Following the standard stochastic frontier model for cross section analysis (Aigner et al., 1977) later extended to panel data (Battese and Coelli, 1995), let

where Y_it denotes own tax revenue (OTR)/ taxes on commodities and services (TaxCom), X_it denotes vector of inputs (viz., Log Per capita GSDP, Square of Log Per capita GSDP and Log share of Agriculture in GSDP) affecting tax revenue for the i ^th state in the t ^th period and β is a vector of unknown parameters. Error component is decomposed into two parts v_it and u_it; v_it is statistical noise term with symmetric distribution while u_it is a non-negative error component representing the time varying technical inefficiency. The inefficiency effects are expressed as an explicit function of state-specific variables and a random error to identify the reasons for differences in predicted efficiencies between states as given below

where W_it is a random variable defined by truncation of normal distribution with zero mean and variance σ² and Z_it denotes vector of variables predicting inefficiency (viz., Ratio of transfers in revenue receipts, Aggregate Expenditure to GSDP, Debt-GSDP ratio and VAT Dummy). The estimates of efficiency can be derived once the point estimates of u_it are obtained by the following expression;