Title: Campaign Finance Data and the 2008 Elections
Members: Kathryn Hurley, Jay Arbiv, Alison Flint, Kumar Garapaty, Patricia Carbajales, Josh Daniel, Mike Dale
Category: Best Insight
We create a model that examines how campaign finance contributions influence election results.
Our base model uses demographic data to predict the percentage of individuals who voted for Obama in 2008 by county. We get this data from the American Community Survey 2006-2010 5-year estimates. We include a range of variables including gender, population, ethnicity, education, and per capita income.
We then compare this to a model that also includes campaign contributions for each candidate. We obtain campaign finance data by geocoding all individual contributions for 2008 using ArcGIS and aggregating data to the county level.
Next, we combine the 2008 election results, demographic data, and campaign finance data in Fusion Tables, and then run two regression models. One is a simple linear regression to on the election result. The other model uses Google's API Predictor to create a prediction model using 80% of the data as a training set. We then use this model to predict the other 20% of the observations.
Our final product will create a visualization showing how the predicted results of our model differ from the actual results in 2008.