Gender Stereotypes in User-Generated Content

Anna Kerkhof and Valentin Reich

No 10578, CESifo Working Paper Series from CESifo

Abstract: Gender stereotypes pose an important hurdle on the way to gender equality. It is difficult to quantify the problem, though, as stereotypical beliefs are often subconscious or not openly expressed. User-generated content (UGC) opens up novel opportunities to overcome such challenges, as the anonymity of users may eliminate social pressures. This paper leverages over a million anonymous comments from a major German online discussion forum to study the prevalence and development of gender stereotypes over almost a decade. To that end, we develop an innovative and widely applicable text analysis procedure that overcomes conceptual challenges that arise whenever two variables in the training data are correlated, and changes in that correlation in the prediction sample are subject of examination themselves. Here, we apply the procedure to study the correlation between gender (i.e., does a comment discuss women or men) and gender stereotypical topics (e.g., work or family) in our comments, where we interpret a strong correlation as the presence of gender stereotypes. We find that men are indeed discussed relatively more often in the context of stereotypical male topics such as work and money, and that women are discussed relatively more often in the context of stereotypical female topics such as family, home, and physical appearance. While the prevalence of gender stereotypes related to stereotypical male topics diminishes over time, gender stereotypes related to female topics mostly persist.

Keywords: gender bias; gender stereotypes; natural language processing; machine learning; user-generated content; word embeddings (search for similar items in EconPapers)
JEL-codes: C55 J16 (search for similar items in EconPapers)
Date: 2023
New Economics Papers: this item is included in nep-big and nep-gen
References: Add references at CitEc
Citations:

Downloads: (external link)
https://www.cesifo.org/DocDL/cesifo1_wp10578.pdf (application/pdf)

Related works:
This item may be available elsewhere in EconPapers: Search for items with the same title.

Export reference: BibTeX RIS (EndNote, ProCite, RefMan) HTML/Text

Persistent link: https://EconPapers.repec.org/RePEc:ces:ceswps:_10578

Access Statistics for this paper

More papers in CESifo Working Paper Series from CESifo Contact information at EDIRC.
Bibliographic data for series maintained by Klaus Wohlrabe ().