Tax Effort of the Indian States from 2001–2002 to 2016–2017: A Stochastic Frontier Approach

Abstract

This article estimates the tax capacity and tax effort of 17 major states of India from 2001–2002 to 2016–2017 using the stochastic frontier panel data model. It is found that per capita income, agriculture activity, infrastructure, labour force and bank credit are the significant determinants of tax capacity, while social sector spending and central transfer to states are significant in determining tax effort. The Goods and Services Tax has reduced the states’ tax powers. Therefore, the states are highly dependent on their limited legislative taxes for revenue mobilization. However, there is little scope for the subnational governments to increase tax revenue as all states have achieved at least 90% of their tax potential.

Keywords

Tax effort tax capacity own tax revenue excise duty stochastic frontier approach Goods and Services Tax (GST)fiscal consolidation stamp duty

I. Introduction

A sound public finance policy ensures the smooth functioning of an economy. All the decisions related to the generation of revenue and utilization of funds for development and welfare come under public finance policy. The fiscal sustainability of the state is a precondition for overall economic sustainability since unstable state finances hinder and jeopardize the accomplishment of the state’s development functions required to address social and environmental issues (Sawhney, 2018). The state’s spending on productive channels financed by direct and indirect taxes boost economic growth (Gemmell, 2001; Trivedi & Rajmal, 2011). The state’s fiscal policy is highly significant in affecting an ordinary citizen’s social and economic life. It, thus, generates great interest in studying the concepts of tax effort, tax capacity, public revenue and spending, fiscal performance and issues related to growth and development. Thus, this article aims to analyse the tax effort of the 17 general category states.

Tax capacity means the maximum level of tax revenue an economy can attain, given its economic, social, administrative, demographic and other characteristics (Garg et al., 2017; Lotz & Mors, 1967; Pessino & Finochietto, 2010; The U.S. Advisory Commission on Intergovernmental Relations [ACIR], 1962). On the other hand, tax effort indicates the ratio of actual tax collection to tax capacity. Tax effort describes the capability and preparedness of a government to collect revenue from its tax capacity. An operative tax system is essential for guaranteeing a secure fiscal environment, which improves governments’ accountability and promotes good governance (Bhatia, 1996; Bird et al. 2008). Therefore, governments must analyse their tax capacity and tax effort at times.

India is a quasi-federal nation with three government levels: the first one is the union, the second is the state government and the third one is the local government. The central and state governments’ respective functions are provided in the union and state list under Schedule VII of the Indian Constitution. The Concurrent list under Schedule VII prescribes the mutual powers of the centre and state. However, in case of any dispute between the central and state governments, residual authority remains with the Centre, which constraints the capacity of the state governments in making discretionary changes, especially with tax revenue policies (Basu, 1982).

On the contrary, the subnational governments undertake more than half of the nation’s total expenditure (Bagchi, 2003; Rao et al., 1999; Sen & Dash, 2013), which ultimately causes a vertical fiscal imbalance between the union and the states. On the other hand, the differences in states’ natural, demographical and structural composition cause horizontal fiscal imbalances. The states rely primarily on two sources of revenue: their own tax revenue and union transfers. The state governments receive more than 40% of their total revenues from centre transfers, including the devolution of union taxes and grant-in-aid (Suhag & Tiwari, 2018). The 14th finance commission has increased tax devolution to states from 32% to 42% (Finance Commission [FFC], 2015). Simultaneously, the union government has reduced the number of Centrally Sponsored Schemes (CSSs) significantly from 66 to 28 by 2017.

Moreover, the design of CSS has also changed, increasing the share of state government contributions from 50% to 60% (Reddy, 2019). That might bring more pressure on the state governments’ finances, especially in the low-income states. Besides, the implementation of the new Goods and Services Tax (GST) in India has constrained state governments’ autonomy. The GST Council, headed by the union finance minister, levies and governs GST.

GST has subsumed a number of union and state taxes. Therefore, the states possess legislative powers on very few taxes. With states’ reduced autonomy, their own tax revenue (OTR) might be reduced by 17% (Suhag & Tiwari, 2018). Despite these constraints, state governments need to increase their tax revenue on the one hand and rationalize their spending, on the other hand, to achieve fiscal consolidation. Due to the insufficiency of centre transfers and OTR, states accumulated debt, which posed a threat to macroeconomic stability in the past. Therefore, the state governments must estimate their tax potential (capacity) and analyse their tax effort.

Most of the recent studies in the Indian context have concentrated on measuring the tax effort of overall OTR. The states need to analyse their legislative taxes, which are not subsumed under GST. The non-GST taxes with states’ legislative power include state excise duty, stamp duty, electricity duty, value-added tax (VAT) on sale of petroleum products, vehicle tax and professional tax. The present article aims to measure the tax capacity and tax effort of states’ non-GST taxes (excluding VAT on sale of petroleum products)¹ of 17 general category states using the stochastic frontier approach (SFA).

The stochastic frontier analysis is an extension of a traditional production function, which measures the maximum output and technical efficiencies with a given set of inputs. Similarly, a stochastic frontier measures the maximum level of tax revenue that a country/state can achieve with some set of inputs. The ratio of actual tax revenue to the predicted maximum level of tax revenue is tax effort. There are two primary differences between a production frontier and a tax frontier. According to Alfirman (2003), the first distinction is that the determinants of output in a production frontier are specific, namely labour and capital. However, the determinants of the tax frontier are not specific. The second main difference between a production and tax frontier is the interpretation of results. Firms do not accomplish the difference between actual and potential output, that is, technical inefficiency. In the case of tax frontier, the shortfall of actual tax revenue in tax frontier includes technical inefficiency and policy issues. The policy issues include tax legislation differences, which authorities can modify (Pessino & Finochietto, 2010; Cyan et al. 2013).

This article uses a set of economic, social, demographic and fiscal parameters guided by recent literature to determine tax capacity and tax effort. This article primarily focuses on the following questions: What factors determine tax revenue, tax capacity and tax effort in states? What are the relative tax capacity and tax effort in Indian states? What is the scope of tax realization for subnational governments from their legislative taxes?

The rest of this paper is structured as follows. Section II includes a concise review of the literature. Section III covers the research methodology as well as summary statistics. Section IV contains the results and discussion. Finally, in Section V, the conclusion is presented.

II. Review of Literature

Scholars have long been interested in the measurement of tax potential and tax effort. There is a substantial amount of literature available on the topic. However, few studies have used the stochastic frontier analysis to measure tax capacity and tax effort, especially at the subnational level. Most of the existing studies estimated tax potential and tax effort using one or more of the three methods, namely income approach, representative tax system and standard regression approach.

Various Methods for Measuring Tax Effort

Lotz and Mors (1967) mentioned that economic development could be a better measure of tax potential. He identified economic development indicators, such as per capita income, literacy level, strict law enforcement and trade openness whose influence moves in the same direction as taxable capacity. The Representative Tax System (RTS) is another method widely used in measuring tax effort. The RTS approach entails identifying close proxies for the tax base respective to each tax category (Bahl, 1972; Dwibedi et al., 2016; Piancastelli, 2001; Purohit, 2006). However, Mikesell (2007) details that finding accurate and dependable tax bases is problematic. The absence of a cautiously defined tax base might estimate the random taxable capacity index. The majority of studies in the existing literature have used the standard regression approach to analyse the impact of various factors explaining inter-regional tax revenue variation (Crivelli & Gupta, 2014; Davoodi & Grigorian, 2007; Gupta, 2007; Le et al., 2012; Lotz & Mors, 1967; Purohit, 2006; Tanzi, 1992). In the regression approach, the error term denotes inefficiency, which contains the random component, which might lead to arbitrary tax effort estimation (Rao, 1993). However, these studies significantly contribute to identifying factors that explain taxable capacity and tax effort in different regions. Some recent studies that used the standard regression approach to identify the independent variable explaining tax potential and tax effort are provided in the subsequent paragraph.

Cross-Country Analysis of Tax Effort

Gupta (2007) used a panel data regression analysis of 105 developing economies; he found that structural factors such as per capita income, the share of agriculture in GDP, trade openness and foreign aid significantly affect tax revenues. Davoodi and Grigorian (2007) used a cross-country regression analysis to analyse Armenia’s tax performance. They found that institutional improvements and a reduction in informal activity size are essential factors in increasing tax performance. Baunsgaard and Keen (2010) used the panel data of 117 countries over 32 years to study the association between tax revenue and trade liberalization. They noticed a positive and significant impact of trade liberalization on tax revenue. However, the relationship was relatively weaker in low-income countries.

Similarly, Le et al. (2012) attempted to identify tax revenue determinants using a panel data set of 110 countries. They found that demographic factors such as population growth, dependency ratio, macroeconomic variables, namely agricultural activity, trade openness and institutional variables like administrative quality and corruption had a significant impact on tax revenue mobilization. In contrast, Crivelli and Gupta (2014) did not find a substantial relationship between tax effort and corruption in their study of 35 resource-rich countries. However, they also identified gross value added (GAV) of agriculture, trade openness and foreign aid are significant factors in determining tax effort.

Tax Effort Using Stochastic Frontier Approach

SFA has been used in the majority of recent studies to estimate tax potential and tax effort (Alfirman, 2003; Garg et al., 2017; Jha et al., 2000; Langford & Ogleberg, 2015; Pessino & Finochietto, 2010). In the earliest attempt, Jha et al. (2000) applied SFA to estimate tax efficiencies. They figured out the tax efficiencies of 15 general category Indian states and discovered the moral hazard problem associated with the vertical transfer of funds. The higher grants from the union government to states tended to reduce the efficiency of tax collection of the state governments. Their study also asserted that the low-income states were less efficient in tax collection. Similarly, Alfirman (2003) studied the tax revenue of Indonesia’s local governments using SFA and found that the local governments were far away from utilizing their tax potential.

Pessino and Fenochietto (2010) estimated the tax capacity and tax effort of 96 countries using SFA. They claimed that tax potential increased in the same direction as per capita GDP, trade and public sector education spending, but decreased as agricultural activity increased and income inequality increased. In parallel, tax effort deteriorates as consumer inflation rises, while it improves as corruption decreases. Similarly, Langford and Ohlenberg (2015) attempted to measure the tax capacity of 85 no resource-rich countries using SFA. Their results also established that tax potential increases with a high level of education and trade openness. They also asserted that tax effort had a positive and significant relationship with better law and order conditions, improved democratic accountability and a low level of corruption.

Tax Effort in Indian Context

The estimation of tax potential and tax effort has attracted the attention of Indian policy economists. Many scholars have attempted to measure and analyse the performance of tax revenue realization in India, especially at the subnational level. Reddy (1975), Thimmaiah (1979), Oommen (1987), Sen (1997), Coondoo et al. (2001), and Purohit (2006) used the regression approach to analyse the inter-state tax variation in India, while the most recent studies by Karnik and Raju (2015), Garg et al. (2017), Mukherjee (2017) and Nayudu (2019) applied SFA to estimate the tax potential and tax effort of Indian states. The following paragraph details the findings of the above-mentioned SFA studies in the Indian context.

Karnik and Raju (2015) used the data of 4 significant taxes of 17 states of India from 2001 to 2011. They asserted that the states were unable to realize the optimum revenue from their tax potential. They stressed that reducing corruption and decentralization of public expenditure is vital to increasing OTR in states. Garg et al. (2017) tried to find the reasons behind the shortfall of tax effort from 1991–1992 to 2010–2011. They asserted that states’ tax effort worsens with higher intergovernmental transfers; on the other hand, it grows with improved law and order, political competition and a higher ratio of public expenditure in GSDP. Their inefficiency equation results reveal that the implementation of FRMBA has resulted in enhanced tax effort. They observed a significant degree of variation in the tax effort of different states.

Furthermore, Mukherjee (2017) tried to measure the VAT efficiency and reasons for its variation across Indian states from 2001 to 2014. He discovered that an increase in per capita GSDP, royalties received from natural resources and anti-incumbency were all responsible for improved tax efficiency. However, an increase in grants and share in central taxes tended to decrease the VAT effort. In this study, the tax efficiency across states did not show convergence. He advocated an in-depth assessment of tax administration that could help in improving the tax efficiency of states. Finally, Nayudu (2019) explains that stamp duty and registration fees are the only significant sources of tax revenue for states after implementing GST. He, thus, estimated the tax effort mentioned taxes of 16 significant states from 2001 to 2014 using SFA. The tax efficiency index had a very high degree of variation across states, and an improvement in the administrative set up could help improve tax efficiency.

III. Research Methodology

The stochastic frontier model is an extension of a regression model that explains the maximum output level that a country can attain with a given set of inputs. Similarly, a stochastic tax frontier represents the maximum tax revenue level that a state can attain given a set of inputs. Aigner et al. (1977) originally introduced a standard econometric stochastic frontier model. Later on, several variants of the model were applied by authors with different structures of the error term and the inefficiency term. Battese and Coelli (1995) defined a two-stage estimation model of SFA for panel data of firms. The first step involves approximating a production frontier and predicted values of the technical inefficiency relative to predicted frontier values. In the second stage, the inefficiency effects are assumed to be a linear function of observable factors. We applied the Battese and Coelli (1995) model to estimate the tax frontier and technical inefficiency of 17 vital states of India using panel data from 2001–2002 to 2016–2017. The specification of the frontier model is as follows:

I n (Y_{i t}) = α + β^{T} (I n X_{i t}) + v_{i t} + u_{i t}

(1)

where

Y_it represents the log of OTR-GSDP ratio of the ith state at time tth time (t = 1, 2, 3, …, T),

X_it represents the log of variables affecting the tax revenue of the ith state at time tth time (t = 1, 2, 3, …, T),

v_it represents disturbance term or error term independent of u_it, and normally distributed. v_it can be positive or negative.

u_it represents technical inefficiency; it is a non-negative random component associated with the factors responsible for non-attainment of a maximum tax capacity of ith state at time t. u_it is independent of v_it.

α represents constant.

β represents the vector of unknown parameters.

Equation (1) represents the stochastic tax frontier, which differs from the standard regression model in terms of its construction of the error term. The standard regression model assumes an error term that can have both positive and negative values. It indicates that technical efficiency can be lower or higher than tax capacity. Thus, the states might seem to either underperform or over-perform. On the other hand, the non-negative error term in SFA warrants that a state can achieve the maximum output level; thus, the actual revenue collection cannot exceed the optimal revenue level (Battese & Coelli, 1995; Garg et al., 2017; Pessino & Finochietto, 2010). The uit represents the technical inefficiency, which is specified as follows:

u_{i t} = Z_{i t} δ + W_{i t}

(2)

where

Z_it represents the vector of variables explanatory variables associated with the technical inefficiency of tax revenue collection over time.

δ represents a vector of unknown parameters.

W_it is a random variable defined by the truncation of the normal distribution with zero mean and variance δ².

Finally, Equation (3) defines the technical efficiency, that is, the tax effort of the ith state in tth year based on the conditional expectations (Jondrow et al., 1982):

T E_{i t} = E [\exp (- u_{i t} | ε_{i})]

(3)

where TE represents the tax effort and ε_i represents the composite error.

There are several alternate specifications available to estimate the stochastic frontier within the panel data models, namely Pit and Lee (1981), Schmidt and Sickles (1984), Battese and Coelli (1995), Lee and Schmidt (1993), Battese and Coelli (1995), Kumbhakar (1990), Green (2005) and Belotti et al. (2013). We have adopted the Battese and Coelli (1995) model to estimate the frontier and technical inefficiency. The Battese and Coelli (1995) model avoids the bias by estimating the parameters of the stochastic frontier and inefficiency equation simultaneously (Mukherjee, 2017; Wang & Schmidt, 2002). Comparing with other models, the Battese and Coelli (1995) model captures the time-varying inefficiency from observable heterogeneity using maximum likelihood estimation (Belotti et al., 2013; Mukherjee, 2017; Wang & Schmidt, 2002).

Variables and Summary Statistics

We have used a panel data set of 17 non-special Indian states over 16 years from 2001–2002 to 2016–2017. These states include Andhra Pradesh, Bihar, Chhattisgarh, Goa, Gujarat, Haryana, Jharkhand, Kerala, Karnataka, Madhya Pradesh, Maharashtra, Orissa, Punjab, Rajasthan, Tamil Nadu, Uttar Pradesh and West Bengal. The list of data sources is given in Appendix A. Table 1 presents the list of variables and summary statistics. The data set comprises two sets of independent variables where the first one explains the tax capacity across states and the second data set determines inefficiency. This article uses a set of economic, social, demographic, fiscal and administrative parameters guided by recent literature to determine tax capacity and tax effort.

Table 1.

Summary Statistics

Variables	Observations	Mean	Std. Deviation	Minimum	Maximum
Dependent variables
Tax–GSDP ratio	272	6.68	1.65	3.61	14.56
The ratio of non-GST taxes to GSDP	272	2.15	0.74	0.61	4.70
Independent variables explaining tax capacity
Real per capita NSDP	272	67277.73	44987.11	11586.77	308823.00
Agricultural GSDP	272	15.51	7.09	2.42	33.11
Labour force (per thousand)	272	448.47	81.84	278.84	683
Road density	272	1.48	1.13	0.04	5.54
Credit-to-GSDP ratio	272	31.11	19.58	8.96	102.27
Independent variables explaining tax effort
Share of social sector expenditure in GSDP	272	7.08	2.58	3.15	19.56
Percentage of total central transfers in GSDP	272	3.64	2.78	0.65	14.52
FRBMA targets	272	0.51	0.50	0	1
Implementation of VAT	272	0.72	0.45	0	1

Source: Summary statistics are based on Reserve Bank of India (RBI) and Central Statistics Office (CSO) data.

Dependent Variables

Own tax revenue as a percentage of GSDP: A log of OTR percentage in GSDP is the dependent variable to measure the overall tax effort. The own tax revenue in the reference period varies between 3.61% of GSDP and 14.56%, whereas the average value is 6.68%.

The ratio of non-GST tax to GSDP: The non-GST taxes with states’ legislative power include state excise duty, stamp duty, electricity duty, VAT on sale of petroleum products, vehicle tax and professional tax. This ratio presents the taxes on which states have legislative powers (excluding VAT on sale of petroleum products) after the implementation of the GST. It ranges between 0.61% and 4.70% of GSDP with a mean of 2.15%.

Independent Variables Explaining Tax Capacity

Real per capita NSDP: per capita net state domestic product (NSDP) is an economic development indicator. A high degree of development brings more demand for public expenditure (Tanzi, 1992), thus a higher level of tax capacity to pay for public spending. Literature suggests that an increase in the per capita income leads to a rise in tax revenue. Thus, a log of per capita NSDP is taken as an independent variable.

Agricultural GSDP: The share of agricultural activity in GSDP across states varies between 2.42% and 33.11%, with a mean of 15.51%. Most previous studies suggest a negative relationship between agricultural activity and tax revenue (Crivelli & Gupta, 2014; Gupta, 2007; Le et al., 2012). The share of the agriculture sector in GSDP of states is declining significantly, and the income from agriculture is mostly exempt (Sengupta & Rao, 2012). Therefore, an expansion in the agricultural sector might inversely impact the tax capacity of a state.

Labour force (per thousand): The labour force represents the ratio of economically active population to the total population. We have used a ratio of total working age (between age 15 to 64) population to total population as measured by the usual status approach of National Sample and Survey (NSSO). It includes both the employed and involuntary unemployed population. A higher proportion of an economically active population enhances tax potential. Hence, we can expect a positive impact of the labour force on tax frontier. The labour force per thousand varies between 279 and 683 in the sample.

Road density: Infrastructure plays a crucial role in the economic development of a region. A sound road transport system establishes a facile connection between farms, fields, factories and markets. This ultimately provides access to employment, education and health and contributes to increased public services revenue. Therefore, we can expect a positive impact of road density on tax potential. Road density is a ratio of total road length in kilometres to the entire area of the state. The average road density is 1.48 kilometres per square kilometre, ranging between 0.04 and 5.54 kilometres per square kilometre.

The ratio of credit by scheduled commercial banks to GSDP: Credit extension stimulates economic activity, which ultimately creates the tax base. Thus, we can expect a positive impact of the ratio of credit by scheduled commercial banks (SCBs) to GSDP on tax capacity. The ratio varies from 8.96% to 102.27% with a mean of 31.11%.

Independent Variables Explaining Tax Effort

Share of social sector expenditure in GSDP: The ratio of social sector spending is crucial for education, public health and labour welfare (Purohit, 2014). A high level of social sector expenditure signifies citizens’ empowerment in the states, which ultimately impacts the young population in the productive age group. Thus, we can expect a positive association between social sector spending and tax effort. The share of social sector spending in GSDP ranges between 3.15% and 19.56%.

Percentage of total central transfers in GSDP: Percentage of total central transfers to the respective state in its GSDP is another essential fiscal variable in determining tax effort. It varies from 0.65% to 14.52% of GSDP with a mean of 3.64%. ‘States’ efforts to increase their own tax collection may be restricted by the anticipation of obtaining a large sum in the form of central transfers (Garg et al., 2017).

FRBMA targets: The Government of India introduced Fiscal Responsibility and Budget Management Act (FRBMA) with a motto of restraining the deficits of the union and state governments. The introduction of FRBMA enabled the states to optimize their spending and revenue. Therefore, we constructed a dummy variable to examine the effect of FRBMA on tax effort. The year in which states met the FRBMA target (a fiscal deficit of less than 3% of GSDP) is assigned a value of 1 (one), otherwise a value of 0 (zero). We expect a positive impact of FRBMA Dummy on tax effort.

Value-added tax introduction: We have constructed another dummy variable to examine the impact of VAT on tax revenue collection. The year of VAT implementation is assigned a value of 1 (one), otherwise a value of 0 (zero). We expect a positive impact of the VAT on tax efficiency.

IV. Results and Discussion

Table 2 shows the results of the SFA estimation for the OTR collection. The macroeconomic variables such as inflation, capital formation and exchange rate do not vary across states; however, they fluctuate across years. Accordingly, to control the time-specific fixed effects, we have included the time dummies in the model.² The results are based on the maximum likelihood time-varying inefficiency model developed by Battese and Coelli in 1995.

The log-likelihood value in the model shows the goodness of fit. A higher log-likelihood measure shows high goodness of fit. Moreover, sigma_u and lambda parameters indicate the presence of technical inefficiency in the model. If the sigma_u and lambda are significant, there is a presence of technical inefficiency, which means there is a deficit in tax effort. Equation (1) presents the tax frontier’s estimation, and Equation (2) presents the inefficiency. In the inefficiency model, a negative coefficient indicates a positive impact on tax efficiency (tax effort) and vice versa.

The log-likelihood value of 371.52 in Table 2 shows high goodness of fit for the estimation of tax frontier and inefficiency. The estimation results show the agriculture share in GSDP, per capita income, labour force and road density, and credit disbursement significantly determines the states’ tax capacity. It signifies that the states with a superior economic, demographic and infrastructural structure have a better tax capacity. It is to note that the agriculture share has a positive impact on tax capacity. However, most studies on the topic deduce that agriculture activity is negatively associated with tax capacity (Crivelli & Gupta, 2014; Gupta, 2007; Le et al., 2012; Sengupta & Rao, 2012).

Table 2.
Tax Frontier and Technical Efficiency of Own Tax Revenue from 2001–2002 to 2016–2017

Dependent Variable: Log (own tax revenue share in GSDP) Coefficient SE p-Value

Constant −1.602* 0.276 0.000

Log per capita NSDP (constant price) 0.240* 0.044 0.000

Log agriculture share in GSDP (current price) 0.191* 0.026 0.000

Log working-age population share in total population 0.366* 0.089 0.000

Log road density 0.046* 0.016 0.003

Log bank credit by SCBs share in GSDP 0.151* 0.030 0.000

Inefficiency equation

Constant 0.467* 0.063 0.000

Log social sector expenditure share in GSDP_(t-1) −0.517* 0.089 0.000

Log share of total central transfers in GSDP_(t-1) 0.110 0.050 0.028

Achievement of FRBMA target 0.006 0.012 0.644

Introduction of VAT 0.037 0.026 0.155

Sigma n 0.039* 0.022 0.001

Lambda 0.792*** 0.037 0.000

Log-likelihood 371.520

Number of observations 272

Dependent Variable: Log (own tax revenue share in GSDP)	Coefficient	SE	p-Value
Constant	−1.602***	0.276	0.000
Log per capita NSDP (constant price)	0.240***	0.044	0.000
Log agriculture share in GSDP (current price)	0.191***	0.026	0.000
Log working-age population share in total population	0.366***	0.089	0.000
Log road density	0.046***	0.016	0.003
Log bank credit by SCBs share in GSDP	0.151***	0.030	0.000
Inefficiency equation
Constant	0.467***	0.063	0.000
Log social sector expenditure share in GSDP_(t-1)	−0.517***	0.089	0.000
Log share of total central transfers in GSDP_(t-1)	0.110**	0.050	0.028
Achievement of FRBMA target	0.006	0.012	0.644
Introduction of VAT	0.037	0.026	0.155
Sigma n	0.039***	0.022	0.001
Lambda	0.792***	0.037	0.000
Log-likelihood	371.520
Number of observations	272

Source: The authors.

Note: *, ** and *** indicate values significant at 10%, 5% and 1% levels, respectively.

Scholars generally attribute the exemption of agricultural income from taxation as a primary reason behind the negative relationship between tax capacity and the agricultural sector. However, it is noteworthy that the relatively poor states such as Madhya Pradesh, Chhattisgarh, Bihar and Rajasthan attained high economic growth rates and outperformed the national growth rate. This high economic growth rate has resulted from a higher growth of the agriculture sector in these states (Kawadia & Philips, 2014). Thus, we can deduce that agriculture-driven economic growth has created the tax base in the above-mentioned states. Thus, there is a positive association between the agricultural sector and the tax frontier in the model.

The significant value of sigma_u and lambda indicates the simultaneous presence of technical inefficiency in the model. The technical inefficiency equation with observable heterogeneity explains the factors determining the shortfall in tax effort. A negative coefficient indicates a positive impact on tax efficiency (tax effort) and vice versa in the inefficiency model. Social sector expenditure and share of central transfers are the significant determinants of tax effort. The results show that social sector spending is positively associated with tax effort, which explains that an increase in social sector expenditure increases states’ tax effort.

In contrast, the share of total central transfers of GSDP harms technical efficiency. ‘The expectation of receiving a significant sum in the form of central transfers may limit states’ efforts to enhance their own tax collection (Garg et al., 2017). The estimation results also describe that the FRBMA and the VAT dummies are insignificant in determining the tax effort.

Table 3 presents the results of SFA and technical inefficiency estimation for Non-GST taxes in states. The model has a significantly high log-likelihood value of 382.50, which shows an excellent fit for the model. Except for road density, the tax frontier shows that all independent variables have a positive and significant impact on the state’ tax potential. Agriculture’s contribution to GDP, per capita income, labour force and credit disbursement are all important determinants of tax capacity. An increase in independent variables increases the potential of revenue collection from states’ legislative taxes.

Table 3.

Tax Frontier and Technical Efficiency of Non-GST Taxes from 2001–2002 to 2016–2017

Dependent Variable: Log (non-goods and service tax share in GSDP)	Coefficient	SE	p-Value
Constant	−2.831***	0.449	0.000
Log per capita NSDP (constant price)	0.119**	0.048	0.013
Log agriculture share in GSDP (current price)	0.406***	0.046	0.000
Log working age population share in total population	0.673***	0.158	0.000
Log road density	0.007	0.028	0.807
Log bank credit by SCBs share in GSDP	0.261***	0.045	0.000
Inefficiency equation
Constant	−1.249**	0.685	0.068
Log social sector expenditure share in GSDP_(t-1)	−0.776***	0.514	0.000
Log share of central transfers in GSDP_(t-1)	2.190***	0.624	0.000
Achievement of FRBMA target	−0.066	0.265	0.265
Sigma n	0.061***	0.020	0.002
Lambda	1.053***	0.021	0.000
Log-likelihood	382.920
Number of observations	272

Source: The authors.

Note: *, ** and *** indicate values significant at 10%, 5% and 1% levels, respectively.

Sigma_u and lambda values are also significant in the model, which determines the model’s technical inefficiency. The results of this inefficiency equation are also similar to the previous inefficiency equation. The share of social sector spending in GSDP and central transfers in GSDP are significant determinants of tax effort. In contrast, FRBMA is insignificant in determining tax effort.

Table 4 presents the tax effort index and states’ rank for the years 2001 and 2016. The tax effort index can have any value between 0 and 1 (or 0–100%). A tax effort coefficient of zero indicates that the state is perfectly inefficient in realizing its tax potential. A coefficient of 1 represents the perfect efficiency of the state in realizing its tax potential.

Table 4.

Tax Effort and the Rank Indian States for 2001 and 2016

State	Tax Effort (own tax revenue)				Tax Effort (non-GST taxes)
State	2001	Rank	2016	Rank	2001	Rank	2016	Rank
Andhra Pradesh	0.92	8	0.90	9	0.99	9	0.94	13
Goa	0.96	3	0.88	13	0.99	5	0.97	6
Gujarat	0.96	2	0.84	16	0.99	1	0.93	16
Haryana	0.95	4	0.88	14	0.99	2	0.96	8
Karnataka	0.88	15	0.89	11	0.99	7	0.98	1
Kerala	0.89	13	0.90	10	0.99	14	0.96	10
Maharashtra	0.91	11	0.82	17	0.99	6	0.95	12
Punjab	0.85	17	0.94	5	0.99	12	0.97	2
Tamil Nadu	0.92	9	0.88	12	0.99	8	0.96	11
Bihar	0.93	7	0.97	1	0.99	15	0.94	14
Chhattisgarh	0.94	5	0.96	2	0.99	4	0.97	3
Jharkhand	0.97	1	0.95	4	0.99	10	0.94	15
Madhya Pradesh	0.90	12	0.92	7	0.99	11	0.96	9
Orissa	0.91	10	0.92	6	0.99	16	0.96	7
Rajasthan	0.93	6	0.92	8	0.99	3	0.97	5
Uttar Pradesh	0.88	14	0.95	3	0.99	13	0.97	4
West Bengal	0.85	16	0.87	15	0.99	17	0.92	17
Summary statistics of tax effort from 2001 to 2016
Mean	0.89	Min.	0.77		Mean	0.95	Min.	0.78
Std. Dev.	0.05	Max	0.99		Std. Dev.	0.04	Max	0.99

Source: The authors.

Considering the whole sample, the tax effort index of states’ tax revenue ranges between 77% and 99% with a mean of 0.89 (or 89%). The tax effort index for OTR collection increased in most of the sample states. It is noteworthy that relatively more prosperous states such as Andhra Pradesh, Goa, Gujarat, Haryana, Maharashtra and Tamil Nadu lost their fiscal space between 2001 and 2016, while relatively poor states such as Bihar, Chhattisgarh, Uttar Pradesh and Madhya Pradesh have benefited from their tax efforts. The actual tax–GSDP ratio in the relatively poor states has increased at a higher rate than in the more prosperous states. Bihar, Chhattisgarh, Uttar Pradesh and Madhya Pradesh have a significant share of agricultural activity in their respective GSDP. It confirms that agriculture-driven economic growth has created a tax base for indirect taxation in the states despite tax exemption on agriculture income. Bihar, Chhattisgarh and Uttar Pradesh were the top three states in realizing their tax potential, while Maharashtra, Gujarat and West Bengal were the bottom three states.

On the other hand, the tax effort index for non-GST taxes ranges between 0.78 and 0.99 in the entire sample. All states realized their full tax potential from the non-GST taxes in 2001; however, the tax effort declined in every state in 2016. Karnataka, Punjab and Chhattisgarh were the top three states, respectively, in realizing their tax potential from non-GST, while West Bengal, Gujarat and Jharkhand had the least tax effort. It is also noteworthy that all states are achieving at least 90% of their tax effort from their legislative taxes, which represents that there is minimal scope to increase the tax effort from non-GST taxes.

V. Conclusion

GST has reduced the states’ legislative powers of taxation. The number of CSSs has reduced significantly, and state governments’ share in the schemes has increased from 50% to 60% (Reddy, 2019). Moreover, the subnational governments undertake more than half of the nation’s total expenditure (Bagchi, 2003; Rao et al., 1999; Sen & Dash, 2013). With reduced autonomy and increased responsibility, the state governments have to achieve fiscal consolidation. Therefore, subnational governments depend on their limited legislative taxes to realize their tax potential.

The stochastic frontier and inefficiency estimation results show that per capita income, agriculture activity, labour force, infrastructure development and credit extension by the scheduled commercial banks (SCBs) are the significant determinants of tax capacity. An improvement in these variables creates a tax base. Simultaneously, an increase in social sector spending increases the tax effort and an increase in the central transfer limits the tax realization of the potential.

The tax effort index for OTR shows that the states with a higher share of agriculture activity had relatively greater technical efficiency. It confirms the hypothesis that agriculture-driven economic growth has resulted in an increased tax base. More importantly, there is minimal scope for states to increase their tax realization from non-GST/states’ legislative taxes as all of the states were achieving more than 90% of their tax effort. However, the government should increase their tax base and strengthen their administrative set up to realize taxes from their limited legislative taxes. The state governments should also look at non-tax sources to create fiscal space.

Footnotes

Acknowledgements

We are grateful to Prof. Lakhwinder Gill and anonymous referees for their insightful comments. Discussions with Mr Madan Dhanora, Mr Nawazuddin Ahmed, Dr Amey Sapre and Dr Era Tiwari helped us refine the work and we gratefully acknowledge their contributions. However, it needs to be noted that the writers are solely liable for any inaccuracies.

Declaration of Conflicting Interests

The authors declared no potential conflicts of interest with respect to the research, authorship and/or publication of this article.

Funding

The authors received no financial support for the research, authorship and/or publication of this article.

Appendix

Appendix A.

Variables and Data Sources

S. No.	Variable	Data Source
1	Own tax revenue	RBI (2019): Handbook of Statistics on Indian States
2	Real per capita NSDP
3	Agricultural GSDP
4	Length of roads
5	Credit by SCBs
6	Social sector expenditure
7	States’ GSDP
8	Labour force	Labour Bureau’s Annual Employment and Unemployment Survey: Various Issues NSSO’s Employment and Unemployment Survey: Various Issues
9	FRBMA achievement	RBI: State Finances: A Study of Budgets: Various Issues
10	Introduction of VAT in states	RBI: State Finances: A Study of Budgets: Various Issues
11	Taxes on profession	EPW Research Foundation (2019): EPWRF Time Series
12	State excise duty
13	Stamp duty
14	Motor vehicle tax
15	Electricity duty

Source: Summary statistics are based on Reserve Bank of India (RBI) and Central Statistics Office (CSO) data.

Notes

References

Aigner

, Lovell

C. K.

, & Schmidt

(1977). Formulation and estimation of stochastic frontier production function models. Journal of Econometrics, 6(1), 21–37.

Alfirman

(2003). Estimating stochastic frontier tax potential: Can Indonesian local governments increase tax revenues under decentralisation? Center for Economic Analysis, Department of Economics, University of Colorado.

Bagchi

(2003). Fifty years of fiscal federalism in India, An appraisal (Working Paper No. 02). NIPFP.

Bahl

R. W.

(1972). A representative tax system approach to measuring tax effort in developing countries. Staff Papers, 19(1), 87–124.

Basu

D. D.

(1982). Introduction to the constitution of India. Prentice-Hall of India.

Battese

G. E.

, & Coelli

T. J.

(1995). A model for technical inefficiency effects in a stochastic frontier production function for panel data. Empirical Economics, 20(2), 325–332.

Baunsgaard

, & Keen

(2010). Tax revenue and (or?) trade liberalisation. Journal of Public Economics, 94(9–10), 563–577.

Belotti

, Daidone

, Ilardi

, & Atella

(2013). Stochastic frontier analysis using Stata. The Stata Journal, 13(4), 719–758.

Bhatia

H. L.

(1996). Public finance. Vikas Publishing House.

10.

Bird

R. M.

, Martinez-Vazquez

, & Torgler

(2008). Tax effort in developing countries and high-income countries: The impact of corruption, voice and accountability. Economic Analysis and Policy, 38(1), 55–71.

11.

Coondoo

, Majumber

, Mukherjee

, & Neogi

(2001). Relative tax performances: Analysis for selected states in India. Economic and Political Weekly, 3869–3871.

12.

Crivelli

, & Gupta

(2014). Resource blessing, revenue curse? Domestic revenue effort in resource-rich countries. European Journal of Political Economy, 35, 88–101.

13.

Cyan

, Martinez-Vazquez

, & Vulovic

(2013). Measuring tax effort: Does the estimation approach matter, and should effort be linked to expenditure goals? (ICEPP Working Paper). Georgia State University.

14.

Dwibedi

J. K.

, Marjit

, & Koushik

K. H.

(2016). Tax effort of West Bengal and Andhra Pradesh: A comparative study based on consumption expenditure. Economic and Political Weekly, LI(7), 63–69.

15.

Finance Commission. (2015). Report of the fourteenth finance commission (Vol. 1). Finance Commission.

16.

Garg

, Goyal

, & Pal

(2017). Why tax effort falls short of tax capacity in Indian states: A Stochastic Frontier approach. Public Finance Review, 45(2), 232–259.

17.

Gemmell

(2001). Fiscal policy in a growth framework (Discussion Paper No. 2001/84). WIDER.

18.

Greene

(2005). Reconsidering heterogeneity in panel data estimators of the stochastic frontier model. Journal of Econometrics, 126(2), 269–303.

19.

Grigorian

M. D. A.

, & Davoodi

M. H. R.

(2007). Tax potential vs. tax effort: A cross-country analysis of Armenia’s stubbornly low tax collection (No. 7-106). International Monetary Fund.

20.

Gupta

A. S.

(2007). Determinants of tax revenue efforts in developing countries (No. 7-184). International Monetary Fund.

21.

Jha

, Mohanty

M. S.

, Chatterjee

, & Chitkara

(2000). Tax efficiency in selected Indian states. In Boadway

& Raj

(Eds.), Advances in public economics (pp. 91–104). Physica.

22.

Jondrow

, Lovell

C. K.

, Materov

I. S.

, & Schmidt

(1982). On the estimation of technical inefficiency in the stochastic frontier production function model. Journal of Econometrics, 19(2–3), 233–238.

23.

Karnik

, & Raju

(2015). State fiscal capacity and tax effort: Evidence for Indian States. South Asian Journal of Macroeconomics and Public Finance, 4(2), 141–177.

24.

Kumbhakar

S. C.

(1990). Production frontiers, panel data, and time-varying technical inefficiency. Journal of Econometrics, 46(1–2), 201–211.

25.

Kawadia

, & Philips

S. S.

(2014). Shedding the BIMARU tag. Economic and Political Weekly.

26.

Langford

, & Ohlenburg

(2015). Tax revenue potential and effort (Working Paper). International Growth Centre. https://www.theigc.org/wp-content/uploads/2016/01/Langford-Ohlenburg-2016-Working-paper.pdf

27.

T. M.

, Moreno-Dodson

, & Bayraktar

(2012). Tax capacity and tax effort: Extended cross-country analysis from 1994 to 2009. World Bank Policy Research Working Paper, 6252.

28.

Lee

Y. H.

, & Schmidt

(1993). A production frontier model with flexible temporal variation in technical efficiency. The measurement of productive efficiency: Techniques and applications, 237–255.

29.

Lotz

J. R.

, & Morss

E. R.

(1967). Measuring “tax effort” in developing countries. Staff Papers, 14(3), 478–499.

30.

Mikesell

(2007). Changing state fiscal capacity and tax effort in an era of devolving government, 1981–2003. Publius: The Journal of Federalism, 37(4), 532–550.

31.

Mukherjee

(2017). Changing tax capacity and tax effort of Indian States in the Era of high economic growth, 2001–2014 (No. 17/196). NIPFP.

32.

Nayudu

A. S. H.

(2019). Tax revenue efficiency in India States: The case of stamp duty and registration fees (No. 19/278). NIPFP.

33.

Oommen

M. A.

(1987). Relative tax effort of states. Economic and Political Weekly, 22(11), 466–470.

34.

Pessino

, & Fenochietto

(2010). Determining countries’ tax effort. Hacienda Pública Española/Revista de Economía Pública, 195(4), 65–87.

35.

Piancastelli

(2001). Measuring the tax effort of developed and developing countries: Cross country panel data analysis-1985/95 (Working Paper No. 818). IPEA.

36.

Pitt

M. M.

, & Lee

L. F.

(1981). The measurement and sources of technical inefficiency in the Indonesian weaving industry. Journal of Development Economics, 9(1), 43–64.

37.

Purohit

M. C.

(2006). Tax efforts and taxable capacity of central and state governments. Economic and Political Weekly, 41(8), 747–755.

38.

Purohit

B. C.

(2014). Efficiency of social sector expenditure in India. Routledge.

39.

Rao

(1993). Taxable capacity tax-efforts and forecasts of tax-yield of Indian States. Institute for Social and Economic Change.

40.

Rao

M. G.

, Shand

R. T.

, & Kalirajan

K. P.

(1999). Convergence of incomes across Indian states: A divergent view. Economic and Political Weekly, 34(13), 769–778.

41.

Reddy

K. N.

(1975). Inter-state tax effort. Economic and Political Weekly, 10(50), 1916–1924.

42.

Reddy

Y. V.

(2019). New approaches to fiscal federalism in India. Review of Development and Change, 24(2), 163–186.

43.

Sawhney

(2018). An analysis of fiscal policy in an emerging economy: Innovative and sustainable fiscal rules in India. Millennial Asia, 9(3), 295–317.

44.

Schmidt

, & Sickles

R. C.

(1984). Production frontiers and panel data. Journal of Business & Economic Statistics, 2(4), 367–374.

45.

Sen

T. K.

(1997). Relative tax effort by Indian states (Working Paper No, 5). National Institute of Public Finance and Policy (NIPFP).

46.

Sen

, & Dash

S. K.

(2013). Fiscal imbalances and indebtedness across Indian states: Recent trends (Working Paper, 2013-119). National Institute of Public Finance and Policy.

47.

Sengupta

D. P.

, & Rao

R. K.

(2012). Direct Tax Code and taxation of agricultural income: A missed opportunity. Economic and Political Weekly, 47(15), 51–60.

48.

Suhag

, & Tiwari

(2018). State of state finances. PRS Legislative Research Institute for Policy Research Studies.

49.

Tanzi

(1992). Structural factors and tax revenue in developing countries: A decade of evidence. Cambridge University Press.

50.

The U.S. Advisory Commission on Intergovernmental Relations. (1962). Measures of state and local fiscal capacity and tax effort (Staff Report). https://library.unt.edu/gpo/acir/Reports/information/M-16.pdf

51.

Thimmaiah

(1979). Revenue potential and revenue efforts of southern states. Oxford & IBH Pub. Co.

52.

Trivedi

, & Rajmal. (2011). Growth effects of fiscal policy of India States. Millennial Asia, 2(2), 141–162.

53.

Wang

H. J.

, & Schmidt

(2002). One-step and two-step estimation of the effects of exogenous variables on technical efficiency levels. Journal of Productivity Analysis, 18(2), 129–144.