Crime modeling and forecasting the number of crimes in the constituent entities of the Russian Federation

The article presents the work-out of a comprehensive econometric research of crime situation in the Russian Federation based on regression analysis. As an object of research we have used empirical data from the Federal State Statistics Service for 2018 to identify the factors that have a significant impact on the number of heavy and especially grave crimes committed in Russia. When establishing correlation between variables, we have built a "fan" of six econometric models of multiple regression. To select the best model, we have carried out Box-Cox and Zarembka tests, which made it possible to extract a linear regression model. A complete econometric study of the problem under consideration also included the analysis of multicollinearity of factors and the study of the heteroscedasticity of the residuals of the linear regression model. Verification of the heterogeneity of observations in the model, which in the course of the study turned out to be the best of all the models considered, was carried out using the White, Breusch-Pagan, Goldfeld-Quandt, Park, and Glazer tests. For all performed tests the hypothesis of homoscedasticity of the residues was rejected. Since, as a result of the study preference was given to a linear regression model, it was on the basis of this model that point and interval forecasts were built. Quantitative relationships of the studied variables have been established.