Download this Paper Open PDF in Browser

Using Google Data to Predict Who Will Vote

27 Pages Posted: 26 Mar 2013  

Seth I. Stephens-Davidowitz

Harvard University - Department of Economics; Google Inc.

Date Written: March 24, 2013

Abstract

This paper argues that Google searches prior to an election can be used to predict turnout in different parts of the United States. For the 2008 and 2010 election, October search rates for "vote/voting," compared to four years earlier, explained 20-40 percent of state-level change in turnout rates. Out-of-sample predictions made prior to the 2012 election were strong. The data might prove useful in predicting candidate performance in midterm elections. If turnout is predicted to be high, the Democratic candidate can be expected to do better than the polls suggest. For presidential elections, the data can be useful in estimating the composition of the electorate, by comparing media market search rates to media market demographics. In the 2008 election, the Google data would have correctly predicted substantially increased African-American turnout. The out-of-sample 2012 demographics predictions using Google data were largely correct. It correctly forecast elevated Mormon turnout. It correctly forecast, contrary to some pollsters’ predictions, that African-American, Hispanic, and youth turnout rates would remain at 2008, rather than 2004, levels.

Keywords: voting, forecasts, Google

Suggested Citation

Stephens-Davidowitz, Seth I., Using Google Data to Predict Who Will Vote (March 24, 2013). Available at SSRN: https://ssrn.com/abstract=2238863 or http://dx.doi.org/10.2139/ssrn.2238863

Seth I. Stephens-Davidowitz (Contact Author)

Harvard University - Department of Economics ( email )

Littauer Center
Cambridge, MA 02138
United States

Google Inc. ( email )

1600 Amphitheatre Parkway
Second Floor
Mountain View, CA 94043
United States

Paper statistics

Downloads
77
Rank
265,174
Abstract Views
517