Evaluating Public Spending Efficiency and its Determinants Towards Social Outcomes: Empirical Evidence from Indian States

Abstract

The topic of public expenditure efficiency has gained a key position in the spheres of public policy goals. This article tries estimating the efficiency of public spending on health and education sectors across Indian states from 1990–1991 to 2016–2017. In input–output linkage, we have computed efficiency using a slacks-based measure model of data envelopment analysis through different orientations. The study also captures the externalities of public spending efficiency with some exogenous factors. It also suggests that these exogenous factors (such as good governance, per capita income and proportion of literates in the household) have significantly improved the output efficiency score of the health as well as the education sectors across Indian states.

JEL Codes: H51, H52, H21, E61

Keywords

Public health spending public education spending efficiency fiscal consolidation

I. Introduction

Public spending is considered one of the real drivers of economic growth in any country across the globe. Studies based on assessing the efficacy of social sector spending along with its associated outcomes have a significant bearing on public policies towards social infrastructure (Panagariya et al., 2014). In public finance literature, governments are considered as producers that produce different outputs with different inputs. A government that produces more outputs with fewer inputs is more efficient than a government that generates fewer outputs with more inputs, while keeping other things constant (Gupta & Verhoeven, 2001).

The topic of efficiency has gained a prominent position in the field of government activities in the form of taxing and public spending across countries (Tanzi, 2004). Due to the lower level of public spending in the case of emerging and developing countries, a small improvement in efficiency would generate substantial resources and thereby attain public objectives (Herrera & Ouedraogo, 2018). In India, the percentage share of total public spending by the states (on average) accounted for 17% of gross domestic product (GDP) in 2016–2017. In public spending, social expenditure constitutes 43.4% by states during this period. Hence, a small improvement in the efficiency of public spending would create sizable resources to achieve better social outcomes. Further, to adhere to fiscal consolidation targets as per fiscal responsibility legislations (FRLs), budgetary constraints faced by the states have put the objective of improvement in public spending efficiency.

The literature affirms that public social spending, mainly health and education, has a more direct bearing on the accumulation of human capital, which leads to potential growth. Therefore, assessing the efficacy of growth-oriented sectors such as health and education assumes critical importance. Hence, this induces our attention towards estimating the efficiency of public spending in these areas. Besides this, we have also attempted to analyse whether Indian states are efficient in performing their spending activities. Also, how can we achieve maximum social output goals through some exogenous factors while using the same spending level?

This study is divided into six sections, including the present one. Section II provides a review of the literature. The sources of the data and methodology applied to work out the technical efficiencies of public spending are explained in section III. Section IV presents the results of the data envelopment analysis (DEA). Section V deals with the Tobit and ordinary least squares (OLS) regressions applied to factors affecting spending efficiency. Section VI presents conclusions and policy suggestions.

II. Literature Review

The literature related to assessing the efficiency of public expenditure has been extensive in the case of developed countries, but there are limited contributions in the case of efficiency in public spending that focus on emerging and developing economies.

Pioneered contributions were led by Afonso and Aubyn (2006) and Gupta and Verhoeven (2001). These studies constructed an index of public sector efficiency by considering public expenditure about various socioeconomic indicators that are desired outcomes for public spending. The studies have employed a non-parametric technique and computed efficiency scores based on the relative difference of inefficient units from the best practices. Another stream of literature deals with specific areas of public service, mostly in health, education, infrastructure, public administration and social protection. In this context, many studies have explored the efficiency, mainly in advanced countries such as the OECD and European countries (Adam et al., 2011; Afonso & Schuknecht, 2019; Dutu & Sicari, 2020; Joumard et al., 2010; Mandl et al., 2008; Sutherland et al., 2009). Further, few studies have analysed these sector-specific efficiency levels in emerging and developing economies (Aparicio et al., 2020; Kapsoli & Teodoru, 2017; Ollivaud, 2017).

Additionally, the major discussion about exploring the determinants of public spending efficiency adds another layer to the literature over the last decade. For instance, Afonso and Aubyn (2006) incorporated property rights, education level, income level and competitive civil service which affect efficiency in European countries. Agasisti (2014) has used technological literacy in addition to socioeconomic variables that positively affect public educational efficiency. Moreno-Enguix and Lorente Bayona (2017) suggested that the state of economic development (GDP), corruption and democracy contribute to higher public sector efficiency. Antonelli and Bonis (2019) explained that higher efficiency in social spending is associated with higher educational and GDP levels, whereas lower population size and lower corruption tend to low efficiency. However, in the case of emerging and developing economies, important studies are enlisted. Rayp and Sijpe (2007) found that civil liberties, rule of law, political stability and urbanisation improve efficiency, while the illiteracy rate and the young population are associated with lower efficiency. Fonchamnyo and Sama (2016) showed that the quality of budgetary allocation has been positively associated with public spending efficiency, while the corruption level has adversely affected the efficiency in the health and education sectors. Grigoli and Kapsoli (2018) confirmed that the level of schooling, population density and sanitation facility have enhanced efficiency scores, while alcoholic consumption, TB diffusion and HIV diffusion have lowered the same.

Although many studies have measured the public spending efficiency at the country level and at the international level; however, few studies are focused only on sub-national government, particularly in emerging and developing countries. In this context, Wodon and Jayasuriya (2007) assessed the efficiency of public health and education spending through the parametric stochastic frontier analysis (SFA) technique on panel data at the sub-national level in both Argentina and Mexico. Per capita income has a favourable effect on health efficiency in both countries. In the education sector, per capita income has contributed positively, but not significantly, to educational outcomes. Although the adult literacy rate affected positively and significantly enrolment ratios but this is not true in testing scores. Lavado and Cabanda (2009) analysed the efficiency of both health and education spending using non-parametric techniques at the provincial level in the Philippines. They explained that there exists potential in efficiency in both sectors. Also, the study suggests that access to safe water positively affects efficiency, whereas fiscal grants and the Gini coefficient adversely affect the same. Campoli et al. (2019) evaluated the efficiency of public education spending through DEA at the sub-national units in Brazil. They found that the inefficient federal units can increase the efficiency of education spending by 4.83% on average in 2011 and by 4.87% in educational outcomes by using the same public resources.

However, in India, only two studies have gauged the efficiency of public spending across Indian states. Yadava and Neog (2019) analysed the efficiency across 19 Indian states during 2006–2015. According to this study, the input and output efficiency scores (57.8% and 85%, respectively) explained that the states have huge potential to increase efficiency if they follow the best practices of peer states. Mohanty and Bhanumurthy (2020) have also assessed the efficiency in the case of health, education and overall social sector expenditure and their respective outcomes among 27 Indian states from the period 2000–2001 to 2015–2016. This study further suggested that public spending efficiency could be enhanced with other factors such as governance, economic growth as well as mother’s schooling. However, these studies have used basic (BCC) models of DEA to compute technical efficiency scores for selected sectors. Therefore, they neglect the slacks in terms of excess inputs and shortfalls in outputs to get an improved efficiency score. Against this background, the study re-examines the efficiency of public social spending among Indian states using more advanced slacks-based measure (SBM) models in the DEA framework. The main advantage of the SBM model in the context of public finance is to mitigate the wastage of public spending by removing slacks on the input side and to enhance social indicators by removing slacks on the output case to get refined efficiency scores.

III. Data and Methodology

Data Descriptions

This study is based on secondary data, culled from different sources for 27 Indian states over the period from 1990–1991 to 2016–2017. Since the availability of data is not continuous (time series), we have chosen a cross-sectional data set to compute social spending efficiency. In the data set, we considered each reference year due to data comparability. Data related to public spending (as inputs) have been compiled from an RBI report titled ‘State Finance: Study of Budgets’, while social indicators (as outputs) were retrieved from two sources such as ‘EPWRF time series database’ and ‘Sample Registration System (SRS) bulletins’, released by the Registrar General of India (RGI). Data on the governance index have been culled from a report ‘Public Affairs Index’ published annually by the Public Affairs Centre at the state level. In the input–output linkage, the relation between inputs and outputs is not considered to be contemporaneous with each other. Therefore, we have followed a common methodology by estimating three years (including the previous two years) averages of public spending from the input side and the current year of health and education outputs from the output side (Lavado & Cabanda, 2009).

DEA Framework

In the DEA literature, two broad categories of models are radial (BCC or CCR) and non-radial or SBM model. The SBM model of DEA was proposed by Tone (2001). Additionally, the SBM model rejects the hypothesis of proportional changes of all inputs or outputs and captures slacks explicitly on efficiency. Therefore, the study has used the SBM model of DEA to compute efficiency. Figure 1a shows inputs on the x-axis and outputs on the y-axis. The formation of the frontier is derived from efficient DMUs (i.e., K, O, L and N). However, DMU ‘M’ is inefficient. The inefficient DMU can become efficient by covering horizontal (input), vertical (output) and diagonal (non-oriented) distances to the frontier. In Figure 1b, each axis represents the input. DMUs ‘D’ and ‘G’ are inefficient from a slack perspective. If we move ‘D’ to ‘E’, DE amount of X₂ input is reduced to produce the same output. Similarly, output slack can be determined. It is imperative to discuss ‘how non-oriented SBM is different from directional technology distance function (DDF)’. The advantage of the non-oriented SBM model over DDF is to internalise slacks in computing efficiency. However, the limitation of non-oriented SBM is that it strictly assumes that all inputs and outputs are positive, whereas DDF allows flexibility to include some inputs and outputs as zero.

In the next section, we discuss the empirical results based on the SBM method.

IV. Empirical Results

This empirical analysis permits us to measure the extent to which public spending in the case of health and education is unproductive relative to the estimated efficiency frontier that explains the distance between the estimated state and the efficient state that lies on the frontier.

Health

The calculation of efficiency scores of the health sector has been estimated through two inputs, that is, health expenditure to gross state domestic product ratio (HE/GSDP) and non-health expenditure to GSDP ratio (NHE/GSDP) for four sub-periods due to data comparability. On the other hand, three outputs, namely, life expectancy (LE), infant mortality rate (IMR) and under-five mortality rate (U5SR), are used. However, mortality indicators are used as good outputs in the form of infant survival rate (ISR) and under-five survival rate (U5SR) are expressed as

ISR = (1000 - IMR) /IMR .

Similarly, U5SR = (1000 - U5MR) /U5MR .

Before moving to the final model of efficiency (i.e., multiple inputs and multiple outputs), we have computed efficiency with a single input such as health spending to GSDP ratio with every single output indicator during 2016–2017. The results of single input and single output analysis (Table 1) reveal that Kerala and Maharashtra are efficient with unit efficiency scores through all different approaches. On average, the output efficiency score is higher than the input and non-oriented efficiency score. Also, we have seen more divergences in output efficiency than input and non-oriented efficiency. It is observed that on average states are more efficient in the LE indicator, while they are highly inefficient in attaining ISR and U5SR indicators in the output orientation. Further, we have also computed efficiency scores with multiple inputs and multiple outputs for four reference years shown in Table 2.

Table 1.

The Results of Slacks-based Measure (SBM) Efficiency (Health) in Single Input Single Output Framework (2016–2017).

States (DMUs)	Input-oriented			Output-oriented			Non-oriented
States (DMUs)	LE	ISR	U5SR	LE	ISR	U5SR	LE	ISR	U5SR
Andhra Pradesh	0.56	0.56	0.56	0.93	0.29	0.23	0.54	0.29	0.22
Assam	0.34	0.34	0.34	0.87	0.22	0.18	0.31	0.13	0.11
Bihar	0.35	0.35	0.35	0.91	0.26	0.23	0.33	0.16	0.14
Chhattisgarh	0.38	0.38	0.38	0.87	0.25	0.19	0.34	0.17	0.13
Gujarat	0.77	0.77	0.77	0.95	0.46	0.38	0.74	0.46	0.38
Haryana	0.83	0.83	0.83	0.95	0.46	0.39	0.80	0.46	0.39
Himachal Pradesh	0.51	0.50	0.50	0.96	0.39	0.47	0.51	0.35	0.42
Jammu and Kashmir	0.30	0.22	0.22	0.98	0.41	0.38	0.30	0.16	0.15
Jharkhand	0.46	0.46	0.46	0.90	0.34	0.28	0.43	0.28	0.23
Karnataka	0.80	0.80	0.80	0.95	0.61	0.51	0.77	0.61	0.51
Kerala	1	1	1	1	1	1	1	1	1
Madhya Pradesh	0.65	0.65	0.65	0.88	0.24	0.19	0.59	0.24	0.19
Maharashtra	1	1	1	1	1	1	1	1	1
Odisha	0.47	0.47	0.47	0.90	0.22	0.19	0.44	0.18	0.16
Punjab	0.76	0.71	0.71	0.98	0.61	0.52	0.76	0.61	0.52
Rajasthan	0.46	0.46	0.46	0.91	0.24	0.20	0.43	0.19	0.16
Tamil Nadu	0.73	0.80	0.74	0.97	0.78	0.71	0.72	0.78	0.71
Uttar Pradesh	0.36	0.36	0.36	0.86	0.22	0.17	0.33	0.15	0.11
Uttarakhand	0.54	0.54	0.54	0.95	0.26	0.34	0.53	0.24	0.32
West Bengal	0.56	0.56	0.56	0.94	0.39	0.40	0.55	0.39	0.39
Mean	0.59	0.58	0.58	0.93	0.43	0.39	0.57	0.39	0.36

Note: LE, life expectancy; ISR, infant survival rate; U5SR, under-five survival rate.

Table 2.

The Results of Slacks-based Measure (SBM) Efficiency (Health) with Multiple Inputs Multiple Outputs.

States (DMUs)	Input-oriented				Output-oriented				Non-oriented
States (DMUs)	1990	2000	2010	2016	1990	2000	2010	2016	1990	2000	2010	2016
Andhra Pradesh	0.44	0.44	0.37	0.54	0.30	0.28	0.36	0.33	0.44	0.16	0.17	0.30
Assam	0.72	0.78	0.48	0.44	0.28	0.26	0.25	0.27	0.72	0.25	0.15	0.19
Bihar	0.39	0.37	0.45	0.36	0.26	0.25	0.31	0.32	0.39	0.12	0.18	0.19
Chhattisgarh	–	–	–	0.45	–	–	–	0.29	–	–	–	0.21
Gujarat	0.72	0.71	0.89	0.88	0.33	0.29	0.49	0.64	0.72	0.26	0.48	0.57
Haryana	1	1	0.91	0.79	1	1	0.58	0.52	1	1	0.55	0.47
Himachal Pradesh	–	0.41	0.41	0.49	–	0.34	0.43	0.53	–	0.17	0.22	0.41
Jammu and Kashmir	–	–	0.28	0.33	–	–	0.37	0.49	–	–	0.13	0.21
Jharkhand	–	–	–	0.48	–	–	–	0.39	–	–	–	0.31
Karnataka	1	1	0.89	0.78	1	1	0.57	0.65	1	1	0.56	0.61
Kerala	1	1	1	1	1	1	1	1	1	1	1	1
Madhya Pradesh	0.50	0.57	0.56	0.54	0.17	0.18	0.23	0.28	0.50	0.14	0.17	0.23
Maharashtra	0.82	1	1	1	0.54	1	1	1	0.82	1	1	1
Odisha	0.82	0.70	0.70	0.50	0.28	0.19	0.27	0.27	0.82	0.17	0.24	0.22
Punjab	0.28	0.75	0.81	0.70	0.34	0.32	0.64	0.65	0.28	0.29	0.57	0.58
Rajasthan	0.25	0.72	0.68	0.49	0.23	0.20	0.27	0.29	0.25	0.19	0.23	0.23
Tamil Nadu	0.25	0.85	0.80	0.79	0.35	0.46	0.63	0.81	0.25	0.46	0.61	0.76
Uttar Pradesh	0.23	0.76	0.48	0.40	0.19	0.25	0.23	0.26	0.23	0.22	0.14	0.17
Uttarakhand	–	–	–	0.63	–	–	–	0.42	–	–	–	0.39
West Bengal	0.24	0.63	0.62	0.58	0.32	0.34	0.50	0.49	0.24	0.27	0.39	0.46
Mean	0.58	0.73	0.67	0.61	0.44	0.46	0.48	0.50	0.58	0.42	0.40	0.43

It is observed from Table 2 that in the case of an input-oriented approach, Kerala has emerged as a technically efficient state for four reference periods. Since 2000, Maharashtra has also been registered as technically efficient as it has secured an efficiency score equal to unity, which implies that the state has incurred the smallest amount of HE to achieve the same output level. Further, Haryana and Karnataka were found to be in technically efficient state until the second sub-period (2000–2001), after that, they became inefficient as reported by their efficiency scores to be less than unity. In 2016, Jammu and Kashmir was found to be the least efficient state with an efficiency score of 0.33, which means that it utilises only 33% of spending on health efficiently and the remaining 67% is used inefficiently. Furthermore, all states on average have improved their input efficiency score from 0.58 in 1990 to 0.61 in 2016, explaining that the states are improving their efficiency score over time but still have huge potential for reducing their spending level by 39% to achieve the same level of health outputs.

The output-oriented approach has also reported a similar result. For the whole period, Kerala and Maharashtra have emerged as efficient states since 2000–2001 as these states have achieved the maximum level of health outputs while using the same level of health inputs (spending). In 2016, the most inefficient state was Uttar Pradesh with an efficiency score of 0.26, which means that 74% fewer outputs were produced with the same resources. Thus, all states on average have improved their health efficiency score from 0.44 in 1990 to 0.50 in 2016. This indicates that all states on average are still inefficient at a 50% level, which means that on average they can improve their outputs by 50% while using the same level of resources.

When we carried out the non-oriented analysis, the same picture also emerged. Again, Kerala was labelled as the most efficient state throughout the period, explaining that it has simultaneously achieved the highest level of health outputs and reduced its health spending to the lowest. The average non-oriented efficiency score decreased from 0.58 in 1990 to 0.43 in 2016. In 2016, the average efficiency score was reported to be 0.43, indicating that on average, the states have huge potential for improving by 57% in terms of both the expansion of outputs and reduction of inputs.

Education

To compute the educational efficiency, two outputs and two inputs have been used in the DEA set-up. It considers two output indicators, namely, gross enrolment ratio (GER) for school education and GER for higher education, while education expenditure to GSDP ratio (EE/GSDP) and non-education expenditure to GSDP ratio (NEE/GSDP) are considered two inputs for measuring educational efficiency.

The analysis of single input single output efficiency (Table 3) suggests that Gujarat and Tamil Nadu have been found to be the most efficient in each input–output model, while Himachal Pradesh and Maharashtra were found to be efficient only in the GER school indicator during 2016–2017. From output and non-oriented specification, states are more efficient in the GER school than in the GER higher education indicator.

Table 3.

The Result of Slacks-based Measure (SBM) Efficiency (Education) in Single Input Single Output Framework (2016–2017).

States (DMUs)	Input-oriented		Output-oriented		Non-oriented
States (DMUs)	GER (School)	GER (Higher)	GER (School)	GER (Higher)	GER (School)	GER (Higher)
Andhra Pradesh	0.65	0.69	0.77	0.69	0.62	0.54
Arunachal Pradesh	0.25	0.25	0.80	0.62	0.25	0.18
Assam	0.34	0.34	0.70	0.37	0.33	0.16
Bihar	0.34	0.34	0.61	0.31	0.29	0.13
Chhattisgarh	0.45	0.41	0.79	0.34	0.44	0.18
Goa	0.86	0.71	0.99	0.60	0.86	0.51
Gujarat	1	1	1	1	1	1
Haryana	0.93	0.90	0.91	0.72	0.91	0.72
Himachal Pradesh	1	0.51	1	0.78	1	0.44
Jammu and Kashmir	0.34	0.36	0.63	0.55	0.30	0.24
Jharkhand	0.57	0.57	0.64	0.38	0.48	0.28
Karnataka	0.92	0.94	0.89	0.82	0.89	0.82
Kerala	0.80	0.66	0.97	0.52	0.80	0.42
Madhya Pradesh	0.49	0.48	0.72	0.43	0.47	0.26
Maharashtra	1	0.92	1	0.76	1	0.76
Manipur	0.35	0.33	0.89	0.75	0.35	0.27
Meghalaya	0.34	0.33	0.77	0.50	0.33	0.20
Mizoram	0.27	0.24	0.87	0.52	0.27	0.15
Nagaland	0.25	0.25	0.58	0.35	0.20	0.12
Odisha	0.55	0.56	0.71	0.45	0.53	0.32
Punjab	0.99	0.90	0.99	0.72	0.99	0.72
Rajasthan	0.60	0.54	0.80	0.44	0.58	0.31
Tamil Nadu	1	1	1	1	1	1
Tripura	0.35	0.31	0.82	0.41	0.35	0.16
Uttar Pradesh	0.43	0.44	0.70	0.53	0.41	0.29
Uttarakhand	0.71	0.68	0.90	0.71	0.71	0.55
West Bengal	0.62	0.58	0.77	0.39	0.60	0.30
Mean	0.60	0.56	0.82	0.58	0.59	0.40

Note: GER, gross enrolment ratio.

From multiple inputs and outputs analysis (Table 4), the result of the input-oriented model reveals that Maharashtra is a single state that has emerged as a fully efficient state as its efficiency score equals unity under the input-oriented SBM model. Also, Tamil Nadu and Gujarat have been labelled as efficient since 2006 and 2011, respectively. These efficient states have managed to reduce their spending to a minimum level to achieve the same level of outputs. In 2016, the most inefficient state was found to be Arunachal Pradesh with an efficiency score of 0.22. The average efficiency score of education has declined from 0.67 in 2001 to 0.60 in 2016. This indicates that states still, on average, can reduce their spending by 40% to achieve the same level of output.

Table 4.

The Results of Slacks-based Measure (SBM) Efficiency (Education) in the Multiple Inputs Multiple Outputs Framework.

States (DMUs)	Input-oriented				Output-oriented				Non-oriented
States (DMUs)	2001	2006	2011	2016	2001	2006	2011	2016	2001	2006	2011	2016
Andhra Pradesh	0.48	0.48	0.49	0.63	0.75	0.77	0.78	0.74	0.42	0.39	0.39	0.55
Arunachal Pradesh	0.46	0.34	0.28	0.22	0.58	0.60	0.70	0.72	0.27	0.20	0.20	0.19
Assam	0.74	0.58	0.55	0.47	0.76	0.45	0.47	0.49	0.58	0.29	0.28	0.31
Bihar	0.37	0.44	0.48	0.36	0.55	0.39	0.45	0.42	0.24	0.19	0.23	0.20
Chhattisgarh	–	0.87	0.70	0.50	–	0.58	0.41	0.49	–	0.53	0.30	0.31
Goa	1	0.58	1	0.86	1	0.74	1	0.75	1	0.45	1	0.65
Gujarat	1	0.97	1	1	1	0.62	1	1	1	0.62	1	1
Haryana	1	1	0.95	0.84	1	1	0.80	0.81	1	1	0.79	0.79
Himachal Pradesh	0.52	0.54	0.68	1	0.95	0.90	0.91	1	0.50	0.48	0.62	1
Jammu and Kashmir	0.45	0.42	0.38	0.33	0.57	0.60	0.66	0.60	0.31	0.28	0.26	0.25
Jharkhand	–	0.57	0.66	0.53	–	0.52	0.39	0.48	–	0.33	0.27	0.33
Karnataka	1	0.97	0.98	0.87	1	0.96	0.93	0.86	1	0.96	0.93	0.81
Kerala	0.79	1	0.95	0.84	0.69	1	0.74	0.68	0.63	1	0.72	0.59
Madhya Pradesh	0.64	0.86	0.72	0.47	0.71	0.84	0.69	0.55	0.54	0.72	0.50	0.33
Maharashtra	1	1	1	1	1	1	1	1	1	1	1	1
Manipur	0.29	0.47	0.31	0.30	0.88	0.88	0.86	0.84	0.28	0.41	0.26	0.29
Meghalaya	0.41	1	0.44	0.34	0.70	1	0.60	0.62	0.34	1	0.27	0.27
Mizoram	1	0.23	0.23	0.26	1	0.73	0.66	0.67	1	0.17	0.15	0.21
Nagaland	0.23	0.26	0.27	0.24	0.58	0.52	0.50	0.45	0.16	0.15	0.15	0.14
Odisha	0.65	0.82	0.71	0.55	0.78	0.62	0.55	0.56	0.59	0.54	0.42	0.40
Punjab	0.75	0.78	1	0.78	0.77	0.64	1	0.83	0.62	0.55	1	0.73
Rajasthan	0.85	0.77	0.80	0.55	0.64	0.53	0.60	0.57	0.58	0.43	0.51	0.40
Tamil Nadu	0.97	1	1	1	0.96	1	1	1	0.93	1	1	1
Tripura	0.24	0.32	0.37	0.33	0.57	0.50	0.47	0.56	0.16	0.16	0.18	0.23
Uttar Pradesh	0.70	0.64	0.60	0.45	0.54	0.59	0.58	0.62	0.43	0.41	0.37	0.35
Uttarakhand	–	0.61	0.79	0.75	–	0.88	0.87	0.81	–	0.55	0.71	0.71
West Bengal	0.64	0.75	0.72	0.61	0.74	0.55	0.50	0.53	0.56	0.45	0.38	0.41
Mean	0.67	0.68	0.67	0.60	0.78	0.72	0.71	0.69	0.59	0.53	0.51	0.50

The output-oriented model also yields similar results. Again, Maharashtra is the top performer throughout the period, followed by Tamil Nadu and Gujarat, which are found to be efficient states since the second and third sub-period, respectively. In 2016, the least efficient state was Bihar, with the lowest efficiency score of 0.42. Furthermore, the average efficiency score decreased from 0.78 in 2001 to 0.69 in 2016. In 2016, the average efficiency score of all states was 0.69, which explains that they can enhance their outputs by 31% while using the same level of spending.

With regard to the non-oriented model, Maharashtra is found to be efficient with the highest efficiency score equal to unity for each sub-period. However, Nagaland is the least efficient state during the same period. Also, the non-oriented model reported that, on average, the efficiency score has declined from 0.59 in 2001 to 0.50 in 2016. It indicates that enormous potential exists among all states as their average efficiency score is 0.50, which means that they can make simultaneous adjustments in terms of expanding outputs and contracting inputs up to 50% by following the best practices of efficient states during 2016–2017.

V. What Causes Variations in the Efficiency of Public Social Sector Spending?

For post-efficiency analysis, there is a need to explore some non-discretionary factors that affect the output efficiency in both sectors. In literature, three kinds of factors could affect efficiency, that is, economic, institutional and social or demographic. Hence, we consider governance (proxied as an institutional factor) has a favourable impact on the efficiency of public spending in the health and education sectors (Rajkumar & Swaroop, 2008). Also, per capita GDP (proxied for economics factor) has been positively associated with outcomes related to public services like health care, education attainment and so forth (Afonso & Aubyn, 2006; Herrera & Ouedraogo, 2018). As per theoretical perspective, we included governance indicator as governments with better institutions and lower corruption practices tend to have higher efficiency scores and vice versa. Further, per capita GDP is chosen to capture the effect of Balassa–Samuelson, implying that if high-income countries tend to be more inefficient due to persistence of higher wages, then its coefficient sign is expected to be negative. Besides this, we added another variable such as the proportion of literate in households (PLH) used as a proxy for social or demographic based on the premise that more proportion of educated people would raise the efficiency level by the successful administration of government and hence, we expect a positive impact of this variable (Hauner & Kyobe, 2010).

To conduct the analysis, we have considered an efficiency score as the dependent variable and three aforementioned factors as independent variables. Therefore, we have executed Tobit models to predict efficiency scores (estimated in DEA). Here, Tobit regression provides better results compared to OLS regression since the dependent variable is limited. To verify this, we have compared these results with the OLS model. The standard Tobit regression equation is:

y_{i}^{*} = x_{i}^{*} β + ε_{i}

Here, subscript i refers to the number of states in India. $y_{i}^{*}$ is the efficiency score and $x_{i}^{*}$ is a vector of explanatory variables; β is a vector of parameter and $ε_{i}$ is an error term. In this methodology, the effect of some exogenous factors on output efficiency only is presented in Table 5 during 2016–2017.

Table 5.

Tobit Results for Effect of Factors on Both Health and Education Efficiency (Output-oriented).

Variables	Health Efficiency				Education Efficiency
Dependent: Efficiency	Tobit (1)	Tobit (2)	OLS (3)	OLS (4)	Tobit (1)	Tobit (2)	OLS (3)	OLS (4)
GOV	2.972^** (0.01)	1.508 (0.17)	2.972^** (0.04)	1.508 (0.24)	1.616^*** (0.00)	0.673 (0.13)	1.616^** (0.01)	0.673 (0.20)
PNSDP	2.8e-06^** (0.03)		2.8e-06^*(0.06)		4.2e-06^***(0.00)		4.2e-06^***(0.00)
PLH	0.018(0.38)		0.018(0.44)		0.016^*(0.09)		0.016(0.16)
GOV*PNSDP		4.9e-06^*(0.09)		4.9e-06(0.15)		8e-06^***(0.00)		8e-06^***(0.00)
GOV*PLH		−0.004(0.93)		−0.004(0.94)		0.032(0.31)		0.032(0.38)
Constant	−1.313^*(0.06)	−0.452(0.43)	−1.313(0.11)	−0.452(0.49)	−0.552^**(0.04)	−0.098(0.63)	−0.552^*(0.08)	−0.098(0.68)
Chi-square	28.21^***	31.10^***			85.35^***	121.54^***
F-test			7.52^***	8.29^***			22.76^***	32.41^***
Adjusted R²			0.56	0.57			0.74	0.75
N	20	20	20	20	20	20	20	20

Note: Figures shown in parentheses are p values and ^***, ^** and ^* denote 1%, 5% and 10% significance levels. The coefficients are estimated using robust standard errors.

From Table 5, in the health case, the coefficients of governance and per capita economic growth are positive and significant (at 5% level). The coefficient of the proportion of household literacy was positive but not significant. Thus, good governance and higher per capita income help in improving efficiency in the health sector. In OLS, similar results have been estimated. In addition, we performed an interaction term to see its impact on efficiency. The interaction term of governance and per capita income (GOV*PNSDP) is found to affect health efficiency (in Model 2) but it has not been found significant (in Model 4).

In education, the coefficients of governance, per capita income and the PLH have positive and statistically significant effects on the efficiency score (Model 1). However, PLH is not found to be significant in OLS estimates (Model 3). Furthermore, the interaction term of governance and per capita income is positive and highly significant (at 1%), suggesting that governance along with per capita income will improve the efficiency of the education sector. It also finds that the interaction term of governance and the PLH have a positive sign, but it is not significant (Models 2 and 4). Thus, we observed that the magnitude of governance is larger than other factors, which implies that governance has a larger impact on health and educational efficiency.

VI. Conclusions and Policy Implications

In this study, we have assessed the efficiency of public spending on health and education sectors across Indian states during reference years. The results using the SBM model are similar to a previous recent study by Mohanty and Bhanumurthy (2020) related to its efficient units, but due to refined measure of efficiency score, the number of efficient units has been reported to fall down in both sectors. In the health and education sectors, the empirical analysis confirmed that Kerala, Maharashtra and Tamil Nadu have emerged as efficient states from input, output and non-oriented viewpoints. At the disaggregate level, we have seen that some states are relatively more efficient than others. This variation in efficiency can be due to some exogenous factors which are carried by the econometric technique, that is, Tobit models. In this analysis, we find that good governance, higher per capita income and PLH have a favourable effect on efficiency. Hence, we recognised that the state can be more efficient if it is good at these indicators. Furthermore, the findings suggest that, in the health case, governance is found to be positively significant with high magnitude, whereas it is not significant in previous studies led by Mohanty and Bhanumurthy (2018, 2020). Also, we introduce another variable such as PLH (proxied as schooling level) which is found to be positive and significant in the education sector.

Overall, the empirical results indicate that states are mostly inefficient except for a few ones. This inefficiency can be mitigated by higher budgetary spending by the states. Alternatively, the study also suggests some outcomes-based policy interventions. This reflects that governance, per capita income and PLH (proxied as schooling level) are key concerns if the policymakers intend to improve social outcomes. Thus, this research could improve budgetary allocations and attain better public social goals among the Indian states.

Footnotes

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.

ORCID iD

Sumit Kumar

References

Adam

, Delis

, & Kammas

(2011). Public sector efficiency: Leveling the playing field between OECD countries. Public Choice , 146(1), 163–183. https://doi.org/10.1007/s11127-009-9588-7

Afonso

, & Schuknecht

(2019). How “big” should government be ? REM Working Paper 078-2019 . REM – Research in Economics and Mathematics.

Afonso

, & St. Aubyn

(2006). Cross-country efficiency of secondary education provision: A semi-parametric analysis with non-discretionary inputs. Economic Modelling , 23(3), 476–491. https://doi.org/10.1016/j.econmod.2006.02.003

Agasisti

(2014). The efficiency of public spending on education: An empirical comparison of EU countries. European Journal of Education , 49(4), 543–557. https://doi.org/10.1111/ejed.12069

Antonelli

M. A.

, & De Bonis

(2019). The efficiency of social public expenditure in European countries: A two-stage analysis. Applied Economics , 51(1), 47–60. https://doi.org/10.1080/00036846.2018.1489522

Aparicio

, Perelman

, & Santín

(2020). Comparing the evolution of productivity and performance gaps in education systems through DEA: An application to Latin American countries. Operational Research , 22, 1443–1477. https://doi.org/10.1007/s12351-020-00578-2

Campoli

, Ferraz

, & Rebelatto

(2019). The efficiency of public expenditures on education: Data envelopment analysis of Brazilian Federative Units between 2011 and 2014. In Mula

, Barbastefano

, Díaz-Madroñero

, & Poler

(Eds.), New global perspectives on industrial engineering and management. Lecture notes in management and industrial engineering . Springer International Publishing. https://doi.org/10.1007/978-3-319-93488-4_9

Dutu

, & Sicari

(2020). Public spending efficiency in the OECD: Benchmarking health care, education, and general administration. Review of Economic Perspectives , 20(3), 253–280. https://doi.org/10.2478/revecp-2020-0013

Fonchamnyo

D. C.

, & Sama

M. C.

(2016). Determinants of public spending efficiency in education and health: Evidence from selected CEMAC countries. Journal of Economics and Finance , 40(1), 199–210. https://doi.org/10.1007/s12197-014-9310-6

10.

Grigoli

, & Kapsoli

(2018). Waste not, want not: The efficiency of health expenditure in emerging and developing economies. Review of Development Economics , 22(1), 384–403. https://doi.org/10.1111/rode.12346

11.

Gupta

, & Verhoeven

(2001). The efficiency of government expenditure: Experiences from Africa. Journal of Policy Modeling , 23(4), 433–467. https://doi.org/10.1016/S0161-8938(00)00036-3

12.

Hauner

, & Kyobe

(2010). Determinants of government efficiency. World Development , 38(11), 1527–1542. https://doi.org/10.1016/j.worlddev.2010.04.004

13.

Herrera

, & Ouedraogo

(2018). Efficiency of public spending in education, health, and infrastructure: An international benchmarking exercise. Policy Research Working Paper , 14(September). http://www.worldbank.org/research.

14.

Joumard

, André

, & Nicq

(2010). Efficiency and institutions. Health Care Systems , 769. https://doi.org/10.1787/5kmfp51f5f9t-en

15.

Kapsoli

, & Ruxandra Teodoru

(2017). Benchmarking social spending using efficiency frontiers. IMF Working Papers , 17(197), 1. https://doi.org/10.5089/9781484315309.001

16.

Lavado

R. F.

, & Cabanda

E. C.

(2009). The efficiency of health and education expenditures in the Philippines. Central European Journal of Operations Research , 17(3), 275–291. https://doi.org/10.1007/s10100-009-0095-1

17.

Mandl

, Dierx

, & Ilzkovitz

(2008). The effectiveness and efficiency of public spending. Economic Papers EU . https://doi.org/10.2765/22776

18.

Mohanty

R. K.

, & Bhanumurthy

N. R.

(2018). Assessing public expenditure efficiency at Indian states. National Institute of Public Finance and Policy Working Papers , 225.

19.

Mohanty

R. K.

, & Bhanumurthy

N. R.

(2020). Assessing public expenditure efficiency at the subnational level in India: Does governance matter? Journal of Public Affairs , 21, e2173. https://doi.org/10.1002/pa.2173

20.

Moreno-Enguix

M. D. R.

, & Lorente Bayona

L. V.

(2017). Factors affecting public expenditure efficiency in developed countries. Politics and Policy , 45(1), 105–143. https://doi.org/10.1111/polp.12194

21.

Ollivaud

(2017). Improving the allocation and efficiency of public spending in Indonesia. Marketing and Branding Research , 4(2), 10–18. https://doi.org/10.1080/09692290.2016.1262446

22.

Panagariya

, Chakraborty

, & Rao

M. G.

(2014). State level reforms, growth, and development in Indian states . Oxford University Press. https://doi.org/10.1093/acprof:oso/9780199367863.001.0001

23.

Rajkumar

A. S.

, & Swaroop

(2008). Public spending and outcomes: Does governance matter? Journal of Development Economics , 86(1), 96–111. https://doi.org/10.1016/j.jdeveco.2007.08.003

24.

Ray

S. C.

(2004.) Data envelopment analysis: Theory and techniques for economics and operations research . Cambridge University Press.

25.

Rayp

, & Van De Sijpe

(2007). Measuring and explaining government efficiency in developing countries. Journal of Development Studies , 43(2), 360–381. https://doi.org/10.1080/00220380601125230

26.

Sutherland

, Price

, & Gonand

(2009). Improving public spending efficiency in primary and secondary education. OECD Journal: Economic Studies , 2009(1), 1–30. https://doi.org/10.1787/eco_studies-v2009-art4-en

27.

Tanzi

(2004). Measuring efficiency in public expenditure. Paper presented in conference on public expenditure evaluation and growth .

28.

Tone

(2001). A slacks-based measure of efficiency in data envelopment analysis. European Journal of Operational Research , 130(3), 498–509.

29.

Wodon

, & Jayasuriya

(2007). Efficiency in improving health and education outcomes: Provincial and state-level estimates for Argentina and Mexico. Estudios Económicos , 22(1), 57–97.

30.

Yadava

A. K.

, & Neog

(2019). Public sector performance and efficiency assessment of Indian states. Global Business Review , 23(2), 493–511. https://doi.org/10.1177/0972150919862664