Abstract:

Purpose- This article predicts tax declaration rates by Markov chain model. Methodology- Four Markov models are constructed for the declaration rates of three tax revenues. Declaration probabilities for the year 2017 are estimated by constructing probability matrices of transitions between classes described for every model. Declaration rates are predicted by the product of the initial probability matrix and transition probability matrix. Limiting matrices of predictions are found. The best Markov model was found by estimating the sum of mean square errors for every model. Findings- Main results of this study are i) transitions of tax declarations are declining in higher states and improving in lower states, ii) in its best model corporate tax declaration rate is predicted to be between 47% and 64.5% in 2017 with a probability of 78% and would be stable in the same interval at a probability of 60% in 2038, iii) in the long run income tax would decrease ₺20,3 billion with a probability of 42% and Value Added Tax would decrease ₺26,2 billion with a probability of 40% and iv) expected declaration rates of income tax, corporate tax and value added tax in 2017 are 52.3%, 61.6% and 62.8%, respectively. Conclusion- Income tax and Value Added Tax payoffs may substantially decrease from 2015 to 2038 and 2023 respectively. This may cause a revenue deficiency to Turkish Revenue Administration. Therefore İncome tax and Value Added Tax audits should be increased. Even though tax revenues increase over time, the declaration rates show a decreasing to stationary or increasing to stationary behavior.

Keywords: