Do Words Matter? Gendered Language and Sentiment Polarity in U.S. News Headlines Across #MeToo and COVID-19
(2025) DABN01 20251Department of Economics
Department of Statistics
- Abstract
- This thesis investigates how women are portrayed in U.S. news headlines through the lens of gender bias and sentiment polarity. Using a dataset of over 130,000 headlines from major American news outlets between 2005 and 2021, the study combines a curated dictionary-based approach to detect stereotypical language with transformer-based sentiment analysis to measure tone. Two major societal disruptions, the #MeToo movement and the COVID-19 pandemic, are examined as focal events for structural change in media language. Time series models, including changepoint detection and ARMA forecasting, are employed to test for statistically significant shifts in bias and polarity. The results show that while #MeToo increased the visibility of women’s... (More)
- This thesis investigates how women are portrayed in U.S. news headlines through the lens of gender bias and sentiment polarity. Using a dataset of over 130,000 headlines from major American news outlets between 2005 and 2021, the study combines a curated dictionary-based approach to detect stereotypical language with transformer-based sentiment analysis to measure tone. Two major societal disruptions, the #MeToo movement and the COVID-19 pandemic, are examined as focal events for structural change in media language. Time series models, including changepoint detection and ARMA forecasting, are employed to test for statistically significant shifts in bias and polarity. The results show that while #MeToo increased the visibility of women’s stories, it had limited impact on the underlying tone or bias of media coverage. In contrast, the COVID-19 pandemic introduced greater volatility and emotional intensity in headlines, reflecting more pronounced editorial shifts. These findings suggest that despite increased representation, the linguistic framing of women remains persistently stereotyped and emotionally narrow, limited to conventional portrayals such as caregiving, vulnerability, or victimhood, with little room for agency, anger, or authority. The thesis contributes to literature on feminist media studies, computational linguistics, and event inference by offering a replicable framework for measuring media bias over time. Future research could examine how women in leadership or professional roles are portrayed in business and economic news, exploring shifts in bias around female CEOs, politicians, or entrepreneurs. This could reveal how media framing influences perceptions of competence and authority, and whether it correlates with real-world outcomes like hiring trends or board representation. Further work could also extend the analysis to social media platforms to build a more comprehensive view of gender representation in digital media. (Less)
Please use this url to cite or link to this publication:
http://lup.lub.lu.se/student-papers/record/9192244
- author
- Ordish, Sophie Kristina LU
- supervisor
- organization
- course
- DABN01 20251
- year
- 2025
- type
- H1 - Master's Degree (One Year)
- subject
- keywords
- Gender bias, news media, sentiment analysis, time series, computational linguistics
- language
- English
- id
- 9192244
- date added to LUP
- 2025-09-12 09:06:04
- date last changed
- 2025-09-12 09:06:04
@misc{9192244,
abstract = {{This thesis investigates how women are portrayed in U.S. news headlines through the lens of gender bias and sentiment polarity. Using a dataset of over 130,000 headlines from major American news outlets between 2005 and 2021, the study combines a curated dictionary-based approach to detect stereotypical language with transformer-based sentiment analysis to measure tone. Two major societal disruptions, the #MeToo movement and the COVID-19 pandemic, are examined as focal events for structural change in media language. Time series models, including changepoint detection and ARMA forecasting, are employed to test for statistically significant shifts in bias and polarity. The results show that while #MeToo increased the visibility of women’s stories, it had limited impact on the underlying tone or bias of media coverage. In contrast, the COVID-19 pandemic introduced greater volatility and emotional intensity in headlines, reflecting more pronounced editorial shifts. These findings suggest that despite increased representation, the linguistic framing of women remains persistently stereotyped and emotionally narrow, limited to conventional portrayals such as caregiving, vulnerability, or victimhood, with little room for agency, anger, or authority. The thesis contributes to literature on feminist media studies, computational linguistics, and event inference by offering a replicable framework for measuring media bias over time. Future research could examine how women in leadership or professional roles are portrayed in business and economic news, exploring shifts in bias around female CEOs, politicians, or entrepreneurs. This could reveal how media framing influences perceptions of competence and authority, and whether it correlates with real-world outcomes like hiring trends or board representation. Further work could also extend the analysis to social media platforms to build a more comprehensive view of gender representation in digital media.}},
author = {{Ordish, Sophie Kristina}},
language = {{eng}},
note = {{Student Paper}},
title = {{Do Words Matter? Gendered Language and Sentiment Polarity in U.S. News Headlines Across #MeToo and COVID-19}},
year = {{2025}},
}