Research

Working papers

Hiring a Look-Alike: Evidence of Preference-Based Homophily in Corporate Law Firms

Abstract: Managers tend to hire workers who are similar to themselves. It is not well understood, however, whether this is due to preferences or productivity concerns. This paper presents a new approach for measuring homophily in organizations that is due to preferences by analyzing the facial similarity between managers and their associates. I construct a novel data set of around 3,600 lawyers who work in 20 of the largest corporate law firms in Germany to show that law firm partners look more facially similar to the associates in their own division (same firm, city, and field of law) than to comparable associates in the same firm. However, this is only true for associates without job experience, suggesting that facial homophily may play a role in the hiring process only when candidate information is scarce.

On Self-Organization: The Formation of Volunteer Fire Brigades in 19th-Century Baden][On Self-Organization

(with Tristan Stahl)

Abstract: How does self-organization emerge? We study the formation of a formal institution for local public goods provision in 19th-century Baden, which at the time was an autonomous state in what is now Germany. During this period, many, but not all, communities established volunteer fire brigades, a form of bottom-up collective action that persists to this day. We find that self-organization emerged where local shocks interacted with pre-existing social structures. Specifically, a major fire significantly increased the likelihood of establishing a fire brigade, but only in communities that already had a singing or gymnastics club.

Publications

CultureBERT: Measuring Corporate Culture With Transformer-Based Language Models

(with Stefan Pasch)

2023 IEEE International Conference on Big Data (BigData)

Abstract: This paper introduces transformer-based language models to the literature measuring corporate culture from text documents. We compile a unique data set of employee reviews that were labeled by human evaluators with respect to the information the reviews reveal about the firms’ corporate culture. Using this data set, we fine-tune state-of-the-art transformer-based language models to perform the same classification task. In out-of-sample predictions, our language models classify 17 to 30 percentage points more of employee reviews in line with human evaluators than traditional approaches of text classification. We make our models publicly available. 

Our language models are publicly available on Hugging Face. For an explanation on how to apply them, check out our GitHub tutorial.