Elites’ Response to Women’s Action Committees and Local Services in Nigeria

This RCT was conducted with over 5,800 women in 450 communities in three southwestern states of Nigeria (Ogun, Osun, and Oyo) during 2023. In both treatment and control communities (i.e., wards), we recruited previously unaffiliated women to join women's action committees (WACs): 1/3 of WACs (control group) received basic training in civic education; 1/3 (treatment group 1, T1) received the same basic civic education training plus intensive training in leadership, organizing, and advocacy, intended to build women's collective agency and improve their articulation of demands to policymakers and policymakers' responsiveness to them; and 1/3 (treatment group 2, T2) received the same training as treatment group 1 plus their husbands were invited to participate in a parallel men's training focused on men's allyship. The effects of the treatments on women's political participation and on women's aspirations and economic activities are pre-registered on OSF and AEA, respectively. In this trial, we aim to test the effects of these treatments on elite responsiveness to petitions by women and about women's issues.

2023-05-28

2024-03-03

To assess the responsiveness of local elites, we employed a forced-choice conjoint experiment in our survey. We presented elites with hypothetical petitions from community members seeking their help to address problems in the community. The petitions varied on: (1) the petitioner, (2) the issue being raised, and (3) the affected demographic within the community. After reviewing the petitions, respondents were asked to choose which of the two petitions to respond to and to explain their reasoning. The question wording and response choices are detailed in the attached pre-analysis plan document.Two outcome measures were asked in the conjoint experiment: a forced-choice question to choose between two petition profiles and a question to explain the respondent's choice. Hypotheses 1-11 examine the effects on the former outcome measure.

To test whether elites perceive an increase in women's participation for Hypothesis 12, we will create an additive index using 6 survey questions enumerated in the attached pre-analysis plan.

To measure elite responsiveness, we examine the choice that elites make in the forced-choice question in which they are asked to select between two petition profiles. Each elite in our survey was asked: "if you could respond to only one of these two petitions, which would you choose?", with the response options being: (1) first petition; (2) second petition; (3) nudged, but don't know; and (4) nudged, but refused to answer.

We additionally asked and will analyze why elites chose the specific profile, using the following question: "Why did you choose the [first / second] petition? Select the option that comes closest to your thinking." The following are response options: Easiest; Know most about; Most needed; Most important; Affects most people; Most likely to be fixed; Other (specify); (UNREAD) Nudged, but Don't Know; (UNREAD) Nudged, but Refused to Answer.

To test whether elites perceive an increase in women's participation for Hypothesis 12, we will create an additive index use the following 6 survey questions:
(1) When thinking about voting, do you think women in your community vote in national elections as often as the men do?

(2) When thinking about voting, do you think women in your community vote in local elections as often as the men do?

(3) When thinking about who contacts you about service delivery problems, do you think women in your community contact you as often as men do?

(4) When thinking about who participates in political discussions, do you think women in your community participate in community meetings as often as men do?

(5) When thinking about who attends community meetings, do you think women in your community participate in community meetings as often as men do?

(6) When thinking about who stands for elections as candidates, do you think women in your community stand for elections as often as men do?

In our attached pre-analysis plan, we additionally note exploratory analyses that we may conduct to determine whether the treatments of men and women in their community influenced the likelihood of elites preferring collective petitions over individual petitions. We also note that we may examine whether, relative to elites in T1 and control communities, elites in T2 communities exhibit shifts in gender norms and social perceptions.

The broader Metaketa V project posits that a key reason behind women's low political participation is psychological constraints---such as a belief that they cannot effect meaningful change. Research in psychology has shown that increasing the salience of group identity and perceived group efficacy can motivate individuals to engage in collective action. Metaketa V draws on the \textit{social identity model of collective action (SIMCA), which emphasizes that group identification, a sense of group-based injustice, and beliefs in collective efficacy are essential drivers of collective action. Metaketa V hypothesizes that an intervention enhancing women's group identity, their sense of group-based injustice, and their beliefs in their own collective efficacy will affect three main outcomes: (1) the quality of women's political participation, (2) the level of women's political participation, and (3) politicians' responsiveness to women's participation.

Our focus here is on the third outcome: responsiveness to women's participation, which we consider a downstream effect of this intervention. While external factors such as top-down pressures, electoral calendars, and patronage dynamics may condition the responsiveness of local leaders, we aim to understand whether T1 and T2 can increase and improve their responsiveness to the types of bottom-up demands which might be generated by empowered women's groups.

As described in our previously registered Metaketa pre-analysis plan, we expect that T1 will have downstream effects on the responsiveness of local leaders (baales), who we simply describe as elites. Additionally, we expect that T2 will amplify these effects by increasing elite responsiveness even further through training both women and their husbands. In patriarchal societies, men control a disproportionate share of resources, including access to healthcare, education, transportation, extension and trainings, technologies, and financial services. They are also significantly more likely to hold leadership roles in government and religious organizations, and wield greater decision-making power within households. As a result, their behaviors and attitudes matter profoundly for women's outcomes. At the same time, however, men can leverage their influence to serve as "male champions" who actively support women's efforts to gain greater voice and agency in public spaces. Therefore, we expect the men's training will produce additional downstream effects on elite responsiveness given how men's support for women's voice and agency will be observed by elites, surpassing the impact of the women's training alone.

This RCT was conducted with over 5,800 women in 450 communities in three southwestern states of Nigeria (Ogun, Osun, and Oyo) during 2023. In both treatment and control communities (i.e., wards), we recruited previously unaffiliated women to join women's action committees (WACs): 1/3 of WACs (control group) received basic training in civic education; 1/3 (treatment group 1, T1) received the same basic civic education training plus intensive training in leadership, organizing, and advocacy, intended to build women's collective agency and improve their articulation of demands to policymakers and policymakers' responsiveness to them; and 1/3 (treatment group 2, T2) received the same training as treatment group 1 plus their husbands were invited to participate in a parallel men's training focused on men's allyship.

The trial registered here tests the effects of these interventions on the responsiveness of local elites (baales) -- measured using a forced choice conjoint experiment. This trial is being registered before any analysis of the elite survey has been conducted.

Not available

For the field experiment, we randomized the 450 wards into control, treatment 1 (women's training), or treatment 2 (women's + husbands' training) after all burn-in meetings and the women's baseline survey took place. Data collected from these activities and a supervisor survey of the communities was used to create blocks for randomization. Please reference the Metaketa V team PAP for more details on the randomization procedures.

For the paired, forced-choice conjoint experiment, respondents evaluated pairs of fully randomized petition profiles and repeated the process five times to ensure robust data on their decision-making patterns. The petitions were formatted in a vignette style, rather than two side-by-side tables. As described in the table in the attached pre-analysis plan, each petition varied randomly across three attributes: 1) the petitioner, 2) the issue, and 3) the affected demographic in the community. Randomization for the conjoint experiment was manually programmed into SurveyCTO, meaning that a plug-in feature was not used. The combination of attribute levels had an equal probability of being selected---i.e. we used a uniform distribution of profiles. The only constraints imposed on the randomization process were to guarantee respondents saw unique draws: they were not presented with the same pair of petitions more than once nor with a pair of identical petition profiles.

Randomization of the training treatments was at the village level.
Randomization for the conjoint experiment was at the level of the paired petition profile. Each elite in the sample receives paired petition profiles five times.

Yes

For the training treatment, there are 450 WACs (1 in each of the 450 wards in the sample), and randomization is assigned at the ward level. There are 13 women in each WAC.

~2,250 (450 elite x 5 paired profiles each)

150 elite in control, 150 elite in T1, 150 elite in T2.