The Project
About the Project
The California Proposition Predictor is a data-driven tool for understanding statewide ballot measures — built to help voters make more informed decisions before Election Day.
Get Involved
This project is nonpartisan and independent. You can get involved by verifying our data sources, sharing the tool before elections, or using our open data for your own research.
Limits & Disclaimer
Predictions are generated by a statistical model trained on historical California ballot proposition data. They are not guaranteed outcomes and should not be treated as such.
Factors that cannot be fully modeled include: late-breaking news events, candidate endorsements, grassroots mobilization, and shifts in voter sentiment close to Election Day.
Campaign finance data may lag real-time spending by several weeks due to Cal-Access reporting schedules.
Methodology
Our prediction model weighs several key factors:
- Historical pass rates — By category and ballot wording type
- Campaign finance ratio — Support vs. opposition spending from Cal-Access
- Polling data — When available and close to the election
- Semantic similarity — ML-based matching to past propositions via embeddings
- Demographic modeling — Census ACS data mapped to county-level patterns
The model is a logistic regression ensemble calibrated on 1980–2024 California ballot proposition outcomes across >1,200 measures.
Browse proposition data →Meet the Team
Built by a group of data scientists and civic technologists interested in making California ballot information more accessible and transparent to voters.
Data Science
Prediction model & statistical analysis
Engineering
API integrations & frontend development
Policy Research
Historical proposition analysis & categorization
This is a civic technology project. Not affiliated with any political party or campaign.