Research

This paper investigates disparities among objective poverty measures, individuals’ subjective perceptions, and poverty-related social media discourse in US counties. We find that while poverty and perceived poverty are positively correlated, poverty-related social media discourse is unrelated to a county’s level of poverty. We document that the county-level predictors of the three poverty dimensions differ widely, suggesting that poverty-related social media discussions do not take place in the counties that are most affected by poverty or perceive themselves as poor. The paper concludes by highlighting discrepancies in social media discourse, revealing a skewed portrayal of poverty, particularly concerning gender and ethnicity.

Social scientists increasingly use demographically stratified social media data to study the attitudes, beliefs, and behavior of the general public. To facilitate such analyses, we construct, validate, and release publicly the representative DADIT dataset of 30M tweets of 20k Italian Twitter users, along with their bios and profile pictures. We enrich the user data with high-quality labels for gender, age, and location. DADIT enables us to train and compare the performance of various state-of-the-art models for the prediction of the gender and age of social media users. In particular, we investigate if tweets contain valuable information for the task, since popular classifiers like M3 don’t leverage them. Our best XLM-based classifier improves upon the commonly used competitor M3 by up to 53% F1. Especially for age prediction, classifiers profit from including tweets as features. We also confirm these findings on a German test set.

Biases in legal decision making are difficult to identify as type II errors (wrongful acquittals) are hardly observable and type I errors (wrongful convictions) are only observed for the subsample of subsequently exonerated convicts. Our data on the first German soccer league allows us to classify each referee decision accurately as correct, type I error or type II error. The potential bias we are interested in is favoritism towards clubs with higher long-term status, proxied by the ranking in the all-time table at the beginning of each session and by membership. Higher-status clubs benefit largely from fewer type II errors. By contrast, the actual strength of clubs has no impact on referee decisions. We find no difference in type I errors and suggest anticipation of the bias as a potential explanation for the difference. We investigate several mechanisms potentially underlying our results; including career concerns and social pressure.

How does competition affect the entry and selection of politicians? I study this question using data on U.S. Congressional elections between years 1998-2014. My identification strategy levies changes in electoral competition due to redistricting. Difference-in-differences estimates reveal a discrepancy between the electorally dominant and weak party. The average candidate in primary elections of the weak party is more experienced and more likely to descriptively represent their district following an increase in competition. The reverse holds in the strong party. Investigating underlying mechanisms, I find suggestive evidence that candidates respond to preferences of party members, which may matter more in competitive elections.

Trust in politicians can influence government turnover, economic and government performance as well as the demand side of policy-making – voters’ preferences over policies. In this paper I study how a lack of trust in politicians influences the supply side – policy provision. Using data on 63,000 legislative documents, 75,000 individual roll-call voting decisions as well as survey evidence for more than 2,000 candidates in German federal elections between 2009 and 2021, I show that low political trust leads politicians to be less concerned with the provision of many types of public goods - most importantly climate protection. In order to establish causality of these results, I follow an instrumental variable approach. My instrument functions similar to a shift-share instrument and leverages variation in internal migration patterns and differential exposure to common state-level shocks to political trust. An analysis of the underlying mechanism suggests that the results are mostly driven by the selection of different politicians rather than pandering to voters’ preferences.

We study the impact of the opening of refugee reception centers on natives’ social media activity and voting behavior in the Netherlands following the unexpected and large refugee inflow in 2015-2016. Using data on more than 100 million geocoded tweets and a difference-in-differences approach, we document a short-lived surge in refugee salience accompanied by a decrease in expressed support for refugees. We also document a sharp increase in discussions related to religious minorities (Islam) after refugee center openings. To understand how social media salience can affect voting behavior, we next turn to uniquely detailed voting data and find a significant increase in anti-immigration voting in areas (very) near newly established reception centers, with this effect diminishing over time and distance. Finally, we show that increased support for anti-immigration parties is driven by areas where salience of refugees responded particularly strongly, whereas areas with high levels of pre-existing salience of refugees do not see a rise in anti-immigration votes at all.