felix iglhaut *

We demonstrate that common results from the literature of discrete choice modeling based on the Gumbel (extreme value type I) distribution can be extended to the Fréchet (extreme value type II) and Weibull (extreme value type III) distributions. We focus on choice models based on multivariate distributions allowing for dependence between structural errors. Here, we generalize existing results to accommodate even more flexible correlation patterns and multivariate distributions. Examples of explicit choice models to be employed in economic modeling are provided. Finally, we demonstrate how to flexible derive correlated choice models and provide closed form solutions for quantities of interest important for welfare analysis.

We examine the degree to which individuals sort into politically homogeneous neighborhoods in the US between 2010 and 2014, comparing sorting along partisan preferences to those on income and ethnicity. Although we refrain from a causal interpretation, i.e. assuming that individuals proactively engage in neighborhood searching based on either dimension, we hypothesize their behavior may mirror this to a still unknown, potentially correlated degree. We make two main contributions. First, we are able to document that individuals sort along party affiliations, and such a sorting pattern cannot be revealed by coarser aggregated data, explaining some of the contradicting findings in the literature. Second, we show that partisan sorting contributes around 25 percent of the yearly increase in geographic polarization. We contrast this finding to sorting on income and ethnicity. For income inequality within neighborhoods, the result is more nuanced: in line with partisanship, sorting would lead to more segregation, despite the overall shift (taking also within-neighborhood changes into account) showing more income inequality within neighborhoods, i.e. less segregation. In contrast, when it comes to race and ethnicity, sorting leads to the opposite by amplifying the reduction in segregation.
Our ongoing work in progress is focused on our second contribution: assessing how large the overlap between these dimensions is and to reveal the heterogeneous response of sorting along either dimension to individual level characteristics. For this purpose, we estimate a random-coefficient random utility model of neighborhood choice in which we use machine learning techniques to construct individual-specific choice sets. Based on the results, we want to assess to what degree sorting along all three dimensions exists and the degree to which these patterns are correlated.

We study the role of perceived threats from other cultures induced by terrorist attacks and criminal events on public discourse and support for radical-right parties. We develop a rule which allocates Twitter users to electoral districts in Germany and use a machine-learning method to compute measures of textual similarity between the tweets they produce and tweets by accounts of the main German parties. Using the exogenous timing of attacks, we find that, after an event, Twitter language becomes on average more similar to that of the main radical-right party, AfD. The result is driven by a larger share of tweets discussing immigrants and Muslims, common AfD topics, and by a more negative sentiment of these tweets. Shifts in language similarity are correlated with changes in vote shares between federal elections. These results point to the role of perceived threats from minorities on the success of nationalist parties.