Forecasting Tax Revenues to the Budget of the Russian Federation Using Econometric and Machine Learning Models

Accurate forecasting of tax revenues is critical for effective budget planning and implementation of the state's fiscal policy. Tax revenues constitute the main part of the budget of the Russian Federation, so their volume directly affects the government's ability to finance social programs, infrastructure projects, and other government expenditures. Inaccuracies in tax revenue forecasts can lead to budget deficits, growth of public debt, and the need to cut planned expenditures. The aim of this thesis is to build accurate forecast models of the dynamics of tax revenues to the budget of the Russian Federation using modern methods of econometrics and machine learning, as well as to compare these models: to identify their strengths and weaknesses, and to analyze their applicability in real conditions. Comparative analysis of the quality of forecasts of various models allows identifying the most effective approaches to solving the problem. The empirical basis of the study is the data of the Federal Tax Service, the Treasury, and statistical service on tax revenues, macroeconomic indicators, and other relevant factors for the period from 2003 to 2023. This sample allows building models that take into account both long-term trends and short-term effects of crisis phenomena, including the crises of 2008-2009 and 2020. The obtained results can be used by the Ministry of Finance, the Federal Tax Service, and other authorities to improve the accuracy of budget planning and the stability of the country's financial system. Moreover, the developed approaches to forecasting tax revenues are potentially applicable in other countries with a similar structure of the economy and tax system.