Analysis of Consistency in Economic Forecasts for Russia Using Unemployment as an Example

The issue of forecasting various macroeconomic indicators in general and unemployment rates in particular has always been acute and discussed in academic circles, but forecasts can vary significantly in some cases depending on the approach to modeling. The purpose of this study is to compare the quality of various unemployment forecasts in Russia, identify the causes of differences in forecasts, regardless of whether the forecasts are bad or good, and draw conclusions about which models work best when forecasting on different horizons. This study will analyze the quality of unemployment forecasts in Russia from 2016 to 2020. Various quality metrics such as RMSE, MSE, MAE, SMAPE will be used to assess the quality of forecasts. In addition, for a deeper analysis of the causes of differences in forecasts, the models used by the authors of the studies will be built on the same data to determine the specific causes of the differences. Among the tested models will be econometric models such as ARIMA, as well as machine learning models such as boosting, perceptron and others. In general, the expected result of this work is to draw conclusions about which models are most suitable for forecasting unemployment and why they are more effective.