Download as pdf or txt
Download as pdf or txt
You are on page 1of 13

Socio-Economic Planning Sciences 68 (2019) 100667

Contents lists available at ScienceDirect

Socio-Economic Planning Sciences


journal homepage: www.elsevier.com/locate/seps

Prioritization of government expenditure on health in India: A fiscal space T


perspective
Deepak Kumar Behera∗, Umakant Dash
Department of Humanities and Social Sciences, Indian Institute of Technology Madras, Chennai, India 600036

ARTICLE INFO ABSTRACT

Keywords: Unequal distribution of fiscal resources and lower prioritization of budget towards healthcare are the most
Government health expenditure important challenges in achieving universal health coverage in India. This study has examined relationships
Fiscal space between government health expenditure and fiscal space (i.e. tax revenue, non-tax revenue, fiscal transfer, and
Universal health coverage borrowings) in twenty-one states of India for the period of 1980–2014. Our panel regression results imply that
Panel data
mobilization of tax revenue has a positive impact, while borrowings have a negative impact on the allocation of
Quantile regression
India
government expenditure on healthcare in the long-run. The panel quantile regression results show that states
associated with the low and middle level of revenue growth have been mobilizing finance through central
JEL classification:
government transfer and borrowings in short-run. Further, the panel vector error correction models show that
H51
sum of the lagged coefficients of borrowings have a greater impact on health financing process as compared to
H61
H70 other sources of fiscal space at short-run, and the speed of adjustment towards long-run equilibrium is relatively
I18 slower. The overall analysis concludes that less domestic revenue mobilization and higher dependency of bor-
C32 rowings for healthcare financing may create fiscal stress on state finances in the long-run, and thereby it could
possibly reduce the prioritization of spending. Therefore, improvement in revenue growth and proper utilization
of fiscal transfer would be appropriate policy implications from this study.

1. Introduction of fiscal space, the prioritization of expenditure towards healthcare


depends on the conducive macroeconomic policies2 in an economy
Healthcare financing is the most important agenda in the 2015- [44,59].
United Nations Sustainable Development for achieving Universal There are some pertinent reasons for increasing the policy attention
Health Coverage (UHC) that ensures equity, quality and financial risk in the generation of fiscal space for health and prioritization of health
protection in healthcare services for all needy people [12]. This was sector financing at the global level. First, very little improvement in the
followed by the 2010-World Health Report [65] that emphasis on the levels and distribution of population health outcome during the MDGs
generation of fiscal space in order to finance healthcare in low-income era-2000-2015 [12]. Second, mostly low-income and middle-income
and middle-income countries. Heller [25] and Tandon and Cashin [58] countries exhibit two major health financing trends over time: a higher
are the frontrunners for the concept of fiscal space1 and its important share of out-of-pocket (OOP) health expenditure and a lower share of
role for the generation of finance towards healthcare. They argue that government health expenditure [21]. Third, achieving health-related
the government could generate fiscal space in three important ways – developmental goals that ensure quality healthcare of all at all ages by
revenue mobilization (tax and non-tax revenue), grants (domestic and 2030.
external), and borrowings (domestic and external). After the generation India is not an exception in health system challenge and suffers a


Corresponding author.
E-mail addresses: hs15d001@smail.iitm.ac.in (D.K. Behera), dash@iitm.ac.in (U. Dash).
1
Fiscal space can be defined as “the capacity of government to provide additional budgetary resources for the desired purpose without any prejudice to the
sustainability of its financial position” [25].
2
Tandon et al. [59] and Mclntyre and Kutzin [44] argue that each source of fiscal space options may bring its own costs and benefit. While increasing revenues may
reduce fiscal constraints, but raising revenue through regressive taxes can do more harm than the good to the overall economy. Similarly, borrowing to finance
current spending may seem like a good idea in the short-run, but could become unsustainable over time because of debt tide with interest payment. External grants is
also not a viable option of generation of fiscal space in an economy because foreign aids for health sector may reduce the other sector grants such as infrastructure
development and it can create negative externalities.

https://doi.org/10.1016/j.seps.2018.11.004
Received 7 January 2018; Received in revised form 5 October 2018; Accepted 8 November 2018
Available online 16 November 2018
0038-0121/ © 2018 Elsevier Ltd. All rights reserved.
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Fig. 1. Share of government health expenditure vis-à-vis out-of-pocket health expenditure. Source: National Health Account [47], WHO.

huge shortage of government finance to deliver healthcare services Russia as well as lower than the OECD countries such as UK and US.
[33]. Fig. 1 compares the share of OOP health expenditure to the total Fig. 2 implies that India has a low level of revenue (fiscal) capacity3 and
health expenditure vis-à-vis the share of government health ex- there is also enough scope for increasing government revenues at
penditure (GHE) to general government expenditure of India in 2014. medium to a high level (35%–46%).
Fig. 1 shows the share of GHE to general government expenditure of In this paper, we examine the prioritization of state (regional)
India is 5%, which is much lower than the other low-income and government towards healthcare financing from a fiscal space perspec-
middle-income countries such as Pakistan, Indonesia, Sri Lanka, Kenya, tive for the period of 1980–2014. Fig. 3 shows the state-wise variation
and Bhutan. The result implies that the government offers low priority between GHE and Fiscal space indicators4 (i.e. tax revenue, borrowings,
to the health sector in the budget. On the contrary, the share of OOP and fiscal transfer). Fig. 3 exhibits that the mobilization funds towards
health expenditure to total health expenditure is higher than the other healthcare across states are depends not only GDP and tax revenue but
low-and middle income countries. So, in order to promote financial also central government fiscal transfer and borrowings. Earlier litera-
protection for moving towards UHC, the share of OOP health ex- ture such as Chakraborty and Dash [15], Behera and Dash [7], Sun-
penditure to total health expenditure should be minimized as a daraman et al. [57], Rao [52], Duran et al. [19], Chakraborty et al.
threshold limit of 30% [42]. [16], and Dholakia and Govil [18], have argued various fiscal space
Fig. 2 shows the share of tax revenue to Gross Domestic Product strategy for the mobilization of resources in state-level of India. Chak-
(GDP) ratio of India is 16.6%, which is comparatively lesser than the raborty et al. [16] and Rao [52] argue that the financial sector reforms
other emerging economies such as China, Brazil, South Africa, and of the central government had adversely affected the state-level fi-
nances in the post-economic liberalization period of India. As a con-
sequence, there was a sharp decline in the central government fiscal
transfer and increasing interest rate of outstanding liabilities had wi-
dened the fiscal gap. They revealed that these macroeconomic policy
shocks had reduced the fiscal space across states resulting the interstate
difference in revenue capacity in order to finance healthcare. Various
fiscal consolidation measures such as higher tax revenue, increase non-
tax revenue, increase central transfer, reduce unproductive expenditure
would generate more fiscal space for the health sector thereby states
would have been reduced the fiscal deficit and debt burden for the
medium-to-long term policy perspective [10,15,18]. Further, the im-
plementation of these fiscal space strategy at the state-level requires
political commitment and good governance [19,57].
To the best of author's knowledge, there are few empirical literature
examines the relationships between government health expenditure
and economic growth using panel unit root and panel cointegration
regression methods in Indian states [6,8,27]. But these studies have not

3
Mclntyre and Kutzin [44] has suggested the ‘rule of thumb’ for government
tax revenue to GDP ratio such as a very low level of fiscal capacity (< 15%);
low fiscal capacity (15%–20%); low to medium (20%–25%); medium
Fig. 2. Trends of Government health expenditure vis-a-vis tax revenue of (25%–35%); medium to high (35%–45%) and very high (> 45%).
emerging economies. Source: World Development Indicators [64], World Bank. 4
Please see Appendix Table A1 for the description of fiscal space indicators.

2
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Fig. 3. State-wise variation in the sources of health financing in India (average 1980–2014). Source: Author's representation.

examined the alternative revenue mobilization options and their long- enhance the healthcare financing. The government should emphasize
run impact on healthcare financing. Therefore, our study is contributing on the improvement of revenue collection and adding other sources of
in this way to examine the effects of fiscal space on government health revenue namely health specific earmark taxes, taxes from natural
expenditure in twenty-one states of India. Further, we observed in Fig. 3 sources etc. The remainder of the paper is structured as follows. Section
is that there is a huge variation in the generation of fiscal space among 2 discusses the literature review. Section 3 contains data and methods.
the states and examine their differential impact on the allocation of Section 4 analyzes the results and discussion. Section 5 contains the
health budget could be another contribution to the existing literature. conclusion and policy implications.
The application of both micro and macro panel-data econometrics, this
study examines the short-run as well as the long-run impact of fiscal
2. Literature review
space indicators (i.e. tax revenue, non-tax revenue, fiscal transfer, and
borrowings) on the prioritization of health expenditure that could be a
This study has divided the literature section into two parts. The first
significant contribution to the empirical literature. The empirical ex-
section reviews the cross-country studies that describes the fiscal space
amination of this study is divided into five parts. First, we test the
options in order to finance the health sector. The second section reviews
stationary property of variables using Levin et al. [35], Breitung [13],
the empirical literature that examines the relationships between fiscal
Im et al. [28], and Maddala and Wu [41] panel unit-root tests. Second,
space and health sector financing by exploring cross-country as well as
we examine the long-run relationship between the series using Pedroni
Individual country-level studies.
[49], Kao [30], and Maddala and Wu [41], panel cointegration tests.
Third, we examine the elasticity of government health expenditure with
respect to fiscal space (i.e. tax revenue, non-tax revenue, fiscal transfer, 2.1. Fiscal space strategy and health sector financing
and borrowings) using panel regression models such as Fixed Effects
(FE), and Fully Modified Ordinary Least Square (FMOLS). Fourth, we Table 1 discusses the sources of fiscal space for the health sector
examine the elasticity of government health expenditure with respect to financing in order to achieve universal health coverage by taking
fiscal space in different quantiles using panel non-additive Fixed Effects country-level studies namely Asian countries, Bhutan, Rwanda, Ghana,
Quantile Regression (FE-QR) methods. Fifth, we examined Granger Indonesia, South Africa, Nepal, Bangladesh, and India. These studies
Causality between the series using the Vector Error Correction Model are based on project report and annual report of Individual countries as
(VECM) approach. well as the report of the international organization such as the World
The overall result relating to fiscal space and its impact on gov- Bank and World Health Organization. Rao and Seth [51], Lane et al.
ernment health expenditure in Indian states implies that domestic [34], Vargas et al. [62], and Meheus and Mclntyre [45] have suggested
revenue mobilization through tax revenue and non-tax revenue would that the domestic revenue mobilization through improving existing
revenue collection and adding alternative tax revenue increases the

3
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Table 1
Country-level experiences in the generation of fiscal space for health.
Author Country Issues and objective Fiscal space options/initiatives

Rao and Seth [51] Bhutan Issue: resources gap in plan expenditure of about 4–5% of GDP due Additional resources mobilization options such as
to a high degree of volatility in the revenue.Objective: financing improvement in revenue collection, enhance revenue
MDGs and reprioritization of education and health sector. productivity by broadening the sales tax base, creating an
enabling environment for private investment- particularly
foreign investment.
Lane et al. [34] Rwanda Issue: More dependency on aid-financed fiscal space for health has Less dependency on foreign assistance and more emphasized
created a problem during the period of economic crisis.Objective: on domestic financing government spending to realize fiscal
fiscal requirement for achieving health-related MDGs. space.
Schieber et al. [55] Ghana Issue: Assesses the sustainability of the National Health Insurance Increase real growth of resources allocation depends on the
Schemes.Objective: creating additional fiscal space for health over ability of the government to significantly improve its
the next three to five years. revenue collection efforts, efficiency gains and improve the
absorptive capacity of existing resources.
Jha et al. [29] Asian countries Issue: High burden tobacco death counties of Asia such as China, Increasing cigarette prices through imposing higher tobacco
India, Philippines, Thailand and Viet Nam. taxes. It will pro-poor in their health benefits.
Objective: Increase fiscal of these high tobacco counties.
Basrin [5] Indonesia Issue: poor health financing systemObjective: evaluation of Expansion of tax base by improving tax administration
sustainability and effective allocation of health budget reform, minimizing tax evasion and generate the sources of
non-tax revenue through the exploitation of natural
resources.
McIntyre et al. [43] South Africa Issues: Higher OOP and poor health care services. Improvement in tax revenue collection through different tax
Objective: fiscal space for financing health care. policy and administrative reform.
Belay and Tandon Nepal Issue: several challenges in the health sector, especially inequalities, Efficiency gains in the health sector through the systematic
[11] malnutrition, and OOP health expenditure.Objective: expand design of existing grants and incentives. So that payment is
government expenditure on health and introduce social health directly linked to performance.
insurance to address the financial risk of ill health.
Vargas et al. [62] Bangladesh Issue: lowest levels of government health spending in the world, Generate additional domestic resources over the short to
inadequate service coverage and a high reliance on OOP. As a result, medium term from economic growth, improved tax
a large number of households fall into poverty each year due to collection and re-prioritization of the budget in the favor of
medical costs.Objective: Increase fiscal space for health. the health sector.
Meheus and Low-and middle- Issue: progress towards UHC and promote inclusive social and Increase revenue through improved tax compliance and
Mclntyre [45] income countries economic development.Objective: examine options for increasing efficiency in revenue collection, maximizing revenue from
domestic government revenue. mineral and other natural resources and increasing taxes
where appropriate.
Behera and Dash [7] Indian states Issue: huge inequality in government health expenditure and More importance to health care by allocating more resource
prioritization of the health budget in the majority of in the budget, emphasize on fiscal capacity by raising
states.Objective: role of state's domestic product in the assessment of revenue.
fiscal space for health

Note: UHC: Universal Health Coverage; OOP: Out-of-pocket; and MDGs: Millennium Development Goals.
Source: Author's representation.

fiscal space to their respective countries, thereby it facilitates health suggests that factors such as ethnolinguistic homogeneity, low level of
sector finance. Schieber et al. [55], Belay and Tandon [11] have em- corruption, democratization, and more women in public office are
phasized on efficiency gains and improving the absorption capacity of correlated with higher shares in government expenditure on health
health sector grants for the generation of fiscal space in the resource- [36,37,40]. However, these findings are not robust and sensitive to
poor economies. Jha et al. [29], Meheus and Mclntyre [45] have argued model specification. In order words, evidence from cross-country stu-
that the maximization of alternative revenues from minerals, natural dies suggests that country-specific public finance policies are key de-
resources, taxes on unhealthy products would possibly generate rev- terminants of government financing towards healthcare and thereby an
enues in an economy. Basrin [5] and Mclntyre et al. [43] have em- economy would expand the breadth, depth, and height of UHC
phasized the administrative reform in the fiscal policy that leads to [21,36,39,40,53].
reduce tax evasion and avoidance, thereby revenue collection will be Reeves et al. [53] examine the relationships between the total tax
improved. The overall objectives of these fiscal policy interventions revenue and government health expenditure in 89 low-income and
lead to reducing the financial burden that occurs through the higher middle-income countries for the period 1995 to 2011. They find that
OOP expenditure and provides financial leverage to the government in domestic tax revenue is a major determinant of progress towards UHC.
order to spend more on healthcare. It implies that indirect tax revenue is positively correlated with greater
investment in public health, access to services and better outcomes. Fan
2.2. Empirical literature and Savedoff [21] examine government revenue mobilization appears
to be a robust factor in explaining government health expenditure in
Although there are a number of empirical studies examine the de- 126 countries for the period 1995 to 2009. They show that at 1% point
terminants of government health expenditure and finds that economic increment in national income leads to 1.15% change in government
growth (per capita income) is the robust indicators to verify whether health expenditure. Liang and Mirelman [36] examine the impact of
health care is a luxury or necessity goods [24]. While the studies em- national income, debt and tax financing on government health ex-
phasis on the prioritization of health expenditure and its relationships penditure in 120 countries for the period 1995 to 2010. They find that
with the fiscal policies is relatively a new domain in health economics. there is a positive relationship between government health expenditure
Therefore, the improvement in public sources of financing towards and domestic borrowings that implies debt financing provides financial
healthcare and prioritization of health budget is the foremost policy leverage to expand government expenditure on health. Further, they
prescription for achieving UHC in low-income and middle-income find a negative relationship between the size of the tax base and the
countries [59]. To date empirical work on prioritization of health allocation of government expenditure on health. Lu et al. [40] sys-
budget has been sparse, available cross-country econometric analysis tematically analyze the sources of public financing of health in

4
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

developing countries and find that domestic revenue mobilization is an intertemporal budget constraint rule in order to understand the concept
important indicator of a government commitment to the health sector of fiscal space for health as suggested by Heller's [25] and Tandon and
financing. The result shows that development assistance towards the Cashin [58]. Eq. (1) illustrates the use of budgetary resources (ag-
health sector has a positive and significant impact of domestic gov- gregate expenditure) to be equal to the sources of budgetary resources
ernment expenditure on health, while debt relief has no effect on health (aggregate revenues). The left-hand side of Eq. (1) represents aggregate
expenditure. Lora and Olivera [39] assess the effects of domestic bor- expenditure (AE ) such as government non-interest expenditure (Gt ) and
rowings (debt), total revenue and fiscal balance on social sector ex- non-discretionary debt interest payment ( Bt ) , while the right-hand side
penditure (especially health and education) in 50 countries for the of Eq. (1) shows aggregate revenue such as tax revenue (Tt ) , borrowing
period 1985 to 2003. They find that the highest debt ratio reduces the (Bt ) , grants (At ) , and non-tax revenue (Ot ) . It was observed that after the
social sector expenditure and government health expenditure hits more generation of aggregate revenue (AR) , the prioritization of health
when debt increases. It also implies that tax revenue is positively and budget (GHEt ) depends on the proportion of (kt ) aggregate expenditure.
significantly affects government health expenditure.
Gt + t Bt = Tt + Bt + At + Ot , GHEt = kt (AEt ) (1)
Xing and Fuest [66], Abbott and Jones [1], Rodden and Wibbels
1

[54], Arena and Revilla [2], and Tornell and Lane [60] have argued In this study, our argument is twofold. First, does the kt of GHE is
that fiscal transfer shows a positive impact on the growth of govern- sufficiently distributed across the population of Indian states; second,
ment expenditure in sub-national level. They argue that central gov- does the aggregate revenue capacity of the state government influences
ernment try to influence the sub-national government spending via the the kt of GHE for the prioritization of health budget. Therefore, we
intergovernmental fiscal transfer system and willing to give more grants demonstrate the effects of fiscal space (aggregate revenue) on the
in the times of economic boom when tax revenue is high. The mobili- prioritization of government expenditure on healthcare in twenty-one
zation finance through the intergovernmental transfer is called ‘Vora- states India for the period of 1980–81 to 2014–2015.6
city effect.5 They argue that the increase intergovernmental fiscal
transfer to sub-national government is justified on several grounds. GHEit = f (TRit , NTRit , FTit , BORit ) (2)
First, the economic costs of raising state-level taxes tend to be higher Eq. (2) illustrates the mathematical functional form that shows GHE
than that of federal taxes; second, it can encourage states to invest in is the function of Tax Revenue (TR), Non-Tax Revenue (NTR), Fiscal
budget items that produce positive externalities, such as education and Transfer (FT), and Borrowings (BOR). The description of the variable
healthcare; third, it can be used to redistribute income from richer are presented in Table A1 (Appendix). We obtained state-wise data on
states to poorer ones, thereby contributing to the reduction of regional GHE, TR, NTR, FT, and BOR from the State Finance-Budget Report of
disparities. Reserve Bank of India [56]. The variables included in the Eq. (2) are
The major limitation of these studies is that they have examined the measured in per capita (divided by the state's population) constant
determinants of government health expenditure using panel Ordinary 2004-05 base year's prices.7 Table 2 presents the result of summary
Least Square (OLS) regression methods namely pooled OLS and FE statistics and pair-wise correlation of variables. The summary statistics
model, while the examination of long-run determinants of government describe a higher standard deviation value of GHE, TR, NTR, and FT
health expenditure is unnoticed in the earlier studies. Additionally, that indicates greater spread in the data from the mean value. It implies
there is a scarcity of literature examines the long-run elasticity of that there is a huge variability in the fiscal space across states. The
government health expenditure by controlling public finance policies correlations analysis indicates that GHE and Fiscal Space (i.e. TR, NTR,
and the majority of studies are considered only economic growth for FT, and BOR) are positively correlated with each other and the re-
their long-run econometric analysis. The studies namely Hitiris [26], lationships are highly significant at 1% level.
Narayan et al. [46], Baltagi and Francesco [4], Wang [63], and Behera
and Dash [6] have examined the long-run relationships between gov- ln(GHE )it = 1 + 2 ln(TR)it + 3 ln(NTR)it + 4 ln(FT )it + 5 ln(BOR)it
ernment health expenditure and economic growth using panel unit- + µit (3)
root, panel cointegration, and panel cointegrating regression methods.
They find that both government health expenditure and economic Eq. (3) shows the log-log panel OLS regression model with real per
growth are associated with the long-run and economic growth causing capita GHE as the dependent variable and real per capita TR, real per
healthcare in the long-run. There has no studied in the context of Indian capita NTR, real per capita FT, and real per capita BOR as our in-
states, which examine the impact of public finance policies on gov- dependent variables in the empirical estimation. The advantages of
ernment health expenditure except Rodden and Wibbels [54] and Be- using natural logarithmic in the empirical models as follows. First, it
hera and Dash [9]. Rodden and Wibbels [54] find that state's own tax reduces the skewness of data because regression can be influenced a lot
revenue and central transfer have shown a positive impact on the by outlier or leverage point of one or both variables. Second, the
growth of government health expenditure. Behera and Dash [8,9] find coefficient can be used to determine the impact of our independent
that per capita tax revenue shows a positive impact on the growth of variables (i.e. TR, NTR, FT, and BOR) on our dependent variable (GHE).
government health expenditure. Earlier studies such as Behera and In precisely, the coefficients in a log-log model represent the elasticity
Dash [7,10], and Hooda [27] have not emphasized the fiscal space of GHE with respect to TR, NTR, FT, and BOR and it is the estimated
concept in their studies. To, the best of author's knowledge, the impact percent change in our dependent variable for a percent change in our
of fiscal space on the government expenditure on healthcare has been independent variables, on average.
unnoticed in the earlier studies. Therefore, this study would provide a Fig. 4 presents the probability distribution of variables GHE, TR,
policy direction for financing healthcare in India as well as other low- NTR, FT, and BOR of Indian states using overlaying histogram, normal
income and middle-income countries. and kernel density plots. The red line presents the kernel density curve,

3. Data and methods 6


India has twenty-nine states. We consider twenty-one states due to data
availability for the stated period. The study period has been restricted till
In this study, we have adopted the algebra of a government's 2014–2015 due to the availability of actual estimates in the state-finance
budget reports and data for the latest years are not available for the sample
states.
5 7
Voracity effects suggest that grants and other resource transfer to the sub- The study has used the GDP deflator (Nominal GDP/Real GDP*100) of the
national government are a result of lobbying or even cronyism and corruption respective states for the conversion of nominal value to a real value in order to
[60]. encounter the effect of inflation.

5
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Table 2
Summary statistics and pairwise correlation.
Variables Definition Summary Statistics Pair-wise Correlation

Mean Std. Dev. GHE TR NTR FT BOR

GHE Government Health Expenditure (Per Capita) 325.117 308.269 1


TR State's Own Tax Revenue (Per Capita) 1638.130 2976.779 0.208* 1
NTR State's Own Non-Tax Revenue (Per Capita) 1033.618 3227.798 0.456* 0.362* 1
FT Fiscal Transfer (Per Capita) 3203.045 5982.691 0.674* 0.659* 0.541* 1
BOR State's Total Borrowings (Per Capita) 8061.438 5928.225 0.622* 0.395* 0.351* 0.481* 1

Note: Balanced panel over 1980–2014 for 21 Indian states. All variables are real constant prices at base 2004–05 in Indian Rupees (INR). * denote the significance at
1% level.
Source: Author's estimation from the State Finance Budget Report [56], Reserve Bank of India (RBI).

Fig. 4. Distribution of Variables. Source: Author's representation.

green line presents a normal distribution curve, and blue bars shows a and drought as well as changes in health technology and management
histogram of numerical data. Fig. 4 implies that GHE and FT are skewed which may affect the GHE. From the Eq. (4), we estimated panel FMOLS
right, while the distribution of TR and NTR are symmetric unimodal long-run cointegrated model proposed by Pedroni [48]. The precondi-
and BOR shows a symmetrical distribution. It implies that variables are tion for applying the FMOLS method is that our dependent variable
not distributed symmetrically from there mean over the period. (GHEit ) and independent variables (TRit , NTRit , FTit , BORit ) should be
cointegrated with slopes which may or may not be homogeneous
3.1. Empirical methods across the cross-sectional entities i .
Second, we have applied panel non-additive Fixed Effects-Quantile
The empirical procedure of this study is divided into three steps. Regression (FE-QR) estimation techniques in order to examine the
First, we have applied panel regression methods namely FE, and FMOLS elasticity of GHE with respect to TR, NTR, FT, and BOR by distributing
in order to examine the elasticity of GHE with respect to TR, NTR, FT, GHE in different quantiles. The QR techniques were originally proposed
and BOR. by Koenker and Bassett [31]. It provides the quantile distribution of the
dependent variables by given set of explanatory variables. In other
ln(GHE )it = 1 + 2 ln(TR)it + 3 ln(NTR)it + 4 ln(FT )it + 5 ln(BOR)it words, it estimates the linear relationships between regressors and a
+ i + t + µit (4) specified quantile of the dependent variable. Koenker and Bassett [31]
QR approach do not take into account the individual-specific un-
By following Eq. (3), we constructed the panel OLS-FE regression observed heterogeneity i and simply estimate the pooled Quantile
model in Eq. (4). In Eq. (4), the intercept 1 shows across- and within- Regression (QR). In order to capture the individual specific unobserved
state level variations, i stands a vector of state fixed effects which heterogeneity i , Koenkar [32] proposed panel QR methods with fixed
denotes time-invariant, t is time period dummy, µit is the residual error effects, which controls the state-specific unobserved distributional ef-
term, i = 1, …, N for each state in the panel and t = 1, …, T refers to fects of independent variables on dependent variables. But the Koenkar
the time period. So, we have controlled the state-specific heterogeneity [32] QR fixed effects model will be considered in the case of a large
as well as time-varying factors in the OLS-FE regression model for the number of cross-sectional individual observations over a relatively
robustness of our results. As noted by Baltagi [3], time-period fixed small number of time periods. The literature has been developed by
effects would control all year specific shocks such as war, epidemics,

6
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

introducing a penalty term in QR fixed effects model in order to likelihood ratio test; and µ is the serially uncorrelated error term. We
minimize the computational problem, that is called additive FE-QR use an instrumental variable estimator to eliminate the correlation
model [14,22,23,32]. We consider the following penalized version ad- between the error term and the lagged dependent variables in the dy-
ditive FE-QR estimators: namic panel data methods. The Optimum lag length was chosen at
K T N n k = 3 because through Vector Auto Regressive (VAR) lag length selec-
min Wk k (Yit i XitT ( k )) + | i| tion procedure. In the GHE Eq. (6a), short-run causality from TR, NTR,
( , )
k=1 t=1 i =1 i=1 (5) FT, and BOR to GHE are tested, respectively, based onH0: B12ik = 0 ik ,
H0: B13ik = 0 ik , H0: B14ik = 0 ik , and H0: B15ik = 0 ik . In the TR Eq. (6b),
In Eq. (5), i the index for states (N), T is the index for the number of
observations per states, K is the index for quantiles, X is the matrix of short-run causality from GHE, NTR, FT, and BOR to TR are tested, re-
spectively, based on H0: B21ik = 0 ik , H0: B23ik = 0 ik , H0: B24ik = 0 ik ,
explanatory variables (TRit , NTRit , FTit , BORit ) , k is the quantile loss
function. In addition Wk is the relative weight given to the Kth quantile, and H0: B25ik = 0 ik . In the NTR Eq. (6c), short-run causality from GHE,
TR, FT, and BOR to NTR are tested, respectively, based on
which controls for the contribution of the k th quantile on the esti-
H0: B31ik = 0 ik , H0: B32ik = 0 ik , H0: B34ik = 0 ik , and H0: B35ik = 0 ik . In
mation of the fixed effects.
the FT Eq. (6d), short-run causality from GHE, TR, NTR, and BOR to FT
In Eq. (5), the additive FE-QR model assumes that the distributional
characteristics of cross-sectional entities do not vary over time. As our are tested, respectively, based on H0: B41ik = 0 ik , H0: B42ik = 0 ik ,
H0: B43ik = 0 ik , and H0: B45ik = 0 ik . Finally, in the BOR Eq. (6e), short-
study period contains the longer time period, we estimated the fol-
lowing non-additive FE-QR approach as suggested by Powel [50]. run causality from GHE, TR, NTR, and FT to BOR are tested, respec-
tively, based on H0: B51ik = 0 ik , H0: B52ik = 0 ik , H0: B53ik = 0 ik , and
QYit ( k | i, t , Xit ) = i + t + XitT ( k ) (6) H0: B54ik = 0 ik . The null hypothesis of no long-run causality in each
Eqs. (6a)–(6e), is tested by examining the significance of the t-statistics
Powel [50] argues that the Koenkar [32] additive FE-QR model is
for the coefficient on the respective Error Correction Term (ECT ) re-
suffered two limitations. First, the additive FE-QR model estimates the
presented by .
distribution of (Yit i )|Xit instead of estimating the distribution of
Yit |Xit , which would not provide the evidence of heterogeneity. Because
observation at the top of the (Yit i ) distribution may be at the bottom
4. Empirical results and discussion
of the Yit distribution. Secondly, the additive FE-QR model does not
provide information about the effects of the policy variables on the
4.1. Panel unit root and cointegration test results
outcome distribution and only give the outcome relative to fixed effect
distribution. The inclusion of time period dummy t in the model
In time-series econometrics, a standard practice is to test the sta-
shows the effects of fiscal space on the outcome distribution of GHE.
tionary properties (the mean, variance, and covariance are constant
Third, we have employed a panel-based VECM approach to identify
over time) of data before applying any linear regression model, other-
the existence and direction of a long-run equilibrium relationship using
wise, it provides spurious regression and statistical inferences (i.e.‘t’
the two-step procedure of Engle and Granger [20]. First, we have es-
test) will not be appropriate [3]. Table 3 presents the results of the
timated the long-run model specified in Eq. (4) in order to generate the
panel unit root tests such as Levin et al. [35], Breitung [13], Im et al.
estimated residuals µit . In the second step, we estimated the dynamic
[28], and Maddala and Wu [41]. The results indicate that the null
Error Correction Model (ECM ) using lagged residuals.
hypothesis of the existence of unit root could not be rejected for all of
ln GHEit = +
q q
the variables at the selected level. However, the unit root null hy-
1j k = 1 B11ik ln GHEit k + k = 1 B12ik ln TRit k
q pothesis for all of the variables at the first difference could be com-
+ k = 1 B13ik ln NTRit k
pletely rejected at the 1% level.
q q
+ B
k = 1 14ik
ln FTit k + B
k = 1 15ik
ln BORit k + 1i ECTit 1 + µ1it The results of panel unit root tests indicate that the variables con-
(6a) tain a panel unit root in the first difference, then we can proceed to
examine whether there is a long-run relationship among the variables
q q
ln TRit = 2j + k = 1 B 21ik ln GHEit k + k = 1 B 22ik ln TRit k GHE, TR, NTR, FT, and BOR. Table 4 presents the results from the panel
q cointegration tests that includes Pedroni [49], Kao [30], and Fisher-
+ k = 1 B 23ik ln NTRit k
q q
+ B
k = 1 24ik
ln FTit k + B
k = 1 25ik
ln BORit k + 2i ECTit 1 + µ 2it Table 3
(6b) Result of panel unit root tests.
Source: Author's estimation.
q q
ln NTRit = 3j + k = 1 B31ik ln GHEit k + k = 1 B32ik ln TRit k Variable ln (GHE) ln (TR) ln (NTR) ln (FT) ln (BOR)
q
+ k = 1 B33ik ln NTRit k
Levels
q q
+ B ln FTit + B ln BORit + 3i ECTit 1 + µ3it LLC 1.939 3.619 2.617 2.879 3.814
k = 1 34ik k k = 1 35ik k
Breitung 7.091 4.393 2.175 5.033 1.659
(6c) IPS 5.396 5.507 −0.480 3.348 0.765
q q Fisher-ADF 14.283 14.165 51.286 24.574 39.716
ln FTit = 4j + k = 1 B41ik ln GHEit k + k = 1 B42ik ln TRit k Fisher-PP 11.087 16.628 147.281* 30.5888 20.045
q First difference
+ k = 1 B43ik ln NTRit k LLC −23.897* −18.392* −6.203* −15.622* −14.491*
q q Breitung −8.272* −8.595* −5.549* −1.566*** −7.129*
+ B
k = 1 44ik
ln FTit k + B
k = 1 45ik
ln BORit k + 4i ECTit 1 + µ4it
IPS −22.133* −19.854* −16.350* −17.779* −14.450*
(6d) Fisher-ADF 439.194* 362.813* 294.797* 344.810* 257.635*
Fisher-PP 1120.200* 709.025* 2234.470* 1122.780* 295.834*
q q
ln BORit = 5j + k = 1 B51ik ln GHEit k + k = 1 B52ik ln TRit k
q Note: LLC and IPS represent the panel unit root tests of Levin et al. [35] and Im
+ k = 1 B53ik ln NTRit k et al. [28] respectively. Fisher-ADF and Fisher-PP represent the Maddala and
q q
+ B
k = 1 54ik
ln FTit k + B
k = 1 55ik
ln BORit k + 5i ECTit 1 + µ5it Wu [41] Fisher-ADF and Fisher-PP panel unit root tests, respectively. The
exogenous variables represented in individual effects and linear trends. ln:
(6e)
Natural logarithms. *, ** and *** denote the significance at 1%, 5% and 10%
Where is the first-difference operator; k is the lag length based on level.

7
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Table 4 in BOR, we expect GHE to change (negative) by 0.19%. Therefore, the


Result of panel cointegration tests. overall result implies that the mobilization of tax revenue has been a
Source: Author's estimation. positive impact while borrowings has been a negative impact on the
Test statistics t-statistics allocation of government expenditure on healthcare in the long-run.
Padroni test In Table 5, we have shown the diagnostic test results of our em-
panel v-statistics −0.521 pirical models. It implies that our estimated regression equations are
panel rho-statistics 0.139
free from serial-serial correlation, no heteroscedasticity, no multi-
panel PP-statistics −2.052**
panel ADF-statistics −4.140* collinearity, no cross-section dependence, and no group-wise hetero-
group rho-statistics 2.361 scedasticity. So, our results follow the unbiasedness and consistency
group PP-statistics −1.058 properties of the classical linear regression model.
group ADF-statistics −4.902* In Table 6, we have examined the elasticity of GHE with respect to
Kao test −4.477*
fiscal space (i.e. TR, NTR, FT, and BOR) in individual states of India
Johansen Fisher test Hypothesized no. of CE None 422.7*
(s) using FMOLS regression methods. The result shows that the impact of
Test trace statistics At most 1 194.4* TR on GHE is positive and statistically significant at 1% level in the
At most 2 104.7* states namely Andhra Pradesh, Assam, Bihar, Punjab, Rajasthan, Tamil
At most 3 70.16*
Nadu, Uttar Pradesh, and Tripura. It implies that at 1% change in TR,
At most 4 61.55*
Johansen Fisher test Hypothesized no. of CE None 288.7* we expect GHE to change by less than 1%. While, the elasticity of GHE
(s) with respect to TR is more than 1% in states like Gujarat, Himachal
Test max-eigenvalue At most 1 122.6* Pradesh, and Tamil Nadu. The result shows that the impact of BOR on
statistics GHE is negative and statistically significant at 1% in states including
At most 2 70.58*
Assam, Gujarat, Himachal Pradesh, Madhya Pradesh, Odisha, Ra-
At most 3 60.91**
At most 4 61.55** jasthan, Tamil Nadu, Tripura, and Uttar Pradesh. It implies that at 1%
incremental change in domestic BOR, we expect GHE to reduce by less
Note: *, ** denote the significance at 1%, and 5% level. than 1% in the long-run.
The overall result implies that almost all states show a positive
type test using Johansen methodology [41]. The results show that a impact of domestic revenue mobilization on prioritization of GHE while
majority of the statistics significantly reject the null of no cointegration. borrowings show an adverse impact on GHE in the long-run. We ob-
Thus, it is evident that GHE and fiscal space (i.e. TR, NTR, FT, and BOR) served that there is a huge difference in the prioritization of GHE with
are moving together in the long-run. respect to fiscal space within states. First, North-Eastern states such as
Manipur, Meghalaya, Nagaland, Sikkim, and Tripura have been out-
4.2. Impact of fiscal space on government health expenditure sourcing their finance towards healthcare through central government
fiscal transfer (FT). Because these states are belong to special category
Using panel log-log regression (i.e. pooled-OLS and FE) methods, status in India and their 90% of expenses in developmental activities
Table 5 exhibits the elasticity of GHE with respect to fiscal space (i.e. are lifted by the central government finances [16]. Second, High po-
TR, NTR, FT, and BOR) in the short-run. We find that the impact of pulous states and low-income states such as Uttar Pradesh, Madhya
NTR, FT, and BOR are positive and statistically significant at 1% level to Pradesh, Bihar, and Rajasthan have been facing many developmental
GHE. It implies that at 1% change in NTR, we expect GHE to change by problems such as poverty and unemployment, thereby there develop-
0.05% keeping other variables constant. Similarly, at 1% change (po- mental expenditure has grown up over the period [8,10,27]. High
sitive) in FT, and BOR, we expect GHE to change by 0.38%, and 0.46% revenue expenditure over the revenue generation capacity of those
respectively. states has increased the debt burden, resulting the debt services burden
In order to check, whether the short-run relationships between GHE deteriorates the developmental expenditure in the long-run. Third,
and fiscal space holds in the long-run or not, for which we applied the highly agricultural developed states namely Haryana, Karnataka, and
long-run cointegrating regression model i.e. FMOLS. The result shows Punjab have shown a negative relationships between GHE and NTR. It
that the impact of TR, NTR, and FT are positive and statistically sig- implies that incremental change in NTR over the period have negative
nificant at 1% level to GHE. It implies that at 1% change in TR, we impact of GHE. There are many factors associated with the generation
expect GHE to change by 0.35% in the long-run. One interesting result, of fiscal space especially tax policy and administration reform, thereby
we have been found in the relationships between GHE and BOR, which it would finance healthcare for achieving UHC (see Table 1). But the
shows negative and statistically significant. It implies that at 1% change

Table 5
Result of Panel Regression. Dependent Variable: ln (GHE) Per Capita.
Source: Author's estimation.
Tests Explanatory Variables (ln Per Capita)

ln (TR) ln (NTR) ln (FT) ln (BOR)

Ordinary Least Square (OLS) 0.121 (0.076) 0.128* (0.031) 0.458* (0.037) 0.036 (0.036)
Fixed Effects (FE) −0.0627 (0.136) 0.0515* (0.028) 0.383* (0.116) 0.460* (0.121)
Fully Modified OLS (FMOLS) 0.353* (0.080) 0.070***(0.039) 0.276* (0.070) −0.190* (0.053)

Diagnostic Test Null Hypothesis Chi-Square Statistics

Breusch-Pagan LM test Homoscedasticity 147.523


VIF test (Mean) No Multicollinearity 2.62
Serial-correlation test No serial-correlation 57.679
Cross-section Correlation test Cross-section independence 2037.895
Modified Wald test No group-wise heteroscedasticity 345.74

Note: Standard error in parenthesis; *, ** and *** denotes the level of significance at 1%, 5% and 10%.

8
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Table 6
Result of FMOLS Regression in Individual States. Dependent Variable: ln (GHE) Per Capita.
Source: Author's estimation.
Sample States Explanatory Variables (ln Per Capita)

ln (TR) ln (NTR) ln (FT) ln (BOR)

Andhra Pradesh 0.698* (0.213) 0.071 (0.199) 0.072 (0.162) −0.194 (0.119)
Assam 0.454** (0.186) 0.156 (0.119) 0.235 (0.178) −0.851* (0.243)
Bihar 0.686* (0.163) 0.006 (0.050) −0.129 (0.157) −0.098 (0.158)
Gujarat 1.174* (0.165) −0.024 (0.110) 0.199* (0.113) −0.518* (0.108)
Haryana −0.335 (0.513) −0.315* (0.107) 0.811* (0.214) 0.219 (0.412)
Himachal Pradesh 1.167* (0.196) 0.128** (0.054) −0.949* (0.233) −0.271* (0.097)
Karnataka 0.182 (0.156) −0.330* (0.049) 0.564* (0.128) −0.057 (0.0920
Kerala 0.057 (0.196) 0.423* (0.053) 0.066 (0.131) 0.193 (0.125)
Madhya Pradesh 0.676 (0.444) 0.032 (0.162) 0.349 (0.324) −0.784* (0.231)
Maharashtra 0.547 (0.511) −0.269 (0.210) 0.361 (0.240) −0.300 (0.269)
Manipur 0.280 (0.189) −0.319* (0.117) 0.774** (0.370) −0.185 (0.177)
Meghalaya −1.350* (0.355) 0.179 (0.209) 1.512* (0.337) 0.514* (0.194)
Nagaland −0.480 (0.522) 0.438 (0.412) 0.701 (0.821) −0.289 (0.493)
Odisha 0.152 (0.307) 0.012 (0.172) 0.460 (0.297) −0.337** (0.141)
Punjab 0.651* (0.149) −0.154* (0.035) −0.166* (0.099) 0.106 (0.072)
Rajasthan 1.665* (0.398) 0.170 (0.164) −0.870* (0.225) −0.637* (0.238)
Sikkim −0.354* (0.081) −0.044*** (0.024) 1.178* (0.138) 0.259* (0.077)
Tamil Nadu 1.218*(0.214) 0.126 (0.112) −0.049 (0.147) −0.631* (0.130)
Tripura 0.749*(0.054) −0.046 (0.036) 0.270* (0.073) −0.469* (0.037)
Uttar Pradesh 0.226** (0.111) 0.127** (0.049) 0.583* (0.109) −0.285* (0.048)
West Bengal 0.012 (0.160) 0.013 (0.049) 0.304** (0.134) 0.140** (0.056)

Note: Standard error in parenthesis; *, ** and *** denotes the level of significance at 1%, 5% and 10%.

determinants of fiscal space is not our scope of analysis and only we can the MCMC optimization procedure and control all time variant un-
imply that domestic revenue mobilization would be the potential observed effects. So, the overall panel QR method shows the impact of
sources of health system financing in the long-run. fiscal space on the growth of GHE by different quantile distribution
Our finding is similar to Xing and Fuest [66], Abbott and Jones [1], namely 5th, 25th, 50th, 75th, and 95th.
Rodden and Wibbels [54], Arena and Revilla [2], and Tornell and Lane [60] Fig. 5 shows the normal quantile-quantile plot of GHE, TR, NTR, FT,
who finds that central government fiscal transfer is the major sources of and BOR. Fig. 5 illustrates that in both higher and lower quantiles of
government expenditure in the sub-national level. Our finding suggests that states show a non-normality because many states fall either below or
domestic tax revenue has no impact on the growth of GHE for the short-run above the normal quantile distribution. We have seen from the Fig. 5
using panel FE model. On the contrary earlier studies, Reeves et al. [53] and that there are many outliers in the data which may affect the regression
Lora and Olivera [39] find a positive impact on GHE using same panel FE estimation. So, the application simple linear regression model has not
model. They find that tax revenue is strongly positively associated with captured the distributional impact of fiscal space on GHE. Therefore, we
greater investment in public healthcare in low-income and middle-income applied the panel QR method to examine the distributional hetero-
countries, while we found that mobilization of tax revenue has positive geneity in the allocation of GHE in Indian states. There are several
spillover effects the GHE in the long-run. We found that domestic borrow- advantages to applying the QR method. First, the QR method provides a
ings is one of the major sources of finance for fiscal space for health in the convenient approach to show a regression model that gives different
short-run. Our findings support the argument of Liang and Mirelman [36] response results of regressors when we divide the GHE into different
who argues that the positive coefficient of government debt (borrowings) distribution [38]. Second, the conditional distribution of GHE shows
provide financial leverage to expand government expenditure on healthcare the asymmetric relationship with the level of fiscal space across states
in low-income and middle-income counties in the short-run. While we find [17]. Third, the QR method is more robust in response to large outliers
the long-run negative impact of borrowings on GHE which has supported by in the data [63].
Lora and Olivera [39]. Lora and Olivera [39] argue that higher borrowings Table 7 shows a comprehensive picture of the effect of the pre-
ratio reduces social sector expenditures as well as health expenditure while dictors (i.e. TR, NTR, FT, and BOR) on the response variable (GHE) by
it hits badly to the growth of GHE in the long-run. using quantile regression. The interpretation of the quantile regression
parameter is mostly similar to linear regression estimation. The quantile
regression parameter estimates the change in a specified quantile (i.e.
4.3. Distributional impact of fiscal space on government health expenditure 5th, 25th, 50th, 75th, and 95th) of GHE produced by a 1% change in the
fiscal space variable (i.e. TR, NTR, FT, and BOR). This allows us to
To facilitate comparisons with panel regression, we have applied the compare how some quantiles of the GHE may be more affected by
panel Fixed Effects-Quantile Regression (FE-QR) in order to examine certain fiscal space characteristics than other quantiles.
the impact of fiscal space on GHE by distributing GHE in different Table 7 presents the result of the non-additive FE-QR model, which
quantiles. The Pooled-QR method of Koenkar8 [32] provides simple shows that the effects of FT on GHE is positive and statistically significant at
pooled quantile regression approach without controlling unobserved 1% level across quantiles. It implies that at 1% incremental change in FT,
distributional heterogeneity and time period effect in the model, we expect 0.28%–0.55% change in GHE across quantiles. In precisely the
whereby the estimation could be biased due to large time series data. In elasticity of FT with respect to GHE is large in the higher quantile as 75th
order to get the robust result, we applied the non-additive FE-QR model and 95th quantiles (mostly upper-middle and high-income states). On the
of Powel [50] that can eliminate all endogeneity problem by adopting contrary, the elasticity of FT with respect to GHE is less in 5th quantiles
(especially poor-income states). The result implies that the central govern-
8
ment fiscal transfer is the most important source of fiscal space for health in
The result of pooled Quantile regression is not reported here and can be
Indian states irrespective of their level of development.
available upon request.

9
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Fig. 5. Normal Quantile-Quantile Plots. Source: Author's representation.

Table 7
Result of Panel Fixed Effects Quantile Regression. Dependent Variable: ln (GHE) Per Capita.
Source: Author's estimation.
Tests Explanatory Variables (ln Per Capita)

ln (TR) ln (NTR) ln (FT) ln (BOR)

th
5 quantile −0.561 (0.125) 0.105* (0.020) 0.279* (0.086) 0.323* (0.075)
25th quantile 0.082** (0.034) −0.072* (0.014) 0.494* (0.018) 0.360* (0.023)
50th quantile −0.276* (0.009) 0.094* (0.003) 0.255* (0.004) 0.410* (0.015)
75th quantile −0.128** (0.056) 0.052* (0.012) 0.624* (0.030) 0.226* (0.082)
95th quantile 0.030 (0.023) 0.186* (0.006) 0.552* (0.024) −0.106* (0.024)

Note: Non-additive Fixed Effects-Quantile Regression (FE-QR) model computed with adaptive MCMC optimization using independent variables as instrument
individual. It captures both individual unobserved heterogeneity and time effects by quantile function of GHE. Standard error in parenthesis; *, ** and *** denotes
the level of significance at 1%, 5% and 10%.

In the FE-QR model, we find that TR is adversely affected the growth of domestic borrowings adversely affects the growth of per capita GHE at 95th
GHE at 5th, 50th, and 75th quantiles, while it has positively affected to the quantiles (mostly high-income states).
growth of GHE at 25th quantile with less significant level as 0.08%. Further, From the overall analysis of different sources of fiscal space in order
we find that there is no significant relationship between TR and GHE at to financing GHE, we observed that states with the low and middle level
95th quantile. In the similar fashion, we find that the effects of NTR are of tax revenue growth (especially 5th, 50th and 75th quantiles), the
positively affected to the growth of GHE at 5th, 50th, 75th, and 95th mobilization of healthcare expenditure growth derived from central
quantiles, while it has adversely affected to the growth of GHE at 25th government transfer and borrowings. There is a paucity of literature
quantile. By analyzing the effects of TR and NTR on GHE, we observed that that examines simultaneously the impact of revenue, central transfer,
the negative or insignificant impact of per capita TR on per capita GHE is and borrowings on the growth of GHE using quantile regression ap-
being compensated by the positive effects of per capita NTR and per capita proach. Very less amount of literature examines the effects of health
Transfer at 5th, 50th, 75th and 95th quantiles. Borrowing (BOR) is another expenditure on economic growth using quantile regression namely
source for the generation of fiscal space but it not a sustainable option for Chen et al. [17], Wang [63], and Linden and Ray [38]. Some of our
health finance in the long-run as compared to the state's own domestic findings are similar to Chen et al. [17], Wang [63], and Linden and Ray
revenue mobilization. In this study, we find that the borrowings show a [38], who argue that lower GHE growth states (especially lower
positive impact on the growth of GHE in all most all quantiles except 95th quantiles) the distribution is skewed to the right, which means less
quantile. It implies that domestic borrowings is one of the potential sources healthcare to be consumed than the high GHE growth states (especially
of finance in order to spend on healthcare in the majority of Indian states higher quantiles).
due to lower revenue mobilization of the government. On the contrary,

10
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Table 8
Result of Panel VECM Granger Causality tests.
Source: Author's estimation.
Dependent variable(s) Direction of causality

Short-run Long-run

∑ ΔlnGHEit-1 ∑ ΔlnTRit-1 ∑ ΔlnNTRit-1 ∑ ΔlnFTit-1 ∑ ΔlnBORit-1 ECTit-1

ΔlnGHEit … 0.065* 0.001** −0.162** 0.129** −0.186*


[9.788] [6.179] [6.941] [6.607] [-11.255]
(0.007) (0.045) (0.031) (0.036) (0.000)
ΔlnTRit 0.107a … −0.003 0.082 0.119 0.012
[2.159]b [2.203] [1.783] [1.388] [1.276]
(0.539)c (0.531) (0.618) (0.708) (0.202)
ΔlnNTRit −0.090 0.499 … 0.156 0.200 −2.63E-05
[2.929] [8.011] [1.051] [2.100] [-0.031]
(0.402) (0.045)** (0.788) (0.551) (0.975)
ΔlnFTit 0.124 0.383 0.087 … 0.161 0.015
[2.895] [16.285] [6.476] [2.627] [1.186]
(0.408) (0.001)* (0.090)*** (0.452) (0.235)
ΔlnBORit 0.008 0.091 −2.00E-03 −0.123 … −0.002
[6.126] [4.040] [9.358] [18.106] [-3.474]
(0.105) (0.257) (0.024)** (0.000)* (0.000)*

Note: a, b, and c denote the sum of coefficients, statistics value, and p-value, respectively; Chi-square statistics testing for short-run causality through the joint
significance of Wald test, and t-statistics testing for long-run causality through the error-correction adjustment coefficient; *, ** and *** denote the significance at the
1%, 5% and 10% level.

4.4. Granger causality between fiscal space and government health paper is divided into three parts. First, we applied panel regression
expenditure methods such as fixed effects and fully modified ordinary least square in
order to examine the elasticity of government expenditure on health-
Table 8 reports the results of the short-run and long-run Granger care with respect to fiscal space indicators (i.e. tax revenue, non-tax
causality tests using panel Vector Error Correction Model (VECM). With revenue, fiscal transfer, and borrowings). Second, we applied panel
respect to Eq. (6a), sources of fiscal space from TR, NTR, and BOR have quantile regression in order to examine the distributional impact of
a positive and statistically significant impact in the short-run on gov- fiscal space on health expenditure across states. Third, we employed a
ernment health expenditure. An examination of the sum of the lagged panel-based error correction model to identify the existence and di-
coefficients on the respective variables indicates that BOR (0.129%) has rection of a short and long-run causal relationship among variables.
a greater impact on GHE than either TR (0.065%) or NTR (0.001%) Our panel regression result implies that the mobilization of tax
which reiterates the importance domestic borrowings in the health fi- revenue has been a positive impact while borrowings has been a ne-
nance process. Moreover, the Error Correction Term (ECT) is statisti- gative impact on the allocation of government expenditure on health-
cally significant at the 1% level and exhibits the expected negative sign. care in the long-run. The result shows that at 1% change in tax revenue,
It implies that 0.186% of short-run disequilibrium originated from the we expect government health expenditure to change (increase) by
past shocks adjusted in the current period, and speed of adjustment 0.35% in the long-run. On the contrary, at 1% change in domestic debt,
towards long-run equilibrium is relatively slower. In terms of Eq. (6b), we expect government health expenditure to change (negative) by
it appears that GHE, NTR, FT, and BOR do not have a statistically sig- 0.19%. The panel quantile regression result shows that tax revenue is
nificant impact on Tax revenue. However, the ECT is statically insig- adversely affected the growth of government expenditure on healthcare
nificant which suggests that tax revenue is not responsive to adjust- at 5th, 50th, and 75th quantiles in the short-run while domestic debt
ments towards long-run equilibrium. In terms of Eq. (6c), it appears shows a positive impact the growth of government health expenditure
that TR has a positive and statistically significant impact on non-tax in all most all quantiles except 95th quantile. From the overall analysis
revenue, while the impact of GHE, BOR, and FT are statistically insig- of different sources of fiscal space, we observed that states with the low
nificant. However, the ECT is statically insignificant which suggests that and middle level of tax revenue growth (especially 5th, 50th and 75th
non-tax revenue is not responsive to adjustments towards long-run quantiles) have been mobilizing finance towards healthcare through
equilibrium. In regards to Eq. (6d), both TR and NTR have a positive central government transfer and borrowings in the short-run. Our panel
and statistically significant impact on the fiscal transfer in the short-run Granger Causality analysis shows that sources of fiscal space from tax
while the impact of GHE and BOR are statistically insignificant. The revenue, non-tax revenue, and borrowings have a positive and statis-
ECT is statistically insignificant, which implies that fiscal transfer is not tically significant impact on government health expenditure in the
responsive to adjustments towards long-run equilibrium. Finally in the short-run. An examination of the sum of the lagged coefficients on the
regards to Eq. (6e), both NTR and FT have a negative and statistically respective variables indicates that domestic borrowings (0.129%) has a
significant impact on the reduction of domestic borrowings while the greater impact on government health expenditure than either tax rev-
impact of GHE, and TR are statistically insignificant. However, the ECT enue (0.065%) or non-tax revenue (0.001%) which reiterates the im-
is statistically significant again with a relatively slow speed of adjust- portance borrowings in the healthcare finance process in the short-run.
ment towards a long-run equilibrium. Moreover, the error correction term is statistically significant at the 1%
level and implies that 0.186% of short-run disequilibrium originated
from the past shocks adjusted in the current period, and speed of ad-
5. Concluding remarks and policy implications justment towards long-run equilibrium is relatively slower.
The overall analysis concludes that less domestic revenue mobili-
In this paper, we examine the effects of fiscal space on the prior- zation and higher dependency of borrowings towards healthcare fi-
itization of government expenditure on healthcare in twenty-one states nancing may create fiscal stress on state finances in the long-run and
India for the period of 1980–2014. The empirical procedure of this

11
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

thereby it could possibly reduce the prioritization of spending. This has been limited and therefore, this study would be helpful for de-
study has suggested some policy recommendation. First, Indian states signing effective public finance policies for strengthening health system
should emphasis on the improvement of revenue collection, increase of an economy.
the tax base and adding alternative revenue i.e. earmark taxes and sin-
taxes in order to broaden the fiscal space. Second, the government Declaration of conflicting interests
should spend central government transfer efficiently and develop an
expenditure management framework in order to prioritize the health The author(s) declared no potential conflicts of interest with respect
sector. Although this study makes a preliminary attempt at under- to the research, authorship, and/or publication of this article.
standing the effects of fiscal space on government health expenditure in
an emerging country like India. This study could extend to examine the Funding
impact of fiscal space on health outcomes in the perspective of universal
health coverage. Further to date, empirical work on the prioritization of This research did not receive any specific grant from funding
government expenditure on healthcare from a fiscal space perspective agencies in the public, commercial, or not-for-profit sectors.

Appendix

Table A1
Descriptions of variables.

Variable Definition

Government Health Expenditu- Medical and public health expenditure, family welfare in both revenue and capital account of the state government.
re (GHE)
Tax Revenue (TR) State's own tax revenue includes direct taxes and indirect taxes.
Direct taxes: Agricultural income tax and profession tax, taxes on land revenue, stamps and registration fees, surcharge on some cash crops and
urban immovable property taxes.
Indirect taxes: Sales tax, central sales tax, surcharge on sales tax etc.
Non-Tax Revenue (NTR) State's own non-tax revenue includes interest receipts, dividends & profits and income from lotteries
Fiscal Transfer (FT) Intergovernmental fiscal transfer includes share in central taxes and central grants.
Share in Central Taxes1: A certain percentage of central taxes revenue is assigned to states under law enactment by Parliament under Article 270 of
the Constitution of India. The share of states in the gross tax revenue as decided by the finance commission of India which is changed the time to
time.
Grants from the Centre: Central assistance in order to mobilize funds for the social programs for the uplift of the poor as well as developmental
activities by the union government. Union government also extends grants to states at the time of natural calamities.
Borrowings (BOR) Total outstanding liabilities of the state government. It includes total internal debt, loans from banks and financial institutions, loans and advances
from centre, provident funds, reserve funds and contingency funds.

Source: State Finance Budget Report [56], Reserve Bank of India (RBI).
1
The Share of tax devolution to states has made to correct vertical and horizontal imbalances among the Indian states. The finance commissions set up under
Article 280 of the Constitution every fifth year to address these imbalances. The recent 14th finance commission has recommended that the share of tax devolution to
states has been increased at 10% points from 32% to 42% of the divisible pool to increase the flow of unconditional transfer to states, and thus provide then with
larger fiscal space [15].

Appendix A. Supplementary data

Supplementary data to this article can be found online at https://doi.org/10.1016/j.seps.2018.11.004.

References Ho J, Rusciano F, Humphreys G. Health in 2015: from MDGs Millennium devel-


opment goals to SDGs sustainable development goals.
[13] Breitung J. The local power of some unit root tests for panel data. Nonstationary
[1] Abbott A, Jones P. Intergovernmental transfers and procyclical public spending. panels, panel cointegration, and dynamic panels. Emerald Group Publishing
Econ Lett 2012;115(3):447–51. Limited; 2001. p. 161–77.
[2] Arena M, Revilla JE. Pro-cyclical fiscal policy in Brazil: evidence from the states. [14] Canay IA. A simple approach to quantile regression for panel data. Econom J
The World Bank; 2009. p. 5144. 2011;14(3):368–86.
[3] Baltagi B. Econometric analysis of panel data. John Wiley & Sons; 2008. [15] Chakraborty P, Dash BB. Fiscal reforms, fiscal rule, and development spending: how
[4] Baltagi BH, Moscone F. Health care expenditure and income in the OECD recon- indian states have performed? Public Budg Finance 2017;37(4):111–33.
sidered: evidence from panel data. Econ Modell 2010;27(4):804–11. [16] Chakraborty P, Mukherjee AN, Amarnath HK. Macro policy reform and sub-national
[5] Basrin OH. Fiscal space in Indonesia: an evaluation of sustainability and effective finance: why is the Fiscal space of the states shrinking? Econ Polit Wkly
allocation. Asian J Public Aff 2013;6(1):19–35. 2009:38–44.
[6] Behera D, Dash U. Nexus between public health expenditure and income: empirical [17] Chen MY, Lin FL, Chang CK. Relations between health care expenditure and in-
evidence from Indian states. J Appl Econ Sci 2016;11(6):1067–77. come: an application of local quantile regressions. Appl Econ Lett
[7] Behera DK, Dash U. Effects of economic growth towards government health fi- 2009;16(2):177–81.
nancing of Indian states: an assessment from a fiscal space perspective. J Asian [18] Dholakia RH, Govil A. Fiscal space with states and the sixth pay commission. Econ
Public Pol 2017:1–22. Polit Wkly 2008:16–20.
[8] Behera DK, Dash U. Examining the state level heterogeneity of public health ex- [19] Duran A, Kutzin J, Menabde N. Universal coverage challenges require health system
penditure in India: an empirical evidence from panel data. Int J Healthc Technol approaches; the case of India. Health Pol 2014;114(2–3):269–77.
Manag 2018;17(1):75–95. [20] Engle RF, Granger CW. Co-integration and error correction: representation, esti-
[9] Behera DK, Dash U. Impact of GDP and tax revenue on health care financing: an mation, and testing. Econometrica: J Econometr Soc 1987:251–76.
empirical investigation from Indian states. Theor Appl Econ 2017;2(611):249–66. [21] Fan VY, Savedoff WD. The health financing transition: a conceptual framework and
[10] Behera DK, Dash U. The impact of macroeconomic policies on the growth of public empirical evidence. Soc Sci Med 2014;105:112–21.
health expenditure: an empirical assessment from the Indian states. Cogent Econ [22] Galvao Jr. AF. Quantile regression for dynamic panel data with fixed effects. J
Finance 2018;6(1):1–21. Econom 2011;164(1):142–57.
[11] Belay T, Tandon A. Assessing fiscal space for health in Nepal. Heath, nutrition and [23] Harding M, Lamarche C. A quantile regression approach for estimating panel data
population discussion paper. Washington, DC: World Bank Group; 2015. models using instrumental variables. Econ Lett 2009;104(3):133–5.
[12] Boerma T, Mathers C, AbouZahr C, Chatterji S, Hogan D, Stevens G, Mahanani WR, [24] Hartwig J, Sturm JE. Robust determinants of health care expenditure growth. Appl

12
D.K. Behera and U. Dash Socio-Economic Planning Sciences 68 (2019) 100667

Econ 2014;46(36):4455–74. Nonstationary panels, panel cointegration, and dynamic panels. Emerald Group
[25] Heller PS. The prospects of creating ‘fiscal space’ for the health sector. Health Pol Publishing Limited; 2001.
Plann 2006;21(2):75–9. [49] Pedroni P. Panel cointegration: asymptotic and finite sample properties of pooled
[26] Hitiris T. Health care expenditure and integration in the countries of the European time series tests with an application to the PPP hypothesis. Econom Theor
Union. Appl Econ 1997;29(1):1–6. 2004;20(3):597–625.
[27] Hooda SK. Government spending on health in India: some hopes and fears of policy [50] Powell D. Quantile regression with non-additive fixed effects. Quantile Treatment
changes. J Health Manag 2015;17(4):458–86. Effects. 2016.
[28] Im KS, Pesaran MH, Shin Y. Testing for unit roots in heterogeneous panels. J [51] Rao MG, Seth A. Fiscal space for achieving the millennium development goals and
Econom 2003;115(1):53–74. implementing the Tenth Plan in Bhutan. Econ Polit Wkly 2009:51–9.
[29] Jha P, Joseph R, Li D, Gauvreau C, Anderson I, Moster P, Bonu S. Tobacco taxes: a [52] Rao MG. Political economy of government finance in India. India Rev
win-win measure for fiscal space and health. Asian Development Bank; 2012. 2015;14(1):58–72.
[30] Kao C. Spurious regression and residual-based tests for cointegration in panel data. [53] Reeves A, Gourtsoyannis Y, Basu S, McCoy D, McKee M, Stuckler D. Financing
J Econom 1999;90(1):1–44. universal health coverage—effects of alternative tax structures on public health
[31] Koenker R, Bassett Jr. G. Regression quantiles. Econometrica: J Econometr Soc systems: cross-national modelling in 89 low-income and middle-income countries.
1978:33–50. Lancet 2015;386(9990):274–80.
[32] Koenker R. Quantile regression for longitudinal data. J Multivariate Anal [54] Rodden J, Wibbels E. Fiscal decentralization and the business cycle: an empirical
2004;91(1):74–89. study of seven federations. Econ Polit 2010;22(1):37–67.
[33] Kurian O. Financing healthcare for all in India: towards a common goal: oxfam [55] Schieber G, Cashin C, Saleh K, Lavado R. Assessing the prospects for fiscal space for
India. New Delhi: Copernicus Marg; 2015. health in Ghana. 2012https://doi.org/10.1596/9780821395660_CH04.
[34] Lane C. Rwanda: fiscal space for health and the MDGs revisited. The World Bank; [56] State finances study of budgets. Reserve Bank of India; 2015. [RBI].
2009. [57] Sundararaman T, Mukhopadhyay I, Muraleedharan VR. No respite for public
[35] Levin A, Lin CF, Chu CS. Unit root tests in panel data: asymptotic and finite-sample health. Econ Polit Wkly 2016;51(16):39–42.
properties. J Econom 2002;108(1):1–24. [58] Tandon A, Cashin C. Assessing public expenditure on health from a fiscal space
[36] Liang LL, Mirelman AJ. Why do some countries spend more for health? An as- perspective. NW Washington, DC 20433: The World Bank; 2010.
sessment of sociopolitical determinants and international aid for government health [59] Tandon A, Fleisher L, Li R, Yap WA. Reprioritizing government spending on health:
expenditures. Soc Sci Med 2014;114:161–8. pushing an elephant up the stairs? WHO South-East Asia J Public Health
[37] Lie GS. Does government health spending increase when women decide? (Master's 2014;3(3):206.
thesis) 2010. [60] Tornell A, Lane PR. The voracity effect. Am Econ Rev 1999;89(1):22–46.
[38] Linden M, Ray D. Aggregation bias-correcting approach to the health–income re- [62] Vargas V, Begum T, Ahmed S, Smith O. Fiscal space for health in Bangladesh: to-
lationship: life expectancy and GDP per capita in 148 countries, 1970–2010. Econ wards universal health coverage. Washington, DC: World Bank; 2016.
Modell 2017;61:126–36. [63] Wang KM. Health care expenditure and economic growth: quantile panel-type
[39] Lora E, Olivera M. Public debt and social expenditure: friends or foes? Emerg Mark analysis. Econ Modell 2011;28(4):1536–49.
Rev 2007;8(4):299–310. [64] World Development Indicators. The World Bank group 2016; Washington DC:
[40] Lu C, Schneider MT, Gubbins P, Leach-Kemon K, Jamison D, Murray CJ. Public United States.
financing of health in developing countries: a cross-national systematic analysis. [65] World Health Report. Health systems financing: the path to universal coverage.
Lancet 2010;375(9723):1375–87. Geneva: World Health Organization; 2010.
[41] Maddala GS, Wu S. A comparative study of unit root tests with panel data and a new [66] Xing J, Fuest C. Central-local government fiscal relations and cyclicality of public
simple test. Oxf Bull Econ Stat 1999;61(1):631–52. spending: Evidence from China. Int Tax Public Financ 2018;25(4):946–80.
[42] Mathauer I, Carrin G. The role of institutional design and organizational practice for
health financing performance and universal coverage. Health Pol Deepak Kumar Behera is a Senior Research Fellow of Economics in the Department of
2011;99(3):183–92. Humanities and Social Sciences, Indian Institute of Technology Madras, Chennai. His
[43] McIntyre D, Doherty J, Ataguba J. Universal health coverage assessment: South primary research interests are in the area of Health economics and public finance. He
Africa. Global Network for Health Equity; 2014. served as a reviewer for many internationally acclaimed journals.
[44] McIntyre D, Kutzin J. World Health Organization. Health financing country diag-
nostic: a foundation for national strategy development. 2016.
[45] Meheus F, McIntyre D. Fiscal space for domestic funding of health and other social Umakant Dash is a Professor of Economics in the Department of Humanities and Social
services. Health Econ Pol Law 2017;12(2):159–77. Sciences, Indian Institute of Technology Madras, Chennai. His primary research interests
[46] Narayan S, Narayan PK, Mishra S. Investigating the relationship between health and are in the economics of health care and macroeconomics. His research projects pursue
questions of equity, efficiency and risk protection in Indian Health System and are funded
economic growth: empirical evidence from a panel of 5 Asian countries. J Asian
Econ 2010;21(4):404–11. by DFID, Rockefeller Foundation and World Bank. He is also involved in collaborative
[47] National Health Account. Global health expenditure data base. World Health research (through RESYST, CREHS, GHLC) with other partners from countries like South
Organization; 2015. Africa, Thailand, Bangladesh, Ethiopia, Nigeria, Tanzania, Kenya and UK. He is also
[48] Pedroni P. Fully modified OLS for heterogeneous cointegrated panels. member of various committees and task force at the national level.

13

You might also like