**Principal investigators:** **Craig Burnett** University of North Carolina at Wilmington Email: burnettcm@appstate.edu Homepage: http://people.uncw.edu/burnettc/_/Home.html **Vladimir Kogan** The Ohio State University Email: kogan.18@osu.edu Homepage: http://polisci.osu.edu/people/kogan **Sample size:** 6101 **Field period:** 3/5/2010-7/13/2010 **Abstract** Can elected officials and powerful interest groups "steal the initiative" by shaping the language of a ballot measure's official title and summary to favor their preferred policy outcome? We explore this question using a survey experiment based on three actual measures that have appeared on the ballot in several states. We find that ballot framing does indeed have the potential to influence outcomes in direct democracy elections, including measures dealing with "easy" issues. We also show, however, that the effect of framing is far from absolute. Exposure to campaign information — in particular, endorsements from prominent interest groups — greatly attenuates the effect of a ballot language framing. Our results suggest that, in campaign environments where millions of dollars in advertising and direct appeals from political parties and other elites bombard voters, ballot measure framing are unlikely to dramatically shift voter sentiment, although framing does have the potential to alter outcomes in particularly close elections. **Hypotheses** The language used to describe a ballot measure will influence individuals' vote choices. **Experimental Manipulations** We presented two versions for each of the three measures, emphasizing different frames embedded in the ballot title and summary. Each subject was randomly assigned to see one of these two versions for each ballot measure. To make our results as realistic as possible, we based each frame on an actual ballot title and summary that voters in some states saw in a previous election. The randomization was independent for each measure, so a subject's assignment to see one frame for the first ballot measure did not determine which versions of the other measures she would see as the survey progressed. **Outcomes** Vote choice on three ballot measures **Summary of Results** Our results suggest that framing the ballot title and summary can have significant and substantively large effects on voters' decisions and, potentially, election outcomes. We also showed, however, that campaign information can greatly attenuate these framing effects. Our conclusion is that savvy political actors can indeed steal the initiative before voters ever see the measure in the voting booth, but the existence of vigorously contested and well-funded campaigns and the type of issues that appear on the ballot greatly limit the frequency with which this type of theft actually occurs. **References** Familiar Choices: Reconsidering the Institutional Effects of the Direct Initiative (Forthcoming at State Politics and Policy Quarterly). Available here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1978509 The Case of the Stolen Initiative: Were the Voters Framed? (Working paper). Available here: http://papers.ssrn.com/sol3/papers.cfm?abstract_id=1643448