Strengthening Female Political Representation in Nepal

2022-03-01

2022-04-30

The main objective of our experimental design is to measure the effect of updating female ward members’
beliefs about the chances of females being nominated and selected for the ward chair position and their
electability on:
— aspirations to run for the position of ward chair (or other, higher authority positions);
— success of being nominated by their party’s nominating committee for the position of ward chair (or
other, higher authority positions);
— success of being selected by their party’s selection committee for the position of ward chair (or other,
higher authority positions) and;
— success of being elected for the position of ward chair (or other, higher authority positions).

Aspirations to run for the position of ward chair are measured as follows:
(1) elected leaders, candidates and nominees are, by construction, considered aspirants;
(2) in all other scenarios aspirations are measured through the post-election survey.

Before commencing with the measurement of our outcomes of interest, all participants are truthfully informed about our objectives and policy partners. We developed our phone-call scripts in close collaboration with policy partners, including Sankalpa Women’s Alliance in Lalitpur, whose mission is to empower women to participate equally in all spheres of Nepali life and collected data in March and April
of 2022. After obtaining consent, we ask all respondents to confirm their identity and phone number and ask about their current party affiliation for a registry being kept by Sankalpa.

Next, we randomize respondents into five experimental groups:
(i) In the first group, comprising a pure control, we end the survey at this point (‘pure control’);
(ii) In the second group (’control with questions’), we ask respondents two additional questions. First, we ask about the size and number of females on their party’s ward committee. Second, we ask about their estimate of how many, out of 6,500 females holding a ward member position, got more votes than the person who won the ward chair race in the previous election in 2017. This experimental arm can then be used to compare female ward members’ beliefs about the gender composition of party selection committees and electability of female candidates with data on the actual composition and actual electoral outcomes; moreover we can test whether simply bringing
these questions to the participants’ top of mind (without providing information about their answers) has any effect on their aspirations;
(iii) In the third group (‘treatment A on committees’), we first ask respondents about the size and number of females on their party’s ward committee. In a second step, we truthfully inform respondents about a newly issued gender quota for selection committees and a finding from our own research that female committee members are more supportive of female aspirants than their male
counterparts. We use the following script: ‘Because of a recently enacted provision in the Political Party Act, every party committee should comprise at least one-third of women. As a result, compared to the last local elections, women are much more involved in the ward committee which nominates for ward-level positions as well as in the higher tier committee which makes selection decisions. Moreover, recent research based on selection committees from Nepal shows that women might favor the nomination and selection of other women, because women are much more likely than men to think of other women as effective leaders.’ We further directly encourage respondents to vie for ward chair positions using the following language: ‘So, as you think about running for local office again, we want to encourage you to consider vying for the position of ward chair because we think that this year women will have a better chance of being nominated and selected by their party for ward chair positions than in the last election.’ This arm allows us to test whether updating beliefs on the chances of female candidates being nominated and selected for the position of ward chair causes more female ward members to vie for the position of ward chair (or higher positions) in the 2022 local election;
(iv) In the fourth group (‘treatment B on votes’), we first ask respondents about their estimate of how many, out of 6,500 females holding a ward member position, got more votes than the person who won the ward chair race in the previous election in 2017. In a second step, we truthfully inform respondents about the actual vote distribution in the 2017 election. We use the following script: ‘Data from the Election Commission shows that in about 2,800 out of 6,500 wards in Nepal (that is, 43% or about one-half of the wards), [female ward members in quota-protected positions/Dalit female ward members/female ward members in open seats] like you secured more votes than the
person who won the race of ward chair in their ward in 2017. Furthermore, research shows that voters’ level of trust in their female local representatives has been increasing in the last four years.’ We further directly encourage respondents to vie for ward chair positions using the following language: ‘So, as you think about running for local office again, we want to encourage you to consider vying
for the position of ward chair because we think that this year women will have a better chance of being elected by their voters than in the last election.’ This arm allows us to test whether updating beliefs on the chances of female candidates being elected for the position of ward chair causes more female ward members to vie for the position of ward chair (or higher positions) in the 2022 local
election;
(v) In the fifth group (‘treatment A on committees and B on votes’), we combine the scripts from experimental groups 3 and 4. This arm allows us to test for the complementarity of providing respondents with both pieces of information.

The randomization is done through Stata.

The unit of randomization is the ward-party. In practice, however, in more than 95% of wards, all female ward members belong to the same party meaning that the ward-party level treatment is effectively a ward level treatment. For brevity, we then describe this as ward level clustering. In assigning treatment, we stratify on province and party affiliation.

Yes

The treatment is assigned at the ward-party level, and we attempt to contact participants from all wards (6,743 in total across Nepal). The experimental sample spans 4,487 wards.

Among the 4,658 ward-party clusters in our experimental sample, we have 954 clusters in group 1 (pure control), 942 in group 2 (control with questions), 904 in group 3 (treatment A on committees), 926 in group 4 (treatment B on votes) and 932 in group 5 (treatment A on committees and B on votes).

There are 13,439 female ward members in Nepal (6,567 Dalit Female Members, 6,742 Female Members and 130 Members in open seats). We obtained phone numbers for all but 611 of these female members using two methods. First, Sankalpa had an existing record of numbers. Second, we contacted the ward chair (whose phone number was obtained from the municipality website) or municipal officials if the phone number for the ward chair was not publicly available to elicit missing contact information of female ward members. This method also allowed us to obtain current numbers for members in the Sankalpa registry who had incorrectly listed numbers. We managed to successfully contact 6,824 female ward members before the election took place. Only 5 respondents denied survey participation. Our experimental sample is defined as female ward members whose phone number we were able to obtain and contact in advance of the elections. Our experimental sample, therefore is 6,819 female ward members. There are two reasons we failed to contact some female ward members. First, no one answered the number that we identified for a female ward member (4,667 cases). Second, we did not have time in advance of the election to call a number that we identified for a female ward member (1,337 cases).

We randomize the universe of 13,439 female ward members into the following groups: 2,700 participants in group 1 (pure control), 2,705 in group 2 (control with questions), 2,684 in group 3 (treatment A on committees), 2,670 in group 4 (treatment B on votes) and 2,680 in group 5 (treatment A on committees and B on votes). Among the 6,819 ward members in our experimental sample, we have 1,460 participants in group 1 (pure control), 1,343 in group 2 (control with questions), 1,312 in group 3 (treatment A on committees), 1,367 in group
4 (treatment B on votes) and 1,337 in group 5 (treatment A on committees and B on votes).

To compute the MDE of any of our treatments on the probability of aspiring to the ward chair position, we assume that we will have data for 1,360 participants per treatment and 680 wards per treatment, and an intra-cluster correlation of 0.8. At conventional levels of significance (5%) and power (80%), we calculate a MDE of about 3 percentage points. MDEs are also about 3 percentage points for the other outcomes of interest (nomination, selection, and election).