🚨New working paper: Compositional Effects, Internal Migration and Electoral Outcomes
osf.io/preprints/so...
#polisciresearchWhen studying how internal migration affects changes in district-level electoral outcomes, it is important to separate the compositional effect (heterogeneity in political behavior between movers and stayers) from the exposure effect (change in political behavior due to exposure to migration).
Nov 13, 2025 13:13We use Rubin's potential outcomes framework to formally define compositional effects and to decompose the total causal effect of internal migration into exposure effects and compositional effects.
These two types of effects are conceptually similar but different to natural (in)direct effects in mediation analysis. We discuss how to identify, estimate, and bound the compositional effect from in- and out-migration.
To illustrate the formal results, we study the case of East Germany after reunification, when more than 9% of East Germans moved to former West Germany. How did this exodus affect electoral support for the incumbent coalition government in the 1994 federal election across East Germany?
Using constituency-level electoral data, we show that in East German constituencies with more out-migration between 1990–1994, the incumbent vote share increased. What explains this? Do voters reward the incumbent for high out-migration (sparked by increasing unemployment)? Unlikely.
We demonstrate that there is a compositional effect: using survey data, we show that leavers were 18 pp less likely to support the incumbent than stayers. Thus, in constituencies with more out-migration, the incumbent vote share increased.
