A Markov-based decision support model for tax evasion in Greece

Abstract

This thesis proposes a Markov-based decision support model which captures the behaviour of a typical firm in Greece, vis-a-vis tax evasion. To maximize its long-term wealth, the firm has two options at its disposal. First, it can manipulate the percentage of its profits that it will disclose (and consequently be taxed on). Second, there is a type of optional tax amnesty that may be offered to the firm by the Greek government. This option, termed “closure”, allows the firm to pay a lump-sum tax, based on its gross income or stated profits, in exchange for eliminating the possibility of an audit against the firm's past income declarations. We describe a dynamical system, aimed at predicting the actions of the average Greek firm, and at evaluating tax policies before they are implemented. The basic model evolves through several iterations, some computationally challenging, depending on the firm's attitude towards risk and the availability of closure. The model proposed in this thesis represents an innovative step towards bridging the gap between classical macroeconomic and game-theoretic approaches on the subject. It takes into consideration the sequential nature of the firm's decisions regarding tax evasion, together with its attitude towards risk, and allows us to: i) analyze the effectiveness of the closure option, ii) show that in the current environment a rational enterprise has no incentive to disclose its profits, iii) identify ``virtuous'' combinations of tax parameters which lead to full disclosure of profits and iv) estimate a firm's risk-aversion. Our analysis can be used to evaluate the effectiveness of various taxation schemes, potentially benefiting both firms and government.