Прогнозирование миграции из России в Германию с использованием Google-трендов
International migration statistics are published with a delay of up to several years. This prevents researchers from making timely analyses of migration flows. The article reviews a method for forecasting international migration flows based on search queries on the Internet using the example of flows from Russia to Germany during 2011-2020. Rosstat, German and OECD data were used to analyze migration. The approach proposed in the paper makes it possible to solve this problem by obtaining an estimate of migration trends with virtually no time delay. Moreover, in some cases it is possible to predict migration events before the actual relocation, which can also be used to evaluate other statistical indicators. To construct the necessary estimates, we employ methods for increasing the data frequency, making it possible to obtain monthly forecasts.
NLP approaches were used to obtain many search queries on migration topics. As a result, the parameters of a linear regression based on Google Trends search query data were evaluated, which made it possible to make a forecast of migration statistics before the publication of Rosstat statistics. The proposed models, in contrast to the model of seasonal autoregressive integrated moving averages (SARIMA), make it possible to take into account structural shifts and shocks in current processes reflected in Internet search queries, providing the opportunity to obtain short-term migration forecasts in real time (nowcasting). The described methods can be used both in the study of other pairs of countries and for the evaluation of other statistical data.