Upotreba analize masovnih podataka za unapređenje efikasnosti naplate poreza u Poreskoj upravi Republike Srbije

Abstract

Tax evasion poses a major problem for the overall business environment in every economy - it endangers competition, reduces resources for budgetfunded public goods and services, and public policies enforcement as well. Moreover, fundamental human rights are denied in an informal labour market. The reform of the Tax Administration of the Republic of Serbia, which has been implemented since 2015, has already contributed to the increase in efficiency of tax collection over the past years; however, there is still significant room for improvement of tax collection, especially individual income taxes, social security contributions, and value added tax. One of the pillars of the Tax Administration Reform refers to the improvement of the analytic function regarding risk management aiming to make the tax control function more efficient, and raise the awareness of voluntary tax declaration. The paper presents the first results of a joint scientific-research project between the Tax Administration and the Faculty of Sciences of the University of Novi Sad, aiming to develop the algorithms for detecting the risk of tax evasion by using advanced methods in big data analytics and the development of artificial intelligence with the help of machine learning. The presented indicator is based on the weighted norm distance between income distribution in a legal entity and the average income distribution in the business sector which it operates in. The results show the sound performance of the developed indicator. In addition to improving the efficiency of field control, the approach is also going to enable an affirmative approach to those taxpayers who are classified in a low-risk category in terms of tax evasion. Furthermore, additional positive effects are expected based on higher self-reporting of risky categories due to higher probability of being a subject of a field control and detecting tax evasion.