points
Set-up (1) (2) (3) (4) (5) (6) (7) (8)
Main Result Female Referral 0.032 0.151 ‐0.332 ‐1.140 *** ‐0.435 ‐0.627 0.972 2.152
Treatment (0.091) (0.110) (0.216) (0.342) (0.270) (0.538) (0.963) (1.349)
Theory Either Gender 0.040 0.017 ‐0.189 ‐0.246 ‐0.172 ‐0.139 0.015 0.879
Treatment (0.086) (0.104) (0.205) (0.324) (0.256) (0.509) (0.910) (1.274)
Network
Performance Pay 0.264 *** 0.143 ‐0.400 * ‐0.465 ‐0.175 0.419 1.832 * 1.604
structure (0.098) (0.119) (0.234) (0.370) (0.293) (0.582) (1.056) (1.479)
Connections
Perf Pay * Female ‐0.320 ** ‐0.292 * 0.402 1.330 ** 0.551 0.232 ‐2.164 ‐2.134
Heterogeneous Treatment (0.138) (0.166) (0.326) (0.515) (0.408) (0.811) (1.468) (2.055)
Networks
Perf Pay * Either ‐0.270 ** ‐0.052 0.368 0.500 ‐0.260 ‐0.372 ‐1.625 ‐4.511 **
Social Treatment (0.136) (0.164) (0.323) (0.510) (0.403) (0.802) (1.448) (2.027)
Incentives Observations 226 227 227 227 227 227 222 222
Notes
Screening 1 The dependent variable is indicated in the column heading.
2 All specifications include CA visit day dummies.
Screening
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
43. Do Job
Networks
Disadvantage
Women?
Conclusions
BKM
Motivation
Experiment • Stylized Fact: women are less likely to receive job referrals
Set-up
Main Result than men (from data in US and Europe)
Theory
Network • Using a recruitment experiment in Malawi, we confirm
structure
Connections
that women are disadvantaged by referral systems
Heterogeneous
Networks
• Men choose not to refer women, when given the choice
Social
Incentives
• Women choose women at about the population average,
Screening
Screening
but make on average low quality referrals
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
44. Do Job
Networks
Disadvantage
Women?
Conclusions: Economics
BKM
Motivation • We test several network constraints that could drive this
Experiment
Set-up
result
Main Result
Theory • Men and women are equally likely to be connected to men
Network and women
structure
Connections • Men are closest to men, but have high quality male and
Heterogeneous
Networks female contacts
Social
Incentives • Women are not socially closer to one gender than the
Screening
other, but have low quality networks of women
Screening
Either Gender • Men can screen men well, cannot screen women; women
versus Restricted
can screen both men and women to a lesser extent
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
45. Do Job
Networks
Disadvantage
Women?
Conclusions: Policy
BKM
Permitting women’s disadvantage in referral rates has three
Motivation
benefits to employers:
Experiment
Set-up
Main Result • It is lower cost for men to refer men than for men to refer
Theory
women (since social incentives are higher)
Network
structure
Connections
• It is lower cost to get high quality referrals if men are
Heterogeneous
Networks making referrals
Social
Incentives • Screening benefits of referral systems are maximized when
Screening
Screening
men are encouraged to refer only men
Either Gender
versus Restricted
• All in all, a hard problem to solve
Conclusions
Bonus Slides • Current policies to address gender gap - such as investing
Comment 1
Comment 2 in girls’ education - will not be enough to overcome this
Comment 3
Comment 4
• Maybe a role for quota systems in hiring policy
Comment 5
46. Do Job
Networks
Disadvantage
Women?
Comment on Attrition
BKM • 80% of applicants make a referral
Motivation • Reference rate is always really similar across genders,
Experiment different across treatments
Set-up
Main Result • Differences in referral quality across gender, within
Theory treatment can be taken at (close to) face value for those
Network who make referrals
structure
Connections
• Difference in referral quality across treatment will be the
Heterogeneous
Networks combined effect of some attrition + population average
Social choices
Incentives
• For employers (and to understand actual trends in
Screening
Screening references), the net effect (including attrition) is the
Either Gender
versus Restricted
relevant dimension in any event
Conclusions
• Implications for e.g. ability to screen are the same if
Bonus Slides
Comment 1
individuals attrit because they know their options are bad
Comment 2
Comment 3 • We also simulate the model and recover the same
Comment 4
Comment 5 predictions on the attrition decision and results within
made referrals
47. Do Job
Networks
Disadvantage
Women?
Can work experience explain
BKM results?
Motivation
Experiment
Set-up
Main Result
Theory
• Men are more likely to have worked at a survey firm in the
Network past than women
structure
Connections • Working at a survey firm may both enhance your network
Heterogeneous
Networks and give you better information
Social
Incentives • While it does not affect any of the interpretations - or
Screening disadvantages women face - it may be an underlying
Screening
Either Gender mechanism
versus Restricted
Conclusions • We find no differential response among people who have
Bonus Slides worked at a survey firm in the past.
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
48. Do Job
Networks
Disadvantage
Women?
Competition
BKM
Motivation • Niederle and Vesterlund (2007) find that women are
Experiment
Set-up
averse to competition relative to men
Main Result
• Making a reference involves introducing the employer to a
Theory
Network
potential competitor for the job
structure
Connections
• May have an incentive to refer someone bad (though, a
Heterogeneous
Networks marginal incentive for an informed decision maker -
Social referral is one additional applicant among many)
Incentives
Screening • May have been particularly salient in our context, as
Screening
Either Gender
applicants not yet hired
versus Restricted • However, certainly a relevant incentive in on-the-job
Conclusions referrals, too
Bonus Slides
Comment 1 • Again, suggests a mechanism, without affecting
Comment 2
Comment 3 interpretations or policy prescriptions
Comment 4
Comment 5
49. Do Job
Networks
Disadvantage
Women?
Cross-randomization
BKM
Motivation
Experiment • We cross-randomized a treatment designed to make the
Set-up
Main Result competitiveness more salient
Theory
• CAs were told the qualification threshold was either
Network
structure
Connections
1 Absolute: scoring better than 60
Heterogeneous
Networks
2 Relative: scoring in the top half of applicants
Social
Incentives
• We hypothesize that the relative treatment makes the
Screening competition more salient (since CAs compete directly with
Screening
Either Gender
referrals to be in the top half)
versus Restricted
Conclusions
• (admittedly, somewhat weak test)
Bonus Slides • Look just at fixed fee referrals to isolate social incentives
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
50. Do Job
Networks
Disadvantage
Women?
BKM
Appendix Table 3: Competition incentives among fixed fee referrals
Motivation
(1) (2) (3) (4) (5) (6)
Experiment CA Referral Referral CA Referral Referral
Dependent Variable Qualifies Qualifies Qualifies Qualifies Qualifies Qualifies
Set-up
Main Result Competitive Treatment 0.021 0.072 0.052 0.014 0.090 0.227
(0.062) (0.069) (0.121) (0.086) (0.095) (0.165)
Theory Female Treatment 0.094 ‐0.024
Network (0.116) (0.177)
structure Either Treatment 0.175 ‐0.160
Connections (0.123) (0.169)
Heterogeneous Competitive * Female 0.007 ‐0.263
Networks
(0.166) (0.236)
Social Competitive * Either 0.103 ‐0.142
Incentives 0.176 (0.236)
Screening
Observations 287 232 232 166 133 133
Screening
CA Gender Men Men Men Women Women Women
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
51. Do Job
Networks
Disadvantage
Women?
BKM
Figure 2: Gender choice in referrals, by CA performance
.8
Motivation
Experiment
Set-up
.6
Main Result
Referral is Female
Theory
Network
.4
structure
Connections
Heterogeneous
Networks
.2
Social
Incentives
0
Screening
Screening 20 40 60 80 100
Either Gender CA's overall (corrected) score
versus Restricted
Referrals of Male CAs Referrals of Female CAs
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
52. Do Job
Networks
Disadvantage
Women?
BKM
Figure 3: Referral qualification rate, by CA performance
1
Motivation
Experiment
Referral's qualification rate
Set-up
.8
Main Result
Theory
.6
Network
structure
Connections
.4
Heterogeneous
Networks
Social
Incentives
.2
Screening
Screening 20 40 60 80 100
Either Gender CA's overall (corrected) score
versus Restricted
Referrals of Male CAs Referrals of Female CAs
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
53. Do Job
Networks
Disadvantage
Women?
BKM
Figure 6: Referral Qualifies , by Male CA performance
.8
Motivation
Experiment
Set-up
.6
Referral qualifies
Main Result
Theory
.4
Network
structure
Connections
.2
Heterogeneous
Networks
Social
Incentives
0
Screening 20 40 60 80 100
Screening CA's overall (corrected) score
Either Gender
versus Restricted Men referring women, fixed Men referring men, fixed
Men referring women, perf Men referring men, perf
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
54. Do Job
Networks
Disadvantage
Women?
BKM
Figure 7: Referral Qualifies , by Female CA performance
Motivation
Experiment .8
Set-up
Referral qualifies
Main Result
.6
Theory
Network
.4
structure
Connections
Heterogeneous
.2
Networks
Social
Incentives
0
Screening 20 40 60 80 100
Screening CA's overall (corrected) score
Either Gender
versus Restricted Women referring women, fixed Women referring men, fixed
Women referring women, perf Women referring men, perf
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
55. Do Job
Networks
Disadvantage
Women?
Social Payments and Qualification
BKM
Motivation • Possible (reasonable?) that social payments increase with
Experiment qualification in the ambient network
Set-up
Main Result
• Referrals give you better social transfers if they get the job
Theory
Network • Consistent with our modelling assumptions
structure
Connections • No assumption made about the joint distribution of
αg , Qjg in the ambient network
Heterogeneous
Networks
j
Social • Selection rule still leads to decreasing relationship among
Incentives
Screening
non-dominated referrals
Screening
Either Gender • However, may change interpretation of social payments
versus Restricted
Conclusions • Incentives aligned with employer
Bonus Slides • differences in quality expectations may lead to women’s
Comment 1
Comment 2
disadvantage if men expect men to be higher quality,
Comment
Comment
3
4
women have wrong quality expectations
Comment 5
56. Do Job
Networks
Disadvantage
Women?
Unbiased Expectations of Quality
BKM
Motivation
Experiment
Set-up
• Model assumed εg was mean 0 - allowed us to estimate Q1
j
g
Main Result
Theory
• Already showed that men’s fixed fee referrals of men ARE
Network
NOT higher ability than men’s fixed fee referrals of women
structure
Connections • And women’s (low quality) fixed fee referrals ARE NOT
Heterogeneous
Networks the highest quality people they know (they know high
Social quality men)
Incentives
Screening • So, if CA’s have unbiased expectations: can conclude that
Screening
Either Gender expectations of quality ARE NOT source of women’s
versus Restricted
Conclusions
disadvantage
Bonus Slides • But, expectations of quality could be biased
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5
57. Do Job
Networks
Disadvantage
Women?
Biased Expectations of Quality
BKM
• If expected social incentives increase in expected referral
Motivation qualification and expectations are biased (for now, against
Experiment women)
Set-up
Main Result • Incentives to refer a qualified person are still strictly larger
Theory
under perf
Network
structure • Would expect to see even more men referred under perf
Connections
Heterogeneous (We don’t)
Networks
• Would expect to see men restricted to refer women attrit
Social
Incentives more under perf (We don’t)
Screening
Screening
• Moreover, some evidence that social incentives are not
Either Gender
versus Restricted strongly correlated with expected referral performance
Conclusions • Men referring other men are choosing not to refer the best
Bonus Slides men they know under fixed
Comment 1
Comment 2 • Men do respond to incentives
Comment 3
Comment
Comment
4
5
• Similar argument holds for women referring low ability
people.
58. Do Job
Networks
Disadvantage
Women?
Selection rule even with positive relationship
BKM
Motivation
Experiment
Set-up
Main Result
Theory
Network
structure
Connections
Heterogeneous
Networks
Social
Incentives
Screening
Screening
Either Gender
versus Restricted
Conclusions
Bonus Slides
Comment 1
Comment 2
Comment 3
Comment 4
Comment 5