TY - JOUR
T1 - Assessing the Accuracy of Google Trends for Predicting Presidential Elections
T2 - The Case of Chile, 2006–2021
AU - Vergara-Perucich, Francisco
N1 - Publisher Copyright:
© 2022 by the author.
PY - 2022/11
Y1 - 2022/11
N2 - This article presents the results of reviewing the predictive capacity of Google Trends for national elections in Chile. The electoral results of the elections between Michelle Bachelet and Sebastián Piñera in 2006, Sebastián Piñera and Eduardo Frei in 2010, Michelle Bachelet and Evelyn Matthei in 2013, Sebastián Piñera and Alejandro Guillier in 2017, and Gabriel Boric and José Antonio Kast in 2021 were reviewed. The time series analyzed were organized on the basis of relative searches between the candidacies, assisted by R software, mainly with the gtrendsR and forecast libraries. With the series constructed, forecasts were made using the Auto Regressive Integrated Moving Average (ARIMA) technique to check the weight of one presidential option over the other. The ARIMA analyses were performed on 3 ways of organizing the data: the linear series, the series transformed by moving average, and the series transformed by Hodrick–Prescott. The results indicate that the method offers the optimal predictive ability.
AB - This article presents the results of reviewing the predictive capacity of Google Trends for national elections in Chile. The electoral results of the elections between Michelle Bachelet and Sebastián Piñera in 2006, Sebastián Piñera and Eduardo Frei in 2010, Michelle Bachelet and Evelyn Matthei in 2013, Sebastián Piñera and Alejandro Guillier in 2017, and Gabriel Boric and José Antonio Kast in 2021 were reviewed. The time series analyzed were organized on the basis of relative searches between the candidacies, assisted by R software, mainly with the gtrendsR and forecast libraries. With the series constructed, forecasts were made using the Auto Regressive Integrated Moving Average (ARIMA) technique to check the weight of one presidential option over the other. The ARIMA analyses were performed on 3 ways of organizing the data: the linear series, the series transformed by moving average, and the series transformed by Hodrick–Prescott. The results indicate that the method offers the optimal predictive ability.
KW - ARIMA
KW - Chile
KW - elections
KW - forecasting
KW - time series
UR - http://www.scopus.com/inward/record.url?scp=85141566821&partnerID=8YFLogxK
U2 - 10.3390/data7110143
DO - 10.3390/data7110143
M3 - Article
AN - SCOPUS:85141566821
SN - 2306-5729
VL - 7
JO - Data
JF - Data
IS - 11
M1 - 143
ER -