When Random Assignment Is Not Enough for Causal Evidence
by Marco Schmandt, Constantin Tielkes, and Felix Weinhardt
Focus and Research Question
The paper examines whether studies that rely on random placement — the random assignment of people to places or groups — can still produce biased results when estimating the effects of local conditions or group characteristics. This matters because random placement is often treated as a gold standard for causal evidence in economics.
Core Idea
The authors show that random placement alone does not guarantee unbiased estimates of local factors, because people are assigned to places, not to specific local characteristics like unemployment or social attitudes. As a result, estimates can mix causal effects with hidden biases.
Data and Setting
The framework is tested using administrative data on more than 69,000 refugees in Germany, who were initially assigned to counties under a random dispersal policy. The data track individuals over time and capture all later moves, which is crucial for studying mobility bias.
🎯 Does random assignment really guarantee unbiased results?
👥 A new study by @marcodavis94.bsky.social, Constantin Tielkes, and Felix Weinhardt shows why this common assumption can be misleading.
Read more 📑 berlinschoolofeconomics.de/about-us/new...
#economicresearch #evidencebasedpolicy