Chronology of key moments in the history of women's rights in Italy

Following Goldin's methodological framework from "Why Women Won," this dataset provides a structured chronology of key moments in the history of women's rights in Italy. Events are classified into four categories: W – workplace regulation; P – political milestone; E – economic/social; B – bodily autonomy. The dataset is used and analysed in our corresponding article "Studying the gender gap à la Claudia Goldin: a historical perspective on women's work in Italy." Version 1 (dataset_italian_events.csv, n = 112, 1867–2022) is a manually compiled chronology. Sources include Camera dei Deputati (2021), Anne Cova (2022), and the websites of Noi Donne, Il Consiglio Nazionale delle Donne Italiane, Unione Femminile Nazionale, and Istituto Luigi Sturzo. Note: As Italy was a founding member of the EU, some events relate to EU directives with impact on Italy for which no implementing legislation could be identified. Version 2 adds dataset_italian_events_extended.csv (n = 3,181, 1800–2022), which combines the original manual dataset with 3,069 events extracted automatically from eight Italian primary sources using a two-stage large language model (LLM) pipeline (for more details, see our accompanying academic paper). In Stage 1, text passages were summarised into English event descriptions using a local instance of Ministral-3b; in Stage 2, summaries were converted into structured JSON records following a fixed schema with four fields: year, category code, description, and confidence level. The pipeline follows the methodological framework of Benoit et al. (2026). A sourcecolumn in the extended dataset distinguishes manually coded rows (manual) from LLM-extracted rows (llm_extracted). Validation: Prompts for event extraction were developed iteratively with manual inspection of outputs at each stage. Once the final prompting scheme was established, 50 randomly selected rows from the LLM-extracted events were checked manually against the source texts. Accuracy was high: 86% of entries were judged correct on both category assignment and event description.