Together We Will: Evidence from a Field Experiment on Female Voter Turnout in Pakistan

1
Together We Will Evidence from a Field
Experiment on Female Voter Turnout in Pakistan

• Xavier Gine Ghazala Mansuri
• DECRG, World Bank

2
Motivation

• Over the 20th century, women have acquired de
  jure rights to participate in democratic
  institutions
• Universal Declaration of Human Rights (1948)
• Convention on the Political Rights of Women
  (1952)
• International Covenant on Civil and Political
  Rights (1966)
• Convention on the Elimination of All Forms of
  Discrimination against Women (1979)
• However, barriers to effective participation by
  women both as voters and as legislators remain
  significant
• Number of efforts to introduce quotas for women
  legislators. Results suggest some impact on
  policy choices as well as perceptions
  (Chattopadhyay Duflo (2004) Bardhan et al
  (2005, 2008) Ban and Rao (2008))
• But women also have
• Lower participation rates as voters
• They are also more likely to vote in accordance
  with the preference of male clan and household
  heads (family voting) unlike men of all ages

3
Why should we care?

• Good governance and development viewed as
  intrinsically linked (Sen 1999) World Bank
  (2005)
• Voting is essential for electoral accountability
  Basic premise of representative democracy is that
  those who are subject to policy should have a
  voice in its making.
• Preference Heterogeneity Women have different
  preferences so their participation could lead to
  different policy choices
• Human Rights/Equity

4
Potential barriers to female participation in the
electoral process

• Costs of Participation
• Social constraints may restrict choices and/or
  restrict womens freedom of movement
• Traditions, social and cultural stereotypes may
  lead to a sense of disempowerment and discourage
  women from participation in electoral processes
  or exercising their own preferences
• Concerns about security in conflict environments
  may have a greater impact on female participation
  
• Information
• Women have fewer and poorer sources of
  information about the significance of political
  participation and/or the balloting process, in
  part due to illiteracy and mobility constraints
• Lack of information may reinforce disempowerment
  and stereotypes

5
What we assess

• How important is information for turnout and
  candidate choice?
• Why?
• Attitudes change slowly but information can be
  provided quickly and may serve to
• enhance equity
• induce a change in attitudes (Beaman et al
  (2007))
• be habit forming (Gerber, Green, Shachar (2003))
• change policy (Edlund Pande (2002) Lott
  Kenny (1999))
• Are there significant peer effects?
• Why?
• Is this a cost effective way to boost
  participation?
• Evidence of spillovers (Duflo Saez (2003)
  Kremer Miguel (2004 2007))
• Evidence of contagion within family
  (Nickerson(2008))
• Does information matter more/less when an
  election takes place in a politically volatile
  environment and is highly contested?

6
Context

• Rural Pakistan
• According to the 1998 Human Development Report,
  Pakistan ranked
• 138 out of 174 on the Human Development Index
  (HDI)
• 131 out of 163 on the Gender Development Index
  (GDI)
• 100 out of 102 on the Gender Empowerment
  Measure(GEM)
• Political parties, by and large, tend to view
  women as a passive vote bank, following the
  dictates of men within their families or clans.
  Even within their own parties, they treat them
  largely as followers to be strategically used for
  election canvassing and public campaigns. Thus,
  most parties do not even have lists of female
  members.
• Aurat Foundation, 2004

7
What we do

• Conduct a door to door voter information campaign
  directed at rural women just before the February
  2008 national elections in Pakistan
• Two treatments
• The importance of voting (T1)
• T1 plus the significance of secret balloting
  Ability to vote in accordance with ones own
  preferences without external pressure (T2)
• The information campaign was developed as a set
  of simple visual messages

8
Study Design-1

• Two districts in Sindh, Sukkur and Khairpur,
  selected because sharp electoral competition
  between two major political parties
• Pakistan Peoples Party (PPP) (secular-left
  leaning)
• and
• Pakistan Muslim League (F) (allied with the
  military, led mainly by large landlords who are
  also religious leaders pirs).
• 6 villages selected from each district,
  where an NGO, MRDO, which mobilizes
• women using a CBD approach was either
  working (or about to start work)
• 3 villages in Khairpur dropped just before
  the elections due to security
• concerns. These had more contested polling
  stations relative to our sample
• villages
• Final sample has 9 villages and 21 polling
  stations
• Average village population approx 300
  households

9
Study Design-2

• Variation in treatment type (T1 or T2) as well as
  treatment intensity to look at peer effects
• Village divided into geographical clusters
• Clusters randomly assigned to get T1, T2 or
  nothing as follows
• start in a random cluster, deliver T1
• leave a gap cluster
• in the next cluster, deliver either T2 or nothing
  using a coin flip
• leave a gap cluster
• deliver either T2 or nothing depending on prior
  coin toss result
• process repeated till all clusters in village
  covered
• Households within clusters selected as follows
• starting from any one end, every fourth household
  selected until up to 18 households covered
• In T1 and T2 clusters every 5th selected
  household left as a control. So 2 to 4 control
  households in treated clusters.
• In controls clusters, all selected household left
  as controls.

10
Timeline
Feb 18
Feb 18-19
Feb 5 -15
March 5-25
HH visits and Pre-Election Survey
Voting Verification
National Elections
Post-Election Survey
11
Data I

• Pre-Election Visit (information intervention)
• Household location (GIS) basic roster of all
  adult women, plus past voting record and the name
  and address of closest friend/confidant in the
  village
• No refusals, so we have 100 compliance
• Post-election verification
• Self report and verification by checking ink
  stain
• One friend per household, randomly selected from
  among women eligible to vote (had NIC or
  claimed to be on the voter list)

12
Data II

• Post election survey
• Household demographics, including caste
  (zaat/biradari)
• Intervention checks
• Mobility constraints
• Access to media
• Knowledge of location of polling station and the
  protocol for casting a vote
• Election day environment
• Knowledge of candidates, party platforms, recent
  political events, election outcomes
• Knowledge of whether other household members
  voted and for whom
• Polling Station data
• Electoral results by gender and by candidate/party

13
Final Sample

• Pre-election visit
• 64 clusters
• 1019 households
• 2735 women
• 2735 friends
• Post-election verification visit
• 64 clusters
• 992 households
• 2637 women
• 727 friends
• 98 women (27 households) lost because of
  temporary or permanent household migration.
  Friends of women in lost households not verified.
• Attrition is orthogonal to treatment
• Ink mark was missing for 135 women who claimed to
  have cast a vote. Err on the safe side by
  treating these women as not having voted

14
Household Characteristics by Treatment Status
15
Woman Characteristics by Treatment Status
16
Randomization worked

• Little difference in household characteristics.
  Treatment households have a little more land than
  control households in some comparisons, but no
  difference in assets or housing quality
• Women in treated households are a little younger
  in some comparisons and have more young kids as a
  result and also appear to have less access to
  cable TV, perhaps due to their lower mobility
• In the analysis, we control for the household
  and woman characteristics that we lack balance on
  as well as the total number of women registered
  to vote in a polling station
• We also control for whether the woman had a
  national id card (NIC), which is needed to cast a
  ballot, since young women are also less likely to
  have an NIC or to have voted in the past

17
(No Transcript)
18
Regression specification-Woman Level

• Average Effect
• For woman i in household h in village v
• Yihv bThv fXihv uv eihv
• Yihv Women voted (1Yes) based on verification
• Thv treatment indicator
• Xihv vector of control variables
• uv village fixed effect
• Standard errors clustered at geographic cluster
  level

19
Table 3 Average Effect of the Information
Campaign on Turnout
20
Table 4 Spillover Effects via Distance I
21
Regression Specification-Peer effects-II

• Similar to Kremer and Miguel (2004). For woman i
  in household h in village v
• Yihv bThv ?dD (gdD NTdD kdD NdD ) fXihv
  uv eihv
• Yihv Women voted (1Yes) based on verification
• Thv treatment indicator
• NTdD number of treated households between
  distance d and D from household
• NdD number of households between distance d and
  D from household
• Xihv vector of control variables
• uv village fixed effect
• Standard errors clustered at geographic cluster
  level

22
Table 5 Spillover Effects via Distance II
23
Table 6 Spillover Effects via Friendship
24
Regression SpecificationPolling Station Level

• For polling station p in village v
• Ypv bNTpv fXpv epv
• Ypv Number of valid votes cast by women
• NTpv Number of women treated in polling station
• Xpv Vector of polling station control
  variables, including the number of registered
  women

25
(No Transcript)
26
Table 8 Effect on Candidate Choice Using Cross
Reports from Family Members
27
Table 9 Contestation and Information
28
Summing up

• Substantial peer effects
• Accounting for spillovers, the information
  campaign increased turnout among sample women by
  about 12 (little more than an additional
  female vote for every 10 women (or about 4
  households treated)
• The polling station level effects are much
  larger. For every 10 women treated, there are
  almost 7 additional votes
• Information campaigns appear to be an effective
  way of reaching poor rural women
• I additional vote cost about 103 Rs. (or 1.51
  US)
• Some evidence that voting is habit forming, so
  sustained impacts from a single intervention are
  plausible
• Information campaigns can affect not just turnout
  but also independence in candidate choice
• Men in treated households have significantly less
  knowledge about womens candidate choice
• Information on electoral rights may be more
  valuable where differences in preferences over
  candidates are larger

29
Typical village
30
Typical street
31
Communication is easy
32
Communication is easy
33
(No Transcript)
34
(No Transcript)
35
(No Transcript)
36
Visual Aids for Treatments
37
Visual Aids for Presentation
38
Visual Aids for Presentation
39
Visual Aids for Presentation
40
Visual Aids for Presentation
41
Visual Aids for Presentation
42
Visual Aids for Presentation Secrecy of Ballot
43
Visual Aids for Presentation
44
Visual Aids for Presentation
45
Visual Aids for Presentation
46
Appendix Tables
47
Table A2 Intervention check
48
Table A3 Gender Differences in Radio and TV
Access and Exposure to News
49
Table A4 Gender Differences in Knowledge about
Current Political Issues and the Results of the
Election
50
Table A5 Gender Differences in Participation in
Village Political and Social Events
51
Table A6 Treatment Check for Measures of
Political Contestation at the Polling Station
Level
52
Table A7 Contestation and Voter Turnout
53
Table A8 Impact of Contestation on Women's
Participation and Candidate Choice
54
Table A9 Contestation Women's Reports
Regarding Election Day
55
Table A10 Impact of treatment on women's voting
report
56
Table A11 Effect of the Information Campaign by
Woman Characteristics
57
Table A12 Effect of the Information Campaign on
Behavior