¿Permiten las búsquedas por Internet generar predicciones sobre la afluencia de turistas en Argentina?

Abstract

Ante la incertidumbre económica, la información precisa y oportuna es un recurso crítico pero costoso. Esta investigación analiza el sector turístico en Argentina, utilizando datos de Google Trends para predecir la afluencia de visitantes en siete destinos específicos. El objetivo es evaluar si el uso de estadísticas de búsqueda en internet permite anticiparse a la publicación de datos oficiales. Metodológicamente, se contrastan modelos univariados (ARIMA) con modelos multivariados (VAR), integrando en estos últimos la información del motor de búsqueda. Los resultados evidencian mejoras significativas: el modelo multivariado redujo los errores de predicción en cuatro de los siete casos, logrando una mejora promedio de 4,6 puntos porcentuales para el total de los destinos analizados. Aunque la inclusión de estas variables incrementa la complejidad de los modelos debido al mayor número de parámetros, los hallazgos demuestran que los modelos multivariados son superiores para fines prácticos. Estos permiten a los agentes económicos disponer de herramientas más eficaces para la toma de decisiones oportunas frente a la evolución de la demanda turística. Finalmente, se plantea para futuras investigaciones el desafío de desarrollar mejoras metodológicas que permitan alcanzar una mayor parsimonia en los modelos predictivos.

Economic uncertainty, inherent to the contemporary landscape, necessitates precise and timely data to optimize decision-making processes. Given that data acquisition and scenario modeling entail significant resource outlays, information remains a costly asset. In this regard, the evolution of search engine technology offers a distinct advantage by providing high-frequency data and revealing behavioral patterns through search statistics. This research focuses on the Argentine tourism sector, utilizing Google Trends as a tool to forecast tourist inflows across seven specific destinations. The study aims to mitigate official publication lags by modeling demand through two distinct methodologies: a univariate ARIMA approach and a multivariate VAR model, the latter incorporating search-engine metadata. The empirical results demonstrate significant gains in predictive accuracy. The multivariate framework outperformed the univariate baseline in four of the seven cases, yielding an average error reduction of 4.6 percentage points for the total destinations analyzed. While the inclusion of Google data inherently increases model complexity due to the expanded parameter space, these multivariate models prove superior for practical applications. They empower economic agents to make more informed and timely decisions regarding tourism demand. Future research should focus on methodological refinements to achieve greater structural parsimony without compromising forecasting performance.This research focuses on the Argentine tourism sector, utilizing Google Trends as a tool to forecast tourist inflows across seven specific destinations. The study aims to mitigate official publication lags by modeling demand through two distinct methodologies: a univariate ARIMA approach and a multivariate VAR model, the latter incorporating search-engine metadata. The empirical results demonstrate significant gains in predictive accuracy. The multivariate framework outperformed the univariate baseline in four of the seven cases, yielding an average error reduction of 4.6 percentage points for the total destinations analyzed. While the inclusion of Google data inherently increases model complexity due to the expanded parameter space, these multivariate models prove superior for practical applications. They empower economic agents to make more informed and timely decisions regarding tourism demand. Future research should focus on methodological refinements to achieve greater structural parsimony without compromising forecasting performance.