The paper shows how textual information can be used to predict and study cause-effect relationships in macroeconomics. I consider a special case of forecasting - nowcasting on the example of unemployment. The key feature of nowcasting is that the forecast is built for a period that has already passed, but which has not yet come out statistics. As textual information, Internet requests are used. The paper is new in several direction. For the first time in the literature, information from two search engines, Yandex and Google, is used at once for forecasting. Information provided by search engines complements each other and allows performing suitable words’ selection from the bunch of users’ internet-requests. For the first time, the popularity of online systems as sources of information on job availability is taken into account. In Russia, the popularity of the Internet as a source of information on the availability of jobs has more than tripled from 2008 to 2018. If the researcher uses only the dynamics of related internet-requests then the results will show the dynamics of internetservices’ popularity rather than unemployment. Most of the models with internet query words show significant quality improvement in fore(now)casting unemployment. The paper proposes the procedure how to use query data for macroeconomic nowcasting

This paper proposes a novel three-step identification procedure of monetary policy and natural interest rate shocks. Altogether, monetary policy and natural interest shocks explain about 90% of total inflation dynamics. The paper exploits (Arias et al., 2019) procedure, which allows getting canonical impulse response functions to monetary policy shocks. I find no evidence of price and output puzzles. The estimated natural interest rate declines from 2015 to 2019 years. Furthermore, Bank of Russia follows mandate and reacts to inflation in monetary policy feedback rule, while does not respond to output fluctuations.