We apply large language models (LLMs) to classify 6,531 speeches from the UK Parliament (Hansard, 1803‑2005) according to the speaker’s stance on women’s suffrage and political representation. The classified texts are then annotated for sexist content using the Ambivalent Sexism Inventory framework.
In parallel, we release an organized, metadata‑enriched version of the Hansard Corpus containing 6.7 million speeches across 1.2 million debates, with an 89% gender‑matching rate for House of Commons MPs.
Our analysis finds that 54% of anti‑representation speeches contain sexist language, compared with only 21% of pro‑representation speeches. The two camps employ fundamentally different sexism: anti‑suffrage rhetoric mixes hostile and benevolent sexism, whereas pro‑suffrage sexism is overwhelmingly benevolent.
Gender‑wise, female MPs support women’s political rights at a 93% rate, while male MPs do so at 70%; this gap narrows only after enfranchisement. These results demonstrate that benevolent and hostile sexism are deployed in distinct rhetorical contexts, consistent with ambivalent sexism theory.
Review