BusinessSurveyCBRCpiForecastingSourceCode

Does the use of the Bank of Russia's business survey data help to increase the accuracy of the inflation forecast for Russia on a short-term horizon of 1, 2, 3 and 6 months? In order to conduct effective monetary policy, the central bank should be guided by the inflation forecast when making decisions. The more accurate this forecast is, the more effective the central bank's policy is both in terms of specific policy decisions and in terms of building trust and inflation expectations of households. Potentially, bysiness survey data have strong predictive properties due to the utilization on the forecasting power of economic agents. In the current paper we test the predictive power of business survey data through the method of comparing out-of-sample CPI forecast errors in pseudo-real time by a set of econometric analysis and machine learning models with different information sets: with and without business survey indicators. The comparison of errors of models with different information sets allowed us to assess the overall contribution of business survey data to the predictive power of the models. The Diebold-Mariano test, adjusted for small samples, was used as a formal test for comparing model accuracy metrics. As a result of modeling, empirical arguments supporting the existence of useful predictive properties of Bank of Russia's business survey indicators for CPI forecast models were obtained. Adding business survey indicators to models generally leads to a reduction in out-of-sample forecast error, especially when forecasting for a horizon of up to three months. This result depends on the choice of model; on average, direct forecast models with regularization better reveal predictive properties of business survvey data. The forecast of CPI by components also works well when using sectoral business survey statistics. In general, indicators from the business survey can be considered as leading indicators and used in short-term forecasting.