Estimating State-Level Tax Effort in India: A Panel Data Model

After the introduction of GST in 2017, the discretionary space for managing state-level taxes has been tangibly reduced for the state governments. For the taxes which have been subsumed in the GST, the autonomy of the states to change their rates, classification schemes, exemptions, and deductions has been effectively vested in the GST Council implying that any state cannot, in practice, bring about changes on its own. Among taxes where states still have discretion, VAT on petroleum products constitutes the largest share. Other non-GST state-level taxes include state excise duty on alcohol, stamp duty and registration fee, motor vehicle tax, land revenues, and electricity duty. In this paper, we consider two major categories of state taxes namely, VAT on PoL products and an aggregate category mainly comprising SGST along with land revenues, electricity duty, motor vehicle tax, etc. Together these two categories accounted for nearly 78% of states’ own tax revenues on average during 2017–18 to 2019–20 considering 17 medium and large (ML) states. One critical feature in determining these taxes is the presence of interdependence between state expenditures and fiscal transfers and state own tax revenues. We resolve this simultaneity by using a panel 2SLS estimation procedure where instruments in the first stage of estimation are drawn from a broader simultaneous equations model for determination of fiscal transfers. Using this framework, we derive a ranking of 17 ML states for these two tax categories according to their tax effort. We find that in terms of relative tax effort, the most efficient state was Gujarat in the case of SGST et al. category and Maharashtra in the case of VAT on PoL products. We estimate the model-predicted tax revenues first by using actual values of exogenous variables and then by benchmarking the states to operate at the frontier of the tax effort. These estimates can serve as relevant inputs in the determining fiscal transfers where normative levels of tax revenues require to be estimated.