Institutions and Political Inequality

NFS check in. Sept 2024

IPI

Who we are

Where we are working

Themes

  • Vertical linkages
  • Horizontal contestation
  • Exclusion
  • Elite connections

Themes

  • Vertical linkages
  • Horizontal contestation
  • Exclusion
  • Elite networks

Vertical linkages

How do citizens control elites (and vice versa)?

Horizontal contestation

What determines between group cohesion and conflict?

Exclusion

How do citizens respond to political exclusion?

Elite networks

What is the structure of elite networks?

How we work

IPI principles


  1. Engaged research •

  2. Ethical research

  3. Premium on quality •

  4. Supportive environment

1. Engaged research

We are outward looking, inclusive, committed to co-production and transfer

  • German development community: DFG / GIZ / Deneb
  • World Food Program
  • US state department, USAID, US Dept of Labor
  • EGAP policy events (with IGO, NGO, government audiences) – The Hague (2018), Geneva (2019), London (2023)
  • Kampala City Authority
  • Sierra Leone Ministry of Agriculture
  • Nigeria Agency for the Prohibition of Trafficking in Persons

Supporting international scholars:

  • EGAP Africa researcher trainings (Humphreys)
  • Kenya Busara trainings (Scacco)
  • Africa journalist training (Garbe)
  • South Asian network mentoring (McMurry)

Our members sit on committees allocating research funding in the development sector:

  • IPA peace and recovery
  • HFG fellows selection
  • Deval

3. Premium on quality

With a commitment to open science practices.

Formal design declarations for transparency and diagnostics

Bayesian approaches to integrated qualitative and quantitative data

Publication strategy: We favor fewer publications in high impact outlets.
Since 2017 have appeared in:

  • Leading general science journals, including Nature Medicine, Nature Human Behavior, PNAS, PLOS medicine and Plos One, and Science Advances (twice).

  • All three of the the “top 3” political science journals: APSR (3 times) , AJPS, and JOP (twice)

  • Two of the leading development journals: World Development and JDE (twice each).

  • The leading methodology journals in political science (PA) and in sociology (SRM).

  • Excellent UPs: 2.5 Cambridge University Press, Princeton University Press.

Project sampler

Motivation

  • Anti-refugee prejudice on the rise (Valentino et al. 2017).

  • Practitioners see contact and empathy education as promising interventions.

  • But: Limited evidence on effectiveness of refugee-native contact in realistic settings.

  • And: Unclear whether contact, empathy education, or their combination is most effective.

  • Research questions:

    • Can intergroup contact improve refugee-host relations?
    • Are there gains from combining contact with empathy education?

Challenges

  • Self-selection into intergroup contact, opportunities to build empathy.

  • How to create an intervention that resonates with participants from both sides of a deep social divide?

  • How to deliver treatment in a naturalistic way?

Design

  • Self-selection into media consumption.
    • Partner with Lebanese NGO to design RCT with 1000 Syrian and Lebanese youth.
    • Randomly assign participants to contact, empathy education in a 2x2 factorial design.
  • Naturalistic intervention that resonates across social divide.
    • Design experiment around the well-established Family Psycho-Social Support (FPSS) program.
    • Work with a local design team to add context-appropriate empathy (and placebo) content.

Design (Continued)

Context: Syrian refugees in Lebanon

  • 1.5 million Syrian refugees arrived since 2011.
  • Syrian refugees make up 25% of the population.
  • Highly restrictive and discriminatory policies.
  • Tensions over access to employment, state assistance, and international aid.
  • Few opportunities for meaningful contact—e.g., temporal school segregation.

Results: Contact

Results: Empathy Curriculum

Results: Contact and Empathy

What have we learned?

  • Realistic experiment with strong treatments suggests:
    • Contact and empathy curriculum each have positive effects on attitudes.
    • Empathy treatment effects are larger and more robust than contact effects.
    • No evidence of interaction effects.
  • Given the (high!) cost of inducing contact in conflict settings, empathy training initiatives may be a better investment.

Intervention details

  • Intervention: 12 weeks of FPSS programming.

    • 2-3 hour sessions, once per week.
    • 10-12 students per class.

  • Participants: 1000 Lebanese and Syrian children, aged 11-15.

    • Outreach through ads at local public schools.
    • Approx 60% Syrians, 40% Lebanese.

  • Course content: Focused on mental health.

    • Empathy: active listening, conflict management, working in diverse communities.
    • Health: nutrition, exercise, substance abuse, digital health.

  • Random assignment: Homogeneous vs. heterogeneous FPSS class; empathy vs. health curriculum.

  • Outcomes: Prejudice, conflict management skills, participation in mixed social and cultural events.

Results: Empathy vs. Contact

Kampala’s Citizen Charter process


In 2019 we were invited to support a set of consultations to provide input into the construction of a “Citizen’s Charter” for Kampala.

The Charter was to outline:

  • principles and clear standards of service provision
  • rights and responsibilities of citizens and bureaucrats

We had a chance to observe:

  • The creation of an institution
  • The points of disagreement and how these were resolved

Bosancianu, Garcia-Hernandez & Humphreys

Goals

Goal 1

The major focus was horizontal contestation in particular the structure of:

  • Input inequality:

  • Throughput inequality:

  • Output inequality:

between social groups.

Goal 2

However we also get to use experimental methods to learn about the vertical linkages in this setting:

  • how much do bureaucratic elites influence citizens
  • which citizens are more or less vulnerable to this

Bosancianu, Garcia-Hernandez & Humphreys

Design: Overview


  • Baseline
  • Small-scale consultation meetings, of around 1-1.5 hours
    • Participants: 6-8 citizens recruited from the same village
    • Facilitated by KCCA officials, or neutral facilitators (trained enumerators)
    • Objective: collect input from citizens for construction of Charter
    • Decisions: made collectively by the group, and recorded by facilitator
  • Decisions relayed to city
  • City drafts charter
  • Endline

Bosancianu, Garcia-Hernandez & Humphreys

Topics


Pre-identified set of topics | Vetted by KCCA | Plausibly contentious

  • Report budget: At what levels should budgets be reported?
  • Channels of communication: At what levels should KCCA engage with citizens?
  • Growth vs equality: Should KCCA focus on growth or inequality?
  • Raising fees and taxes: Should there be higher taxes for better services
  • Monitor Charter: What level of oversight should there be?

Also: Is KCCA going in the right direction?

Bosancianu, Garcia-Hernandez & Humphreys

Baseline sampling


Bosancianu, Garcia-Hernandez & Humphreys

Elite influence


Strategy:

  • Meeting facilitators are not meant to influence outcomes
  • The random assignment of facilitators to meetings however lets us assess the extent to which they do

  • Variation both between facilitators and between types of facilitator
  • Evidence for such influence in 4 of 5 topic areas.
  • However, relatively weak: ca 20% of variation explained by leader effects

Bosancianu, Garcia-Hernandez & Humphreys

Are they more influential in disadvantaged communities?


Is the influence of facilitators weakened when communities are more privileged?

  • For communication channels with citizens we see:
    • Effects of facilitators are eliminated in privileged communities
    • Facilitator preferences drive meeting outcome more in disadvantaged communities
  • Similar dynamic, though fainter, for a second topic
  • No evidence for remaining three topics
  • Overall consistent with flat decision making and limited manipulation of weaker groups

Bosancianu, Garcia-Hernandez & Humphreys

High level outcomes

Fairly clear answers about local priorities

  • Often contrary to bureaucratic preferences
  • Often in line with values of disadvantaged

Limited manipulation:

  • Either by bureaucrats
  • Or by locally powerful groups (traditionally influential groups are least satisfied with outcomes )

Transfer: Did it matter?

  • We think so: Results and report used by KCCA to develop draft which is (still!) before the city council

Bosancianu, Garcia-Hernandez & Humphreys

Background

  • GIZ and Berghof Foundation are implementing development projects in 24 Lebanese municipalities using participatory planning to ensure successful project outcomes and enhance trust and cohesion.
  • This “Rigorous Impact Evaluation” (RIE) focuses on assessing the direct effect of citizen participation and aims to evaluate how elite contact and participatory planning influence citizens’ attitudes and behaviors towards local government.

Challenge

  • Engagement with government obviously endogenous to trust
  • Evidence based for merits of high level participation in development projects weak
  • German development community relatively inexperienced with RCTs

Research Design

  • BF organizes participatory planning and decision-making processes between representatives of municipalities, civil society, and citizens across 24 municipalities.
  • Working groups (WG) comprising 15 members will be formed within these municipalities to:
    • discuss the socioeconomic needs of the area,
    • identify gaps and necessary measures or projects, and
    • jointly prioritize and select the needed measures.
  • The 15 WG members will include: 10 chosen members from experts, municipality officials, and influential community leaders, and 5 randomly selected members of the local community. (Possibility to increase this to 8 randomly selected?)

Theory

Hypothesis

  • Direct participation in a WG positively affects trust in local government.

However:

  • Impact of exposure to local government depends on initial perceptions as well as local capacity.

Effects are expected to be largest among those with low initial levels of trust (in high capacity locations). But may be negative for those with high levels of trust (in weak capacity locations)

  • Project hypothesis: effects will be strongest for minorities and for women, who are otherwise more likely to be excluded from government.

Measurement

The index for levels of trust is a simple mean of standardized responses to the following questions:

  • I am confident that the municipality works for the benefit of its citizens.
  • I believe that the resource distribution by the municipal board is fair.
  • Corruption is widespread in local government.

Measurement

I trust that my local government acts in the interests of its citizens

Behavioral outcome

We also measure trust using a behavioral outcome:

  • Money allocation:

“We are interested in your views on who can best spend development money in your municipality. Imagine a grant of 10,000,000 Lebanese Pounds by a development agency is made available to alleviate poverty in this municipality. How much do you think should be managed by the municipal authorities, and how much should be transferred directly to the poorest people in this town?”

Endline Data

  • Perception of project decisions.
  • Endline measures of attitudinal measures, e.g., trust in authorities.
  • Endline behavioral measures: Grant allocation.
  • Include a second behavioral measure regarding attending a townhall meeting. Respondents are informed about an upcoming municipal town hall meeting on a subject unrelated to the intervention.

Status

  • This seems very foreign to our partners and a lot of teething issues

  • Assignment to treatment serious hampered by contactability and refusals, itself indicative of low trust

  • German involvement possibly particularly sensitive

  • Launch next week

Motivation

  • Large-scale efforts to improve vertical citizen-state legibility across developing countries

  • Example: introduction of biometric identity cards (eID)

  • Two aspects of expansion of state capacity:

    • Improved service delivery
    • Greater scope for state surveillance

  • Differential implications across social groups:

  • Question: Does eID exacerbate existing intergroup inequalities, and if so, under what conditions?

Garbe, McMurry, Scacco & Zhang

Challenges

  1. Sampling and surveying marginalized populations

  2. Explaining complex policy choices in an accessible way

  3. Assessing actual effects of eID on political inequality

Garbe, McMurry, Scacco & Zhang

Design

  1. Sampling and surveying of marginalized populations

    • In-person survey in Kenya (N = 2073) with members of dominant (Kikuyu, Kalenjin), opposition (Luo), and securitized (Somali) groups

  1. Explaining complex policy choices in an accessible way

    • Hypothetical policies highlighting different benefits and costs

  1. Assessing effects of eID on political inequality

    • Policy conjoint experiment
    • Focus on heterogeneity in attitudes across groups
    • Implications for changes in group-level political behavior

Garbe, McMurry, Scacco & Zhang

Conjoint attributes


  • Hypothesized benefits








  • Hypothesized costs

Garbe, McMurry, Scacco & Zhang

Results 1: Main effects

Figure 1. Willingness to register for eID

  • Overall positive reactions to all eID policy features

  • But securitized group less positive, especially about surveillance

Garbe, McMurry, Scacco & Zhang

Results 2: Secondary effects

Figure 2. Left panel: ‘I would be worried about the police using my personal information’, right panel: ‘I would be worried about being punished for expressing my political views’
  • Opposition group more worried about consequences of surveillance for political participation

Garbe, McMurry, Scacco & Zhang

Results 3: Qualitative Evidence

Why are Kenyans largely positive about eID?

  1. Digitalization as a means of reducing discrimination:

“I think it will really help reduce discrimination against Somalis. If we can just do everything online, we won’t have to have these bad face-to-face interactions. We won’t have to face harassment, or stereotyping, or have to be asked to give bribes.”

  1. Perceived lack of state capacity reduces potential concerns:

“In Kenya, the police are not very reliable. They are typically not very well educated … If they give the police this information, I think it’s OK. What can they even do? In general, the police you meet in the day to day, they are so corrupt, they are just taking bribes. If they are so incompetent, I am not too worried about them having information, since I don’t think they would even know how to use it.”

Garbe, McMurry, Scacco & Zhang

What have we learned?

  • Overall: Citizens appreciate stronger link with government through eID
  • But: Some evidence that particular groups may suffer from
    • Direct exclusion from access to service (securitized)
    • Indirect exclusion via reduced participation (opposition)
  • Ambiguous effects for inequality

Garbe, McMurry, Scacco & Zhang

Motivation

  • Elite connections have political and economic implications. For example,
    • Firm executives with government experiences can access and influence the policy-making process.
    • Government officials with business exposure can respond to market changes through policy adjustments.
  • The project seeks to answer three questions:
    • How can we measure political-economic elite networks?
    • What shape the formation of political-economic elite networks?
    • How do political-economic elite networks influence vertical linkage and horizontal contestation in societies?

Guo et al

Challenges

  • Collecting network data on elites in a scalable way
  • Interpreting network data on elites in a meaningful way

Guo et al

Design

  • Collecting network data on elites in a scalable way
    • Search co-appearance of elite names across a large set of news media
  • Interpreting network data on elites in a meaningful way
    • Use large language models (LLMs) to assess the co-appearance according to a set of questions regarding to what extend the two elites engaged
    • Use human coders to validate the LLM results
    • Use human investigators to validate the news itself

Guo et al

Pilot results

  • search 5204 pairs of political-economic elites (incl. Olaf Scholz, Markus Kamieth, or Christian Sewing)
  • download 4002 articles over the past decade from 30+ German news media
  • code the articles via OpenAI gpt-4o API
    • over 90% consistent with human codes for all 11 questions
  • identify a link if both names refer to the right elites AND the two elites have met OR knew each other historically

Guo et al

What have we learned?

  • It is possible to develop a spatiotemporally scalable method of measuring political-economic elite networks

Next steps:

  • apply the method to German federal MPs and top-100 firms’ management board members
  • explore the nature of the connection reported in the news qualitatively
  • consider possible media reporting bias
  • study the political-economic elite connections as dependent or independent variables

Guo et al

Wrap up

The Future

Clusters

  • Vertical linkages
  • Horizontal contestation
  • Exclusion
  • Elite networks

Directions

  • We expect a deepening of between-cluster analyses: How do horizontal contestation enhance or weaken vertical linkages?
  • We expect a deepening of work on elite networks Can we generate a global database of political connections?
  • We remain committed to deepening collaboration with Global South researchers

Agenda

Need finishers: KCCA project, Nigeria migration project, School segregation, Covid correlates?

Looking forward: Zimbabwe project, Scacco and Grbe projects, MH book

Some issues:

  • Internal: finishers
  • External: Research professor frustrations, GS challenges, admin burden, IT responsiveness

Fin

Extra projects

Motivation

Contact hypothesis:

  • If cooperative and egalitarian, social contact can reduce prejudice
  • But: Real-world opportunities for contact in conflict settings often limited
  • And: Even given opportunities, anxiety can lead to avoidance
  • Mass media interventions as potential solution

Research question: Can media interventions improve intergroup relations in settings of horizontal contestation?

Scacco

Challenges

  1. There is self-selection into media consumption

  2. How to generate media content that resonates with audiences on both sides of deep social divide

  3. How to deliver in a naturalistic way with broad reach

Scacco

Design

  1. Self-selection into media consumption

    • Use experimental design to encourage subset of study subjects (N=1750) in conflict setting (Kaduna, Nigeria) to watch TV drama each week
      • Weekly reminders and incentivized quizzes to boost take-up

  2. Create content that resonates across social divide

    • Work with local TV scriptwriters to design relatable storyline focused on transformative Christian-Muslim friendship

  3. Deliver content in naturalistic way with broad reach

    • Embed storyline into 10 episodes of primetime show, “Dadin Kowa”

Scacco

Kaduna: a divided and unequal city


  • Christian-Muslim clashes:
    • 2000, 2002, 2011, 2012
  • 58% personally witnessed a Christian-Muslim clash
  • 46% had family member killed in Christian-Muslim clash
  • Stark segregation
  • Political and economic inequality

Scacco

Results 1: Negative Sentiment


  • Note: Effects stronger for politically dominant Muslim group

Scacco

Results 2: (Weaker) Positive Sentiment

  • Note: Effects stronger for politically dominant Muslim group

Scacco

What have we learned?

  • Current focus on divisive effects of media
  • But: Media interventions can reduce prejudice, threat perceptions and increase intergroup interaction in conflict setting
  • Next stages: Study knock-on effects of greater acceptance by politically dominant groups for social inequalities

Scacco

Motivation

  • How does horizontal contestation affect outcomes at the vertical level?

  • Three forms of horizontal contestation: gender, group based, and the interaction between the two (intersectional discrimination)

  • Vertical outcomes: How does discrimination affect the choice and quality of political representation?

  • How are these moderated by institutional features commonly used to tackle discrimination?

    • Gender quotas
    • Information on ability
    • Collective choices

Foong, Humphreys & Kasara

Challenges

  • Self-selection into candidacy
  • Endogenous selection of institutions
  • Mixed incentives

These all make causal inferences from observational patterns difficult

Foong, Humphreys & Kasara

Design 1

  • We conduct a candidate choice experiment in Nairobi

  • Participants were asked to pick candidates to perform a task

  • Controlled setting features dual aspects of elections:

    • direct rewards to candidates for selection independent of quality
    • direct rewards to electors from selection conditional on quality

Foong, Humphreys & Kasara

Design 2

  • Avatars represent real candidates

Foong, Humphreys & Kasara

Results 1: Who gets chosen?

Term

Beta

SE
Baseline discrimination
outgroup -0.08*** 0.017
female -0.03* 0.014
female * outgroup 0.05 0.037
Institutional effects on discrimination
quota * female 0.14*** 0.016
information * female -0.05** 0.020
discussion * female 0.00 0.037
quota * outgroup 0.02 0.021
information * outgroup 0.02 0.024
discussion * outgroup 0.18* 0.079
Institutional effects on intersectional discrimination
quota * female * outgroup -0.01 0.048
information * female * outgroup -0.04 0.053
discussion * female * outgroup -0.08 0.174

Foong, Humphreys & Kasara

Results 2: How institutions affect candidate quality

Foong, Humphreys & Kasara

What have we learned?

  • Little evidence for causal interactions between ethnicity and gender
  • Evidence that institutions moderate the relationship between contestation and candidate quality:
    • Quotas remove (in fact reverse) gender bias but weaken candidate quality
    • Information hurts women but improves candidate quality
    • Discussions remove only outgroup bias and improve candidate quality

Horizontal biases affect the quality of representation, but biases can be moderated by institutional context

Foong, Humphreys & Kasara