Моделирование агрегированных показателей внешнеэкономической деятельности России с помощью блочных BVARX-моделей

Экономический журнал Высшей школы экономики. 2026. Т. 30. № 2. С. 215–244.

Slabolitskiy I. S.

This paper proposes a new approach for modeling and forecasting key indicators of Russia's foreign economic activity based on a block-based architecture of a Bayesian Vector Autoregression model with exogenous variables (BVARX). In conditions of high external uncertainty and the absence of publicly available data on the physical volumes of foreign trade since 2022, the model addresses two interconnected tasks: scenario forecasting and the reconstruction of statistical time series.

The methodology is based on dividing the system of variables into four substantive blocks (foreign exchange market, oil and gas exports, terms of trade, imports), which allows for the formulation of economically interpretable prior constraints, avoids overfitting, and ensures compu-tational efficiency. The model is estimated using quarterly data from 2000 onwards. The results show that the forecasting quality of the BVARX model on horizons from one to eight quarters significantly exceeds the accuracy of basic autoregressive models (AR/ARX) for most variables. As an applied result, reconstructed values for indicators of oil and gas exports and imports for the period 2022–2025 are presented.

Keywords: forecasting, data imputation, Bayesian vector autoregressive model, прогнозирование внешнеэкономической деятельности России, восстановление данных, байесовская векторная авторегрессионная модель

Publication based on the results of: New directions in applied stochastic analysis (2025)