Women as Policy Makers: Evidence from a Randomized Policy Experiment - - PowerPoint PPT Presentation

Women as Policy Makers: Evidence from a Randomized Policy Experiment in India

by Raghabendra Chattopadhyay and Esther Duflo Econometrica, 2004

Beia Spiller September 24, 2008

Introduction

• Women are underrepresented in all political positions
• Quotas for women
• Women and men differ in their policy preferences
• Should political outcomes depend on gender of politicians?
• Median voter model
• Coasian bargaining model
• Do political outcomes depend on gender of politicians?
• Evidence unclear

Background

• Gram Panchayat (GP) system in India
• Village system of representation
• 10,000 people per GP (5-15 villages)
• Voters elect council, which elects Pradhan (chief)
• 1992: 73rd Amendment of Constitution
• GP has power to implement development programs as needed
• Local infrastructure and welfare programs
• Reserved 1/3 of Pradhan positions to women

Reservation for Women

• Random Selection of GPs to be reserved for women
• Every third GP is reserved for a woman for the first election
• The reserved GP cycles through all GPs in each state
• Policy Question: Does the existence of reserved GPs influence the policy
• utcomes relative to the non-reserved GPs?

Model

• Osborne and Slivinski (1996) and Besley and Coate (1997)
• Elected representatives are “citizen candidates”
• Citizens decide whether to run or not depending on cost of running
• Game:
• 1. Citizens decide whether to run or not
• a. W > M
• 2. Citizens elect representative
• 3. Policy is implemented
• Policy:
• In the interval [0,1]
• Preferred policy option for citizen i: i
• Women’s preferences: [0,W]
• Men’s preferences: [M,1]

M W, or M W

Model

• Utility of citizen i if outcome xj is elected:

–| xj – i | or –| xj – i | – i

• Élite influence, prefer option '
• ' > m (median voter is not represented by élite)
• xj = j + (1 – ) '
• is constant across all candidates
• Default decision: '
• Citizen’s utility: –| ' – i |

Women’s Decision to Run for Pradhan PROPOSITION 1: Under the following conditions, no woman will run for office in the absence of reservation- (i) W – .5* M > ' – m; (i.e. no woman runs unopposed) (ii) W > m – (1– ) ' . (i.e. no woman runs because costs are too high) LEMMA 1: If W > ' – (1–) '= ', no women run under reservation regime. If cost of running is so high for women, even the most extreme woman (ie i = 0) would prefer to not run and take the default option.

Welfare Impact of Reservation PROPOSITION 2: If W > ', ' – [M + (1–)'] M and ' > max(m + .5M , 2m – [M + (1–)']), the reservation leads to an unambiguous loss in the utility

• f the median voter and that of women.

If no one runs under reservation (W > '), but a more moderate man than the élite/default option could have run (' – [M + (1–)'] M) and been elected (' > max(m + .5M , 2m – [M + (1–)'])), then reservation harms welfare. PROPOSITION 3: If ' – (1–)' W, and the conditions in Proposition 1 are satisfied (no woman runs without reservation system), then the reservation system: (i) always increases the utility of the median female voter if ' – [M + (1–)'] min(m + .5W , W + (1–)', ' – W ) (ii) always increases the utility of the median voter and of the median female voter if condition (i) is satisfied and, in addition, ' – [M + (1–)'] > 2m – max((1–) ', (m – .5W)).

Ex-Post Lobbying and the Median Voter

• Ex-post lobbying leads to a more extreme “pro-male” outcome
• Electing a female Pradhan can moderate this right-leaning tendency
• Reservation can affect policy outcomes
• Can move policies in a pro-woman direction
• Helps both women and median voters

Model Limitations 1) ' can be influenced by reservation if it is lobbying by ordinary citizens

• a. This can move ' to the left

2) The assumption that is constant for all candidates

• a. Before reservation, only women with high would run
• b. Men also would only run with high in order to attract median voters
• c. Thus, candidates’ characteristics are endogenous to system of

reservation, can bias estimates 3) Model ignores other effects of reservation system

• a. Possibility of strategic behavior and incentives of elected official
• i. Women will not run again after reservation term (if Prop. 1 holds)
• ii. Men who are elected before reservation are lame ducks

Measures of Preference

• Data on formal requests and complaints brought to Pradhan
• Simple Example: Series of binary policy decisions
• A random villager expresses his preference for the decision.
• Cost of speaking is bi: cost is independent of reservation
• Pradhan’s prior belief of village preference is .5 and assigns weight of to

the prior, and 1 – to the signal.

• Villager expresses preference if .5(1 – )(1 – ) bi
• Frequency at which a good appears is unbiased estimate of frequency which

this good is preferred to every other good

• −

=

m m i w w i i

N n N n D

: strength of difference between women’s and men’s preferences

• +

=

m m i w w i i

N n N n S 2 1

: strength of preference in aggregate population (if Nw = Nm)

Testing the Assumptions Simplifying assumption: cost of complaints does not change with reservation. Test if cost is affected by Reservation Policy: difference in frequency of requests in reserved/unreserved GPs.

• ' will be farther to the left in unreserved GPs if cost is affected by

reservation Finding summary: while reservation does increase participation by women, it does not change how extreme these complaints are.

• ' is unchanged by reservation

Data Collection Locations: Two districts- Birbhum in West Bengal, Udaipur in Rajasthan. Method: in 2000 (WB) and 2002 (R), all GPs were surveyed.

• Interview with GP Pradhan.
• Survey of 3 villages in GP
• group survey, questions on what infrastructure was installed since May

1998 (for West Bengal), and from 2000 (for Rajasthan)

• Survey of number of complaints brought to Pradhan

Empirical Strategy

• Due to randomization of reservation policy, can merely compare outcomes of

interest in reserved and unreserved GPs.

• Let Yij be outcome of interest for good i in village j, Rj is a dummy for

reservation in village j: Impact of reservation on outcome = E[Yij | Rj = 1] – E[Yij | Rj = 0]. This is valid due to the very low numbers of women in unreserved Pradhan positions.

• Use OLS for regressions of determinants on public good provision.

Empirical Strategy Want to test if there is more investment in goods preferred by women in reserved GPs:

• =

+ + ⋅ + ⋅ + =

N l ij il l j i j ij

d R D R Y

1 3 2 1

ε β β β β

and

• =

+ + ⋅ + ⋅ + =

N l ij il l j i j ij

d R S R Y

1 6 5 4

ε β β β β

The model predicts: 3 0, and potentially 6 0 (gender of Pradhan positively affects policy outcome in reserved GPs)

Empirical Strategy Want to test whether difference in policy comes from greater responsiveness of female Pradhans to complaints by women in a specific village:

• =

+ + + + ⋅ + ⋅ + ⋅ + ⋅ + =

N l ij il l ij ij j ij j ij j i j ij

d D S R S R D R D R Y

1 13 12 11 10 9 8 7

ε β β β β β β β β

If 10 = 0 and 11 = 0:

• Village specific complaints are drawn from distribution of preferences

common to the district: ' is unchanged by reservation

• Pradhans are selected for their specific preferences and this affects policy
• utcome: women Pradhans are no more responsive to needs of women than

men; ' = 0.

Results: Reservation Effects on Political Participation of Women

Results: Requests of Men and Women

Si Di Si Di

Results: Effects of Reservation Policy on Public Goods Provision

Results: Are female Pradhans more responsive to women than men?

Robustness Checks: Controlling for Other Effects of Reservation

Robustness Checks: Controlling for Other Effects of Reservation

Robustness Checks: Controlling for Other Effects of Reservation Are the results driven only by gender?

Robustness Checks: New Pradhans, Lame Ducks, Social Status Impacts

Robustness Checks: New Pradhans, Lame Ducks, Social Status Impacts

New Pradhan Lame Duck Social Status

Robustness Checks: New Pradhans, Lame Ducks, Social Status Impacts

Conclusion

• Women Pradhans under reservation policy invest more in public goods

demanded by women.

• Results driven mostly by gender of Pradhan
• Contradicts median voter model and Coasian bargaining model of politics
• Direct manipulation of identity of policymaker will impact the policy
• utcomes.

Possible Issues

• Is it possible that Pradhans are significantly different from the rest of the

women in the community? Paper uses only aggregate community women’s interests as measure of Pradhan’s preferences.

• Issues with ':
• The model assumes that ' is extreme, but the actual input in these GPs

are just a random selection of people: i.e. ' = m.

• The coefficients on Dij*Rj and Sij*Rj = 0: i.e. ' = 0 or = 1.
• What is the default outcome?
• Election system: model posits that citizens vote for candidate directly,

however, Indian citizens vote for council, who vote for candidate.

• Can cause some issues of strategic play that is overlooked