What Modern democracy relies on representation, but elected representatives do not look like citizens at large. Legislatures in Western democracies are disproportionately male, educated, and otherwise unrepresentative, creating a deficit of descriptive representation. Origins of Descriptive Representation (ODER) focuses on the first step towards improved descriptive representation, i.e. the first-ever election of a candidate from an underrepresented group, which I refer to as 'breakthrough candidacies'. How does the political system react to breakthrough candidacies? Answering this question is critical for understanding the impediments to improving future descriptive representation for all underrepresented groups. Why ODER focuses on the underrepresentation of women and workers in politics, groups at the heart of a rich literature on descriptive (under)representation. However, most existing studies of gender and class representation are based in contemporary contexts, where group representation is long established (if still imperfect) and sustained by equality norms. As a consequence, we may seriously underestimate the set of challenges faced by present and future breakthough candidacies. ODER sheds new lights on this question by considering reactions to historical cases of breakthrough candidacies. The project enriches our understanding of past and future consequences of changes in descriptive representation, and by extension the set of challenges facing currently underrepresented groups. How The project is structured around three subprojects. Subproject 1 undertakes an ambitious data collection effort, digitizing the universe of candidate-level Danish general election results. We partner with Statistics Denmark to make these data publicly available, radically expanding the data coverage of Danish electoral history. Using candidate names and occupations listed in election records, we can include data on candidate gender and class. Subprojects 2 and 3 focus on descriptive representation of women and worker candidates respectively. In each subproject, we implement so-called politician characteristic regression discontinuity designs, which allow us to credibly estimate the effects of women's and workers' breakthrough candidacies on a host of outcomes. Specifically, we consider effects on voter support in subsequent elections as well as the nomination strategies and political rhetoric of winning parties and their competitors.