Open Model Project

If polls are unreliable, then election models based on polls are unreliable.

Help us create the world's first open-source, collaborative election forecasting model at OpenModelProject.org.

About OMP

The Open Model Project is an initiative of American Civics Exchange, aiming to bring crowd-based wisdom to traditional election modeling.

We’ve adopted a radically transparent and messily iterative approach to election modeling, giving everyone free and open access to the model and encouraging you to spin the dials, pull the levers, and see how it actually works.

And we’re inviting you to help us make it better.

We don’t pretend to have a recipe for the perfect election forecast. But we do recognize that the current state of the art is sorely lacking.

The opaque, overly poll-credulous professional models (with ever shifting standards of self-evaluation) have fooled us twice in a row, so shame on us. OMP is the antithesis to the centralized, black box approach to forecasting.

While we use polls as a key input, our ever evolving, transparent, crowd-driven approach favors a mosaic approach, drawing on predictive insights wherever we may find them, in hopes of triangulating more reliable, less rigid probabilistic forecasts.

In addition to polling data, we may draw on economic and public health fundamentals, prediction market pricing, early/mail voting patterns, new voter registration data, fundraising activity, social media influence, and so on.

And we’re focused on building tools that enable people with zero coding skills to make their own tweaks, apply their own assumptions, and test the model’s logic and sensitivities. And when you find something that doesn’t make sense, we ask you to help us improve it.

Prototypes: Georgia Senate, NYC Mayor

We used the twin Georgia Senate runoff elections on January 5th as a proof of concept. We  built the kernel of the model, some visualization and interaction tools for you to play with, and gave you some forums to debate and collaborate what works and what doesn’t.

And to better facilitate the Georgia prototype, we also commissioned our own poll of Georgia voters, which we published with the same focus on transparency and interactivity. Every poll makes use of (sometimes subjective) weights to normalize the demographic or ideological composition of the respondents. Pollsters often make additional judgments about the likelihood of their respondents to actually vote- another opportunity to enhance the accuracy of the poll, but another opportunity for subjectivity to unduly influence the results.

Our interactive poll explorer not only links you straight to the raw data, but offers you user-friendly ways to change any and all of the demographic weights and likely voter screening factors.

Once you’ve had a chance to check it out, we hope you’ll join the conversation - whether as a collaborator, a lurker, or a leader of the Open Model Project.