America's Next Top (Statistical) Model Hosted By DrivenData


Resources for Presidential Election Forecasting

There are many, many, many people and organizations that are interested in creating election forecasts for US Presidential elections. Given the interest, there is lots of data available and many people using these models to predict the outcome of the election.

For the most part, models rely on two primary data sources: polls and fundamentals. Polls are when a sample of the population is called on the phone or asked on the internet who they would vote for. These polls are conducted by polling organizations and the results published. Most models today incorporate the results from many different polls. Fundamentals are information about the population of a state: e.g., demographic information, income, spending, the stock market. By combining these data, models come up with election forecasts.

Your goal is to use whatever data sources you want to predict the exact percentages for each candidate, for each state in the US.

Background information

Existing predictions

Other possibly useful data sources

Ideas for other datasets

  • Social media posts, interest levels, and engagement
  • News media mentions, articles, and sentiment
  • Campaign contributions and spending