A large body of migration literature uses random dispersal policies (RDP) to estimate the importance of local factors for integration.
This paper demonstrates that RDP is not sufficient for causal identification for two reasons. First, while RDP ensures that local conditions are exogenous to immigrant characteristics, they still correlate with other observed and unobserved local factors. Second, onward mobility requires careful consideration, as it can be endogenous to factors at the initial location. We theoretically show that estimates from continuousinstruments based on RDP contain three components: the causal effect of interest, ”multiple-treatment bias” (MTB), and ”mobility bias” (MB). The extent of these biases depends on the interrelations of local factors and onward mobility, which can be partly observed. We empirically investigate these biases using novel administrative data from Germany that cover the universe of all refugees between 2013 and 2018 and feature random dispersal.
The central empirical finding is that estimates that ignore MB and MTB cannot be compared and can even change signs.
Joint: Marco Schmandt, Constantin Tielkes, Felix Weinhardt
Date
12.11.2025
, 11 a.m. until noon
Venue
Institute for Employment Research
Regensburger Straße 104
90478 Nürnberg
Room Re100 E10
or online via MS Teams
Registration
Researchers who like to participate, please send an e-mail to IAB.Colloquium@iab.de