Abstract
This study employs a predictive lead regression model to optimize the selection of Google Trends search outcomes, enhancing the accuracy of Unemployment predictions at the state level. The investigation centers on evaluating the correlation between Google Trends searches utilizing terms aligned with the Permanent Income Hypothesis and their capacity to forecast alterations in the Unemployment rates of various US states. Our empirical scrutiny integrates Google Trends data and state-level Unemployment records spanning a decade. Notably, our findings indicate that searches related to motor vehicles contribute significantly to the enhancement of Unemployment rate predictions for a state, offering improved foresight up to 12 months in advance, surpassing the predictive capacity of searches solely based on Unemployment-related terms. Moreover, broadening the geographical scope of Google Trends searches demonstrates the potential for further refining predictive outcomes in specific states. This empirical evidence underscores the efficacy of incorporating theoretically grounded multiple search terms within a broader context, thus advancing the utility of Google Trends for more accurate and resilient predictive results, transcending the limitations of narrowly focused topic-specific search terms. The Permanent Income Hypothesis plays a crucial role in this framework, further validating the importance of integrating economic theories with innovative data sources like Google Trends to predict Unemployment more effectively.
Keywords: Unemployment, Google Trends, Permanent Income Hypothesis
Purpose: This paper is designed to find a better fit for a predictive model of unemployment. Predicting unemployment using Google Search trends for the word unemployment has been done before and has some success. However, an external event, like a national discussion on unemployment, can skew these results causing the predictive model to fail. By adding other words that are related to unemployment we can improve the model and reduce failures.