Trending in the Right Direction: Using Google Trends Data as a Measure of Public Opinion During a Presidential Election

During the 2016 presidential election, public opinion polls consistently showed a lead in the popular vote and Electoral College for Hillary Clinton over Donald Trump. Following Trump's surprise victory, the political pundits and public at large began to question the accuracy of modern public opinion polling. Fielding a representative sample, convoluted and opaque methodologies, the sheer amount of polls, and both the media's and general public's inability to interpret poll results are among the flaws of the polling industry. An alternative or supplement to traditional polling practices is necessary. This thesis seeks to investigate whether Google Trends can be effectively used as a measure of public opinion during presidential elections. This study gathers polling data from the 2016 presidential election from states that were considered swing states. Specifically, this study examines six total polls, three from states that swung in the way the polls predicted they would – Nevada and Virginia – and three from states that swung against the prediction – Michigan, Wisconsin, and Pennsylvania. Answers to the "Most Important Issue" question in each poll are compared to their corresponding topics in Google Trends by calculating Pearson product moment correlations for each pair. Results indicated that in states that swung as predicted, Google Trends was an effective supplement to traditional public opinion polls. In states that did not swing as predicted, Google Trends was not an effective supplement. Implications of these results and future considerations for the polling industry and Google are discussed.