Estimating Consumers’ Risk Preference from Histories of Search Queries

Estimating Consumers’ Risk Preference from Histories of Search Queries

Risk preferences shape consumer behavior, but measuring them is often costly and time-intensive. This paper examines whether Google search histories can reveal consumers’ risk preferences. Search queries reflect what people attend to, the problems they try to solve, and the decisions they prepare for, many of which involve uncertainty. We collected Google search histories from 998 consumers, covering up to 16 years and 150,000 queries per person, and measured their risk preferences. Using large language models, we predict individual risk preferences from these queries. The results show modest but statistically significant predictive power, suggesting that search histories contain relevant signals. In particular, we find that consumer interests reflected in search behavior, such as addiction, smoking, and betting, are informative predictors of risk preferences. These findings highlight the potential of search data for scalable preference measurement and show that consumers’ digital traces can provide meaningful insights into economically relevant individual differences.