Slides prepared for a presentation at Centre for Advanced Research in Social Science (CARSS) at University of Dhaka

1. Challenges of change: Training women to
manage in the Bangladeshi garment sector
Atonu Rabbani, Department of Economics, University of Dhaka
Joint work with Rocco Macchiavello (LSE), Andreas Menzel (CERGE) and
Christopher Woodruff (Oxford)
Centre for Advanced Research in Social Science (CARSS)
University of Dhaka, September 24, 2017

2. Why are women so under-represented
in management positions?
Or as leaders in general?

3. Broad Question
• About 65-70% of the workers in Bangladeshi RMG sector
are women, ~80% in production lines
• Less than 5% of the supervisors are women, ~0% in the
higher levels
• Why?

4. Managing people is important

5. Management is
important
Data suggests a long
tail of “bad” firms at the
left tail of the
distribution in the
developing countries.
one quarter of cross-
country productivity
differences can be
explained by
differences in
management
practices.
- Bloom , Sadun and
Van Reenen (2015)
y = -0.1753x + 2.287
R² = 0.7647
0
0.1
0.2
0.3
0.4
0.5
0.6
0.7
0.8
0.9
1
7 8 9 10 11 12
ProportionofFirmsbelow
GlobalMedian
Log (GDP per employed Labor)

6. Moreover…
There is a huge
potential to gain in
terms of productivity
within firms as well.
Note: Based on 42,000
line-days worth of data
from 7 factories.
Source: Machiavello,
Rabbani and Woodruff
(2015)
75th percentile line is more than 1.5 times more
productive compared to the 25th percentile line!

7. Context and Questions

8. Addressing some of these issues at RMG Sector in
Bangladesh
• Very large: about 4m people involved in the sector
• Moreover, 65% is female
• At the sewing lines: 70-80%
• However, less than 5% of the supervisors are women
• Why?
• Marginal female supervisor is worse than male supervisor
• OR, NOT, suggesting managerial talent is misallocated
• Worse or not, women are perceived as being worse in supervisory
roles
• Not trying them as supervisors can contribute to persistence in such
perceptions

9. Firm size distribution is a little peculiar here.

10. Supervisor Quality is Declining
Supervisors’ education falling
3Source: Surveys of 559 supervisors and 22.7 million males in the census (moving average shown)
4
5
6
7
8
9
10
11
38 37 36 35 34 33 32 31 30 29 28 27 26 25 24 23 22
Yearsofeducation(movingaverage)
Worker's age in 2015
Average education by worker's age in 2015
male population
male supervisors
• Perhaps a natural outcome
in a growing economy
• Dearth of qualified of
supervisors is often a
crying rally of the local
management.

11. Lower management positions can be fulfilled by women
operators.
Supervisors’ education falling
4
Source: Surveys of 2155 female and 849 male operators in 150 factories;
Bangladesh 2011 population census.
4
5
6
7
8
9
10
11
38 37 36 35 34 33 32 31 30 29 28 27 26 25 24 23 22
Yearsofeducation(movingaverage)
Worker's age in 2015
Average education by worker's age in 2015
female operators
male operators
• SVs are typically ex-line
operators.
• Recent cohort of LOs are
equally educated on average.
• Marginal promotees may be
as good, but how do we know
that?

12. Our understandings so far.
A summary of the results.

13. Constraints to promote women in SV roles
• Three:
• Self-confidence and beliefs about self
• Beliefs and resistance from other workers
• Uncertainty from management
• Once we overcome these, lines with (marginal/newly
appointed) female supervisors are almost equally
productive as the (marginal/newly appointed) male
supervisors
• Overcoming these constraints are not without costs to the
factories (suggesting trade-offs)

14. Design

15. Design and Data Collection
• Just observational data will not be useful: too few women
supervisors and endogenous assignments
• So we work with 24 suppliers of a single large foreign
buyer
• Factories nominated operators for a GIZ-designed six-
week offsite training on production, quality and
HR/compliance
• Factories selected 96 men and 121 women for training –
72 men and 73 women finally completed the training

16. Randomization
• Factories were randomly allocated into either rounds 1 and 3 or
rounds 2 and 4
• Trainees from each factory were randomized either 1 or 3 and
similarly either 2 or 4
• Then trainees were randomly assigned among the trial lines
• We collected data…
• …before the trainings started (Findings 1 and 2): skill assessments,
perceptions
• …right after the six-eight week of trials and a further two-four month
follow-ups (Findings 3): production data
• …further follow-ups (Findings 4): “production management” games

17. Findings

18. Finding #1
• In broadly defined eight sets of tasks, women are
perceived as weaker supervisors
• Consistent across the whole managerial hierarchy,
including the operators
• Women are considered less able to understand
machines and operations
• These perceptions are slightly weaker among the
female operators and those who have worked under
female supervisors

19. PerceivedAbility by
Gender in
Supervisory Roles
Negative means
men are perceived
as better
supervisors
Most people think
so.
And across almost
all tasks.
-1.00
-0.80
-0.60
-0.40
-0.20
0.00
0.20
0.40
HR Managers Production Managers Line Chiefs
Line Supervisors Line Operators

20. PerceivedAbility
cont.
Female trainees have
more favorable views
followed by female
operators.
Average male
operators have the
least favorable views.
Exposure to female
supervisors mitigate
some of the bias.
-1.00
-0.80
-0.60
-0.40
-0.20
0.00
0.20
Female Operators Male Operators LO's worked w/ female SV
Female trainees Male trainees

21. Finding #2
• Before the training, using extensive skill
assessments, we find
• Female and male trainees have similar technical
knowledge of machines and operations contrary to
widely-held beliefs
• However, women are chosen less in simulated
management games
• Women themselves rate themselves lower in the same
eight broad sets of tasks

22. How accurate are the perceptions: Leadership?
• We conducted a leadership
exercise in which groups of
trainees were given 5 minutes to
organize themselves into a
production team – LC, LSV,
operators. Two outcomes:
• In mixed-gender groups, men are
more likely to be the LCs / LSVs
• Enumerators observed
discussion and recorded who
spoke, etc. No sig. difference
0%
5%
10%
15%
20%
25%
30%
35%
40%
45%
Male Female Male Female
Mixed-Gender Single-Gender

23. Self-Confidence
Before training: Self confidence
-1
-0.5
0
0.5
1
Teaching
Giving orders
Understanding
machines
Motivating operators
Communicating with
operators
Correcting mistakes
Communicating with
management
Organizing resources
Trainees Before Training: Female self-assessment is significantly
lower for all skills except "Teaching" and "Motivating operators"
male
female
same as typical supervisor
After training: Self confidence
-1
-0.5
0
0.5
1
Teaching
Giving orders
Understanding
machines
Motivating operators
Communicating with
operators
Correcting mistakes
Communicating with
management
Organizing resources
Trainees Just After Trial: No significant difference between female
and male self-assessment except for "Correcting Mistakes"
male
female
same as typical supervisor

24. Finding #3
• Using detailed line level productivity data, we look
at efficiency, quality defects and absenteeism
• Immediately after the training, initially the female
supervisor perform worse
• This gap disappears after few months at the line

25. Finding #3
Efficiency Defects
-0.0093
0.0236
0.0176
0.0196
Female trainee Male trainee** Female trainee Male trainee
Trial period Post-trial
-0.002
-0.0042
-0.0055
-0.0013
Female trainee Male trainee Female
trainee*
Male trainee
Trial period Post-trial

26. Finding #4: Resistance from the Workers
• The experiment also allows us to understand change in attitude
towards female line supervisor
• Female trainees are ranked almost a point lower
• Male operators rank the female trainees lower than female operators (-1.39 as
oppose to -0.65)
• Male operators also hold more pessimistic view toward prospect of promotion
to supervisory roles in future and report leaving the factory sooner
• In simulated management exercises, the female trainees outperform the male
counterparts (by 0.29 SD higher payoffs), but NOT when a male operator is in
the group
• The promoted trainees do even better
• Female leaders are ranked lower and less able to identify the “correct”
strategies

27. Higher-level management internalizes some of these
concerns (and some more)
• Managers selecting the participants for training express more
confidence in their ability to select the right males.
• However, they are less certain that they are selecting the right
females.
• Aided by higher drop-out rates among the female trainees.
• We also find a big variance in the proportion of female trainees
that are promoted across factories.
• Many promote all the female trainees. Some promote none of them.
• This may come either (a) from the greater difficulty selecting the right
females, or (b) from beliefs among key decision-makers that females
will not be effective supervisors.

28. Conclusions?

29. All in all
• Nominated (marginal) female trainees may be as good as the
nominated (marginal) male trainees.
• Both the factories and even the trainees themselves seem to
think they are not capable.
• Signaling may be difficult as women are not tried in supervisory
positions, a real challenge to change the existing equilibrium
• This perception subsides once they are tried in actual
production.
• Along with initial workers’ resistance and management’s
reluctance, the situation persist overtime leading to
• less than optimal management practices and
• resulting in lesser women representation in managerial positions

30. Thanks for listening