Using Google Searches to Predict Voter Registration
A recent paper in the journal Political Analysis attempts to estimate the effects of voter registration deadlines, finding that 3 million to 4 million more people would have registered to vote in the 2012 presidential election if the states that did not allow Election Day registration had offered it.
The data included each state’s daily Google search volume for “voter registration” and related variations, as well as the daily registration totals from 16 states. These two measures show a strong correlation, and both peaked on the states’ respective registration cutoff dates. However, a surge in searches in the waning days of the campaign and on Election Day suggests that many people were interested in registering after the deadlines had passed.
Researchers used the information to estimate the total registrations filed in each state and then applied those findings to the number of searches conducted after the registration cutoffs to estimate the potential additional registrants.
The paper also demonstrates some of the challenges of working with search data. In the case of Google, the volume of searches for voter registration information in each state is based on where the query originated. In the District of Columbia, the data were abnormally high because of the large number of out-of-state commuters who work in the city, and ultimately the data had to be discarded. Further research in this area will help determine if the overall estimates are similarly high or if they are a suitable gauge of demand.