Forecasting Domestic Migration Using Internet Search Queries

Statistics on internal migration are published with a long delay, which makes it difficult to carry out timely analysis of migration processes. The study examines the suitability of Internet search queries for predicting internal migration on a monthly basis and their impact on the accuracy of predictive models. The analysis used monthly data on internal migration and Internet queries from Google Trends for Moscow and St. Petersburg in the period from 2011-2022. To build the necessary models, an increase in the frequency of data on socio-economic factors was used using cubic spline interpolation. In the work, univariate and multivariate time series predictive models were built. Empirical analysis has shown that including search terms in the model can improve the forecasting of population movements within a country, as forecast errors are reduced after their introduction.