Intergroup contact and avoidance

Outgroup Avoidance in Online Dating: Evidence from a Large-Scale Field Experiment
  [Preregistration]  [PAP]

Recent meta-analyses suggest that, while contact can reduce prejudice, its effects are on average small, especially in well-powered field experimental studies with delayed outcome measurement. One of the reasons why intergroup contact may fail to produce larger effect sizes is that individuals prone to avoiding outgroup contact are both less likely to benefit from and less likely to engage in such encounters. Yet, research on the drivers of intergroup avoidance is scarce. In this study, I systematically investigate patterns and drivers of intergroup avoidance in an online dating context. In partnership with large European online dating platform, I first analyze user data to examine the extent and covariates of ethnic bias in swiping behavior. Next, users will be randomly assigned to one of three brief video treatments that are delivered on the platform and designed to test three psychological mechanism that potentially drive intergroup avoidance. Behavioral data is complemented by an endline survey, including measures on intergroup contact preference, intent, and behavior.



Intergroup Contact, Empathy Education, and Refugee-Native Integration: Evidence from a Field Experiment in Lebanon
With Salma Mousa and Alex Scacco
  [Preregistration]  [PAP]

Intergroup contact and empathy education are widely used strategies to foster trust, tolerance, and cooperation across group lines in deeply divided societies. We test these approaches directly against each other in the context of refugee-host relations in Lebanon, where Syrian refugees comprise a quarter of the population. Our study embeds a popular, existing family psycho-social support (FPSS) program targeting vul- nerable youth in a field experiment, where Syrian and Lebanese youth were randomly assigned to either mixed (Lebanese-Syrian) or homogeneous (Syrian or Lebanese) FPSS classrooms, and to an empathy education or placebo curicculum focused on health and nutrition. We find that intergroup contact had little effect on prejudice, and actually somewhat reduced study participant willingness to engage in future intergroup activi- ties — an effect concentrated among Lebanese participants, and amplified when combining contact and empathy education. We find evidence consistent with dissonance, status threat, and negative experiences for dominant-group members. Empathy educa- tion, by contrast, modestly reduced prejudice and increased support for inclusive poli- cies without leading to negative behavioral effects. We find no evidence of spillovers to study participants’ parents. These findings suggest that empathy education may more be more effective than intergroup contact in improving migrant-relations in conflict settings.



“Native Flight” in Urban School Districts? Evidence from Primary School Registration Data in Berlin
With Bernd Beber, Moritz Matzner, and Macartan Humphreys

We investigate parental school choice in Berlin, Germany. Anecdotal evidence suggests that nonmigrant “natives” take steps to avoid or seek out catchment area schools due to the ethnic composition of student bodies. Indeed, school segregation has been documented across Europe, but the lack of high-resolution data on school choice and migrant populations has made it difficult to precisely show the extent to which schools are more (or less) segregated than we would expect given catchment area demographics and to analyze the process that leads to such segregation. We contribute to this debate with an analysis of a never-before-used dataset of 35,000 school choice forms that parents must complete to enroll their child in elementary school. We combine this data with a newly compiled set of administrative data covering Berlin’s several hundred elementary school enrollment zones for the years 2000 to 2018, which includes demographic and socioeconomic zone characteristics, student population indicators, and various school attributes and performance metrics. We explore in particular the extent to which we see Schelling sorting, where even weak preferences for separation from another group (and small changes in zone composition) can dynamically spiral to produce extreme patterns of segregation, and we develop policy implications for how to counteract such dynamics.



LLM and webscraping

Political Connections: Measuring Political and Business Elite Networks
With Lisa Garbe, Yuequan Guo, and Macartan Humphreys

Elite connections shape political and economic outcomes but are difficult to measure. Traditional methods rely on surveys or shared affiliations, which lack scalability and clear interpretation. We propose a scalable, interpretable approach to map elite networks by compiling elite lists, detecting co-appearances in media, coding interactions with large language models (LLMs), and constructing network ties. This complements conventional methods and suits dynamic elite politics. Applied to Germany (2013–2023), our data includes 750 parliamentarians and 1,051 top executives. We collected 42,366 articles from 30 newspapers, identifying co-appearances. Open-source LLMs code interactions such as meetings or conversations, enabling nuanced network definitions. We validate networks qualitatively and quantitatively by linking them to events like elections and comparing them with education-based networks. This method expands elite network research beyond typical scopes, offering scalable, interpretable insights into how political and economic elites connect and influence each other.



Other work

Power Law Distributions in Intrastate Conflict Intensity Dynamics
With Christoph Trinn
Published in Conflict Management and Peace Science, 40(4), 2022
  [Research article]

The discovery of power laws in conflict intensities has spurred numerous explanation attempts. Two different interpretations have persisted: the notion that power laws are spurious results of random processes and the opposing view that power-law distributions attest to endogenous dynamics linked to self-organized criticality (SOC). We substantiate the SOC forest-fire model for intrastate conflicts, conceptualizing conflict potential as social pressure, measured by horizontal inequality. This potential is triggered by infinitesimal events. Their occurrence depends on the interaction density between conflict actors, operationalized as the conjunction of state capacity and non-state governance. In a global analysis of 143 conflict dyads, we find that 40 conform to a power law and 33 to a stretched exponential distribution, the two outcomes predicted by the model. We find evidence that the forest-fire model is a plausible approximation of the dynamics of intrastate conflicts, accounting for both the conformity and the non-conformity to power laws.



Towards Genuine Collaboration in International Research
With Nora Chirikure
Published in WZB Mitteilungen 184, 2024
  [Blog article]

Researchers worldwide continue to face significantly unequal starting opportunities, particularly in terms of research funding, academic career paths, visas, and networking. Nora Chirikure and Lennard Naumann explore what is needed to establish truly equitable South-North research partnerships and to ensure the success of research originating from the Global South.



Attributing agnostically detected large reductions in road CO2 emissions to policy mixes
With Nicolas Koch, Felix Pretis, Nolan Ritter, and Moritz Schwarz
Published in Nature Energy 7, 2022
  [Research article]

Policymakers combine many different policy tools to achieve emission reductions. However, there remains substantial uncertainty around which mixes of policies are effective. This uncertainty stems from the predominant focus of ex post policy evaluation on isolating effects of single, known policies. Here we introduce an approach to identify effective policy interventions in the EU road transport sector by detecting treatment effects as structural breaks in CO2 emissions that can potentially occur in any country at any point in time from any number of a priori unknown policies. This search for ‘causes of effects’ within a statistical framework allows us to draw systematic inference on the effectiveness of policy mixes. We detect ten successful policy interventions that reduced emissions between 8% and 26%. The most successful policy mixes combine carbon or fuel taxes with green vehicle incentives and highlight that emissions reductions on a magnitude that matches the EU zero emission targets are possible.