This is a joint work with Judith Kas.

Abstract: Social science research has produced evidence of an 'integration paradox': Perceived discrimination among immigrants and their descendants is stronger, the better their access to mainstream middle-class society as broadly indicated by their education, length of residence, or generational status. Social scientists have proposed several explanations of these counterintuitive patterns and produced a rich assembly of mixed empirical evidence. To identify which reliable insights this line of research has produced across the various samples and measures used, this study presents the first (pre-registered) meta-analytic review of the integration paradox. As a review, the first contribution of this study is to systematize key theoretical arguments put forward in the integration paradox literature into a coherent theoretical synthesis. We derive from this synthesis six hypotheses about the integration paradox, about the processes driving it, and about two moderating conditions. As our second contribution we then assess the reliability of the evidence in favor of these hypotheses using recent developments in multilevel meta-analysis. The pre-registered pilot results imply that the core claims of the integration paradox are reliable, especially driven by increased social exposure to mainstream members, and not the product of publication bias. That said, we also identify a fundamental flaw: Prior research suggests systematic under-perceptions of discrimination among the less integrated but lacks a measure of actual discrimination to assess this assumption. I will conclude my talk by elaborating on a new (trust game) experimental research design that overcomes this problem and allows to measure individual-level mis-perceptions of discrimination. We currently implement this novel experimental design among a random sample 3,000 persons living in one of Germany's five largest cities.