Revealing the mood of economic agents based on search queries

Currently, user behavior on the Internet is becoming a key source of information about the market sentiments and the public mood. In this regard, it becomes relevant to study the usefulness of such indicators in modeling macroeconomic indicators. This article proposes an approach to assessment the mood of economic agents using Google Trends search queries. A key particularity of the article is the selection of keywords based on the analysis of RBC news from January 2010 to March 2020. The results showed that sentiments in financial and money markets based on a principal component analysis strongly correlated with the financial stress index of ACRA and Rosstat consumer confidence index, which confirms the possibility of use search queries in the development and analysis of economic policy. We use search query indices for the consumer confidence index forecasting with mixed data sampling (MIDAS). In out-of-sample forecasting, our results show that MIDAS which includes the indicator of sentiment in the money market gives the best forecast performance for the next quarter, and MIDAS with the sentiment in the financial markets has a high accuracy of forecasting the consumer confidence index for 3 quarters.