FROM TAX ADMINISTRATION TO INTELLIGENT TAX ADMINISTRATION: ARTIFICIAL INTELLIGENCE, TAX CONTROL EFFICIENCY, AND THE REDUCTION OF TAX EVASION

Abstract

Nevertheless, tax evasion still leads to losses of significant public funds in developing countries' economies due to poorly organized control mechanisms. Artificial intelligence may serve as one of the means of increasing the efficiency of such mechanisms; however, not much research has been done regarding the performance of intelligent systems at tax authorities. In view of this, the study examines the impact of artificial intelligence adoption on the reduction of tax evasion and checks whether this impact occurs directly or through the intermediary of tax control efficiency, taxpayer compliance, and administrative cost reduction. Empirical data was collected among 215 employees of tax authorities, accounting departments, financial management departments, and information technology departments using a questionnaire consisting of 28 questions on the five-point Likert scale. Internal consistency of five variables was satisfactory, with Cronbach's alpha equal to .77-.86. Then three OLS models were estimated, and their diagnoses for collinearity, autocorrelation, heteroskedasticity, and normality of residuals were performed. The results show that artificial intelligence adoption had the biggest impact on the reduction of tax evasion (B = 0.342, p < .001); in addition, this model explained 67% of the variance. Furthermore, it significantly increased taxpayer compliance (B = 0.318) and reduced administrative costs (B = 0.286). Decomposition of effects revealed that taxpayer compliance and cost reduction pathways conveyed 23.7% of the effect of artificial intelligence adoption on evasion reduction, and the rest was conveyed directly. Thus, all hypotheses have been empirically tested.