Tagged: regression discontinuity design

Direct democracy aggravates anti-immigration policy in Switzerland

Brett Gall recently alerted me to Jens Hainmueller and Dominik Hangartner’s fascinating study of Swiss naturalization decisions. They used changes of laws in different Swiss municipalities for a fairly sound regression discontinuity design to assess the effects of direct democracy on immigration.

Foreigners don’t simply apply to live in Switzerland. They apply to live in specific Swiss municipalities. Historically, some municipalities used representative democracy to admit immigrants, but some managed immigration requests via direct democracy, by actually sending the resumes of all applicants to all citizens of the municipality to approve or reject.

Hainmueller and Hangartner’s methodology is stunningly beautiful. There was a series of rulings by the Swiss Federal Court from 2003 to 2005 that required different municipalities to transition from direct democracy to representative democracy for their immigration decisions. The immigration application process takes about 4-5 years, so applicants weren’t able to anticipate any institutional changes to the municipalities to which they had applied.

Hainmueller and Hangartner collected panel data of different Swiss municipalities forced to transition, and used a regression discontinuity design to compare similar applicants whose immigration applications were processed, by the same municipalities, at almost the same time, via direct democracy or representative democracy.

The least xenophobic municipalities that switched didn’t see any change in immigration patterns. The most xenophobic municipalities that switched drastically increased their amount of immigration.

Why does this happen? Xenophobic citizens voting privately don’t need to justify their xenophobia to anyone. In contrast, public servants facing public scrutiny hesitate to reject applications on the basis of poorly reasoned xenophobia.