Understanding modeled data

A highlight of TargetEarly is its early and absentee partisan affiliation reporting. For some states, TargetSmart is able to collect official party registration information for voters. In other cases, state governments do not make this information available. When official party registration information is not available, TargetSmart does the next best thing which is to predict the party affiliation of voters. Like any prediction there may be inaccuracies. We are careful to label our predictions as "Modeled" so you can understand this is a prediction of party affiliation.

Party registration and predicted party alignment statistics are not intended to predict votes or election outcome

Voters don't always vote lock step with their party registration/alignment. Please do not mistake TargetSmart's prediction of party alignment with election outcome predictions.

How are modeled Democrat and Republican party alignment determined?

TargetSmart has developed a proprietary partisanship classification model to determine the likely political affiliation of a voter. This predictive model is trained using survey data and incorporates various data points such as vote history, party affiliation (where available), consumer interests, and demographic information to make a prediction of how likely a person is to align with the Democratic party. Those marked asUnaffiliated did not fall into a score range where we could make a confident prediction of affiliation.

The partisan model was constructed using over 56,000 survey respondents, both over the phone and online. Model scores are expressed from 0-100, representing the probability that a person identifies as a Democrat. The model was used to score over 256 million voting age persons nationwide.