Working papers:
Land, power, and property rights: Conjoint evidence from Senegal and Cote d’Ivoire [Link to paper]
Only 15 percent of households in rural Africa possess formal titles for their agricultural landholdings, despite the availability of titles. This paper explores whether households’ distrust in formal institutions constraints the perceived returns to land tenure security. I administer a field conjoint experiment to 1,965 household heads across rural Senegal and Cote d’Ivoire, in which respondents selected which of a pair of profiles would win a hypothetical land dispute. Across levels of trust in both formal and informal institutions, a profile having a formal land title consistently dominates other conjoint profiles. After accounting for switching error, respondents selected a conjoint profile with a title to win the hypothetical land dispute almost 100 percent of the time. A subset of respondents in Cote d’Ivoire provide similar answers when asked how likely households would be to prevent government expropriation. I also estimate a structural topic model using free-response answers to the paired conjoint experiment which shows how respondents are more likely to discuss the profile’s sex and level of investment on the land when titles do not distinguish profiles. This research suggests that neither cultural idiosyncrasies surrounding land ownership nor distrust of titles explain the puzzling lack of land tenure formalization across sub-Saharan Africa.
Getting to youths: development programming, conflict resolution, and political violence in Niger [with Kathryn M. Lance, Ryan Sheely, and Ifeoluwa M. Olawole ]. [Link to paper]
What types of interventions are effective at reducing young people’s participation political violence? A growing literature evaluates a diverse array of economic, civic engagement, and psychosocial interventions but little is known about how different types of interventions interact. This paper presents a cluster-randomized control trial in the Maradi and Tillaberi regions of Niger. In the context of a broader vocational training and civic engagement program, young people (aged 15-34) in randomly assigned villages participated in trainings in Interest-Based Negotiation and Mediation (IBNM), an approach to conflict management. In a follow-up survey of 1,734 youth across 118 villages, we show that youth who participated in both Youth Connect and IBNM training are less likely to support violence than youth who only participated in Youth Connect a year after IBNM trainings took place. We find no difference between youth who participated in Youth Connect by itself and a pure control group. A difference-in-differences approach using geo-referenced conflict incidents corroborates these results: villages which received both Youth Connect and IBNM experienced fewer violent incidents than villages which only received Youth Connect. This research evaluates a promising intervention which is low-cost, light-touch, and can be layered atop conventional violence prevention programming. I have been invited to revise and resubmit this paper to the American Political Science Review.
Intra-respondent reliability and enumerator quality in field conjoint experiments [Link to paper]
Conjoint experiments ask respondents to consider multiple `treatments’ simultaneously, leading respondents to make errors in their responses. Field conjoint experiments, which are often administered by teams of enumerators, introduce another error point because enumerators may be poor quality. However, calculating enumerator intra-respondent reliability (IRR) at the enumerator level provides researchers a tool to monitor enumerator performance specifically for the conjoint experiment. Other common proxies for enumerator quality do not correlate with IRR, which stabilizes at the enumerator level after as few as 15 completed surveys.
Pitfalls and tradeoffs in measuring support for violent extremism: Evidence from Niger and Burkina Faso [with Ryan Sheely and Adam Lichtenheld] [Link to paper]
Academics, policymakers, and other researchers use a mixture of disparate strategies to measure growing support for violent extremism in the Sahel and elsewhere. This paper leverage the confluence of five different measures of support for violence to explore differences in how each captures the underlying phenomenon. These measurements include an original survey on support for violence among 1,772 youth in Niger and Burkina Faso, qualitative rankings of village-level vulnerability to violent extremism by local elites, and data from ACLED. Together, this data allows us to 1) provide insight into validly and reliably measuring violent extremism – a challenge for scholars and policymakers – and 2) explore the extent to which commonly used measurement strategies capture the same underlying phenomena.
In Vi(vi)no Veritas? Expertise, review accuracy and reputation inflation [with Rebecca Janssen] [Link to paper]
Review systems including quantitative measures as well as text-based expression of experiences are omnipresent in today’s digital platform economy. This paper studies the existence of reputation inflation, i.e. unjustified increases in ratings, with a special focus of heterogeneity between experienced and non-experienced users. Using data on more than 5 million reviews from an online wine platform we compare consistency between numerical feedback and textual reviews as well as sentiment measures. We show that overall the wine platform displays strongly increasing numerical feedback over our time period from 2014 to 2020 while this is not the case for our control measures. This gap appears to be even stronger for users with less experience or expertise in wine reviewing. We conclude, that online platforms as well as potential customers should be aware of the phenomenon of reputation inflation and simplifying feedback to one number might do a disservice to review platforms goal of providing a representative quality assessment.