Google Trends as an advanced way of indexing consumer sentiment and predicting private consumption

This thesis investigates the potential high-frequent alternative to survey-based consumer sentiment indices in the US. The high-frequent indicators are timely information and a prerequisite for sound policymaking, especially during turbulent periods in economic activity. A description of how US GDP, private consumption expenditure, and consumer sentiment are connected is discussed, alongside comparing of currently used indices of consumer sentiment, their drawbacks, and already created alternatives based on web search data. The primary dataset used for the analysis is the data from Google Trends, reflecting the relative popularity of web searches in the US. The selection of relevant search categories and methodology of constructing consumer sentiment index is inspired by Della Penna and Huang (2010). The autoregressive model of private consumption expenditure is employed to compare the predictive power of the synthesized index. The forecasting performance is assessed by comparison to the commonly used survey-based indicator – the University of Michigan Consumer Sentiment Index. The main value-added of this thesis is the addition of factors reflecting the public concern about news that was not utilized in any of the previous research. The results show that the index based on web search data is a significant predictor of one-month ahead variations of private consumption expenditure; however, its performance compared to the survey-based index is questionable. The integrated news category is shown to be a valuable contributor in predicting the swings in private consumption expenditure.