Estimation of general budget tax revenues using Markov model

The aim of this study is to describe the behavior of the sample data and to estimate the realization rates of tax revenues by one step stochastic Markov chain model. The realization rates of the tax revenues are calculated by using 2000-2014 gross annual data extracted from Turkish Revenue Administration. Four Markov models are constructed for the realization rates of every tax revenue. The realization probabilities for the year 2016 are predicted by constructing probability matrices of transitions between classes described for every model. Tax realization rates are also forecasted by the product of the initial probability matrix and transition probability matrix. Limiting matrix of forecasts are found. The best Markov model is found by estimating the sum of mean square errors for every model. Validity of the models are tested statistically. The results are compared and interpreted.