Contribution of Infrastructure to Output Growth in India

Abstract

This article examines the effect of infrastructure on output performance of India taking into consideration different types of infrastructure facilities and also creating indices of infrastructure using principal component analysis. The methodology involves time series techniques where Granger causality is also tested to determine the nature of relationship between infrastructure and output using the Toda-Yamamoto two-step procedure for non-stationary variables. The results indicate that for India there exist unidirectional relationship from infrastructure to output and the impact of the same was then determined. Upon considering the impact of growth rate of various infrastructure indicators, it is found that electricity and telecommunication growth rates have had significant and positive impact on output growth. The coefficient for electricity generation is 0.35 and of tele-density growth rate is 0.15. These results are significant considering that to ensure and enhance the growth potential of Indian economy, care must be taken to remove any bottlenecks in the provisioning of these infrastructure variables and to further improve upon the delivery of these infrastructure sectors.

Keywords

Infrastructure index usage of infrastructure output growth

Introduction

The rationale behind using infrastructure as an argument for economy-wide growth stems from infrastructure’s direct impact—acting as intermediate input to production, enhancing private sector productivity, and being complementary to private capital formation—to its indirect effects like reducing adjustment cost, helping increase durability of private capital, increasing competitiveness of a region, impacting demand, and supply of health and education services. A wide debate on the influence of infrastructure on output levels and growth has led to attempts to quantify this effect and estimate the contribution of infrastructure as a factor of production to aggregate output.

However, it must also be mentioned here that infrastructure development alone cannot determine economic growth. The various processes underlying economic performance stem from several theories ranging from neoclassical model stressing on exogenous technical progress (Solow, 1956); endogenous growth theories focusing on new knowledge (Romer, 1990), and innovation (Aghion & Howitt, 1992); new economic geography models asserting the role played by location of economic activity, agglomeration, and specialization and recently, institutional economics which underlines the role played by non-economic factors like institutional factors (North, 1990), socio-cultural factors (Knach & Keefer, 1997), political determinants (Brunetti, 1997), and demography (Brander & Dowrick, 1994). Hence, infrastructure development is only one of the factors that can help determine growth and other factors like investment climate, institutions etc. are also equally important.

In recent years India has come to enjoy the reputation of a fast growing developing nation. India’s economy was growing at more than 9 percent before the global financial crisis hit the world economy. The recovery from the impact of the crisis was also swift and sharp and the economy achieved a growth rate of 8.6 percent in 2009–2010 despite a severe drought. But there has been a decline in growth rates since then with the recent growth rate for 2013–2014 being reported as 4.7 and 6.6 percent according to the old and new estimates, respectively (Government of India, 2015). The coming decade is crucial for India if it aims to move from a low income to a middle income country and the required rate of growth for this to happen is estimated to be around 8–9 percent per annum (Rakesh & Kapur, 2014). In order also to deal with the various socioeconomic problems that malign the Indian economy, a strong growth rate will provide leverage to be able to invest in schemes that will help provide employment, education, and better health facilities to its vast population and in helping broaden the scope of social safety nets. Thus, in order to ensure that nothing hinders this growth process or its furtherance it is important to identify the major bottlenecks to growth and the channels through which they may operate.

With supply-side bottlenecks being described as the main reason for slowdown in the Indian economy, it is necessary for instituting schemes like “Make in India” which imply producing for India, and for the World, in India, but to have this idea be translated into reality, India needs a much better and stronger network of infrastructure facilities along with development of good human capital. There have been several new policies and initiatives taken by the government in this direction especially post 2005—Bharat Nirman launched in 2005–2006 for building rural roads, telephony, electrification etc., sector-specific measures undertaken to increase infrastructure spending and minimize the impact of global recession on domestic industry, recasting Industrial Infrastructure Upgradation Scheme to create quality infrastructure, increase in budget allocation for infrastructure expenditure especially in National Highway Development Program and Urban Infrastructure development initiatives. Thus there are attempts being made to build more infrastructure and ease the supply side bottlenecks that come in the form of infrastructure deficiencies. In this vein, it is necessary to find what has been the impact of infrastructure growth on output growth and what has been infrastructure’s contribution in India’s economic growth trajectory under new sets of policies.

Infrastructure’s relationship with economic growth varies across countries, over time as well as within countries. Initial empirical studies on macroeconomic impact of infrastructure engendered in the 1980s as a result of the initial failure to account for the productivity slowdown in developed nations particularly the United States of America. In India as well, there were few early studies investigating into availability of infrastructural facilities and their role in development such as those by Shah (1970), Prakash (1977), Gulati (1977), Rao (1977), Tewari (1983, 1984), Elhance and Lakshamana (1988), and Arun Kumar (1993) etc. But the results from these studies have been mixed and their economic significance doubted upon as the results are biased in the absence of usage of appropriate estimation methodology. It has been argued that in some of the studies the contributions are overstated and importance of other factors of production is ignored. The results are also not found to be robust to the use of more sophisticated econometric techniques and this fact was highlighted by Holtz-Eakin (1995) and Gramlich (1994). Issues like stationarity properties of the data (common trends in capital and output) and the reverse causality aspect are ignored in many of these studies. Thus, use of appropriate econometric techniques is imperative when using time series data.

In much of the literature, especially aggregate level studies, public capital has been considered synonymously for public infrastructure with the effects of public investment or estimates of public capital often assumed to be the effects of infrastructure on growth or output levels (Aschauer, 1989; Demetriades & Mamuneas, 2000; Duggal, Saltzman, & Klein, 1999; Lall, 1999 etc.). However, this approach is problematic because public capital and public investment entails more than what is considered to be infrastructure (e.g., Public administration and defence). This study is not an attempt to answer all these criticism, but it is an effort to add to the literature examining the relationship of infrastructure availability and national output specifically in the context of India. This study differs in sample, time period analyzed, estimation technique, and in explanatory variables considered as growth determinants.

With this perspective in mind, to study the effect of infrastructure on output performance of India, different types of infrastructure facilities have been selected and indices of infrastructure have been created using principal component analysis (PCA) which takes into account the collinearity among infrastructure variables. These have then been utilized to examine the impact on net domestic product (NDP) at factor cost of India. The organization of this article is as follows. Section 2 provides the methodology applied for the purpose of estimating the relationship between infrastructure and NDP. This involves the latest time series methodology of testing for non-stationarity of data and co-integration. Granger causality is tested to determine the nature of relationship between infrastructure and output using the Toda-Yamamoto two-step procedure for non-stationary variables. Section 3 consists of the empirical regression results for growth equations for aggregate output. The article ends with a few words in the manner of conclusion.

Econometric Methodology

Present study intends to examine the effect infrastructure has on output growth in the Indian context. For this purpose, certain infrastructure variables were selected as indicators of infrastructure development1

We have excluded agriculture-related infrastructure variables and focused on more general infrastructure indicators. There has been considerable amount of work that has dealt with agriculture infrastructure and output relationship.

. For making an index, the one main requirement is to get appropriate weights for each of the variable under each category. The conventional methods on index construction give ad hoc and fixed weights to various facilities but this approach has a limitation in that these weights may actually vary over time and space. To overcome this limitation, use has been made of PCA which follows from the multivariate technique of factor analysis. PCA is a way of identifying patterns in data, and expressing the data in such a way as to highlight their similarities and differences. Once these patterns are found, PCA allows compressing the data by reducing the number of dimensions without much loss in information. The time period of analysis is from 1970–1971 to 2011–2012. The indicators from various infrastructure sectors have been selected and their growth rates calculated.

First, the literature examining the direction of causality between infrastructure and output has not provided a concrete answer. There are short- and long-run dynamics involved and we attempt to provide evidence in the context of Indian economy on the direction of causality between infrastructure and output. For this purpose, we use indices of infrastructure in order not to lose too many degrees of freedom, as well as considered few infrastructure sectors individually. After standardizing the infrastructure variables, indices of infrastructure that have been used are

Composite Index for Infrastructure (II): includes all the infrastructure variables of interest.

Index for Usage of Infrastructure (IUI): includes variables that are indicators of usage of the physical infrastructure that exists—electricity consumption, number of registered motor vehicles density, railways traffic (includes both passenger and freight traffic), tele-density, credit–deposit ratio of all scheduled commercial banks, port traffic (freight), and airport traffic (both passenger and freight).

We begin by following a three-step procedure as suggested by Basu et al. (2003). At first, it should be tested that whether the data are stationary or not. Time series data tend to exhibit time trend and can therefore be non-stationary, that is, the variables have means and variances and covariances that are not time invariant. In order to then test for the relationship between infrastructure and output growth and to find if there is bidirectional or unidirectional causality between infrastructure and output variables Granger causality tests are performed. It has to be emphasized here that when testing for Granger causality using Wald test on the parameters of a Vector Auto-regression (VAR) model, if data are non-stationarity, then Wald test statistic do not follow asymptotic chi-square distribution under the null hypothesis. Hence we follow the Toda-Yamamoto two-step procedure for testing causality.

Next, Engle and Granger (1987) argued that the direct application of Ordinary Least Square (OLS) or Generalised Least Square (GLS) to non-stationary data can produce regressions that have problems like misspecification or are spurious in nature. These regressions tend to present results like high R2 values and t-statistics which then lead to Type I errors. However, a combination of non-stationary variables can, in certain cases, result in co-integration and an appropriate relationship. Hence, upon establishing the existence of unit root in time series, as a second step co-integration is tested for a long-run co-integrated relationship between two variables.

However, if the series are found to be non-stationary in levels and not co-integrated, then OLS regression can be estimated by first differencing the series and removing unit root from the series.

Test for Unit Root and Co-integration

For the time series modeling analysis, the properties of each individual series are studied to assess the presence of unit roots and keeping in line with the standard practice in time series first difference the series if unit roots are found. If the series are found to be stationary in levels they are considered integrated of order zero, I(0) or if they are found to be non-stationary and have a unit root and are therefore integrated of order d, I(d) where d is the number of times the variable has to be first differenced to make it stationary. The conventional Granger causality test is based on VAR methodology and cannot be applied to non-stationary variables as the stability condition of VAR is not met rendering the test statistic of Granger causality invalid. If the series are found to contain unit root and are integrated of the same order, this is followed by a Johansen trace test for testing the presence of co-integration between the series. This is followed by vector error correction model to test for relationship between non-stationary variables.

To test for the presence of unit root, in this study use is made of the Augmented Dickey Fuller (ADF) test (lag length selected based on Akaike information criterion (AIC)) and Phillips–Perron (PP) (1988) test (lag length based on Newey West test). When the variables are found to be non-stationary in levels the same tests are undertaken in first differences. If the variables are found to be stationary in first differences then we move on to the co-integration techniques. Table 1 presents the results of the unit-root tests. It can be seen that all series when in logarithm fail to reject the null of no unit root for both ADF and PP test but do so when taken in first difference. This implies that the order of the series is I(1) and this is true even for the infrastructure indices and the output series (per capita net state domestic product). The first difference of the logarithm of these variables is the growth rate and thus the tests confirm that the growth rates of infrastructure, output and labor and GFCF (gross-fixed capital formation) are stationary in nature and OLS estimation can be conducted.

Table 1

Tests for Unit Root

Variable	ADF		Phillips–Perron		Order of Integration
	Level	Δ	Level	Δ
LnElecCons	–2.51(0.31)	–4.47(0.00)***	–1.67(0.74)	–4.55(0.00)***	I(1)
LnElecGen	–0.86(0.95)	–5.73(0.00)***	–1.04(0.92)	–5.77(0.00)***	I(1)
LnElecGenCap	–3.00(0.14)	–3.77(0.00)***	–1.16(0.90)	–3.71(0.00)***	I(1)
LnRoad	–2.25(0.44)	–6.13(0.00)***	–1.20(0.66)	–6.20(0.00)***	I(1)
LnSurfRoad	–2.84(0.19)	–6.15(0.00)***	–2.91(0.16)	–6.15(0.00)***	I(1)
LnVehicle	–0.78(0.95)	–5.14(0.00)***	–1.25(0.88)	–5.24(0.00)***	I(1)
LnRail	–2.92(0.16)	–8.89(0.00)***	–2.92(0.16)	–9.16(0.00)***	I(1)
LnElecRail	–0.39(0.98)	–4.65(0.00)***	–0.79(0.95)	–4.69(0.00)***	I(1)
LnRailFreight	–1.45(0.82)	–4.45(0.00)***	–0.95(0.93)	–4.22(0.00)***	I(1)
LnPassRail	–1.24(0.88)	–5.70(0.00)***	–1.35(0.85)	–5.68(0.00)***	I(1)
LnTele	3.44(1.00)	–5.74(0.00)***	4.06(1.00)	–5.83(0.00)***	I(1)
LnBankBranch	–2.15(0.50)	–2.73(0.07)*	–3.68(0.03)	–2.73(0.07)***	I(1)
LnCDratio	–0.62(0.97)	–5.53(0.00)***	–0.67(0.98)	–5.52(0.00)***	I(1)
LnSchooling	–1.63(0.75)	–3.00(0.14)*	–1.68(0.73)	–3.83(0.02)***	I(1)
LnLifeExp	0.51(0.99)	–3.38(0.01)***	–4.18(0.00)***	–1.52(0.80)	I(1)
LnIMR	0.77(0.99)	–8.45(0.00)***	1.14(0.99)	–8.45(0.00)***	I(1)
LnPassAir	–2.02(0.27)	–5.71(0.00)***	–2.37(0.38)	–5.71(0.00)***	I(1)
LnFreightAir	–2.47(0.33)	–3.34(0.02)***	–1.64(0.75)	–5.56(0.00)***	I(1)
LnTraffPort	–2.23(0.45)	–5.70(0.00)***	–2.26(0.44)	–5.74(0.00)***	I(1)
LnGFCF	–0.51(0.97)	–6.03(0.00)***	–0.35(0.98)	–6.08(0.00)***	I(1)
LnII (PC1)	0.53(0.99)	–5.93(0.00)***	0.53(0.99)	–5.95(0.00)***	I(1)
LnIUI (PC1)	0.51(0.99)	–2.9(0.05)**	0.54(0.99)	–5.16(0.00)***	I(1)
LnLabor	–1.62(0.76)	–4.08(0.01)***	–1.46(0.82)	–4.19(0.01)***	I(1)
LnPCNDP	–0.82(0.95)	–5.46(0.00)***	–0.67(0.96)	–5.56(0.00)***	I(1)
LnPCCAP	–0.26(0.98)	–5.73(0.00)***	–0.10(0.99)	–5.83(0.00)***	I(1)

Source: Authors’ calculations.

Notes: Unit-root tests conducted using Stata 12; Ln = natural logarithm; Δ = first difference; I(m) = series is integrated of order m. ***, **, and * denote level of significance at 1, 5, or 10 percent, respectively.

ElecCons = per capita consumption of electricity; ElecGen = per capita generation of electricity; ElecGenCap = per capita electricity generating capacity; road = road density; SurfRoad = density of surfaced road; Vehicle = vehicle density; rail = railways density; elecRail = electrified rail route density; RailFreight = per capita freight carried by railways; PassRail = per capita number of passengers carried by railways; Tele = tele-density; BankBanch = number of scheduled commercial bank branches per one lakh population; CDratio = credit–deposit ratio; Schooling = average years of schooling obtained; LifeExp = life expectancy at birth (in years); IMR = infant mortality rate; PassAir = passengers carried by Airways; FreightAir = Cargo/freight carried by airlines (domestic and international); TraffPort = port traffic; GFCF = gross-fixed capital formation; II = index of infrastructure; IUI = index of usage of infrastructure; Labor = labor in India; PCNDP = per capita Net Domestic Product; PCCAP = per capita gross-fixed capital formation.

Once it is found from the unit-root tests that the variables are non-stationary in levels form then the next step is to apply co-integration analysis to examine whether a long-run co-integration relationship exists among those variables. Here, the co-integration tests used is based on multi-variate Johansen approach (1991) which uses two statistic tests—Trace test and Max-Eigen value. The likelihood ratio test is based on the trace statistics (λ trace) which tests the H₀: r ≤ q against H₁: q = r is calculated thus: λ_trace(r) = –T∑ln(1 − λi) where λ_r+i + …. λ_n, are the least value of eigenvectors (p − r). The second test is the maximal eigenvalue test (λ_max) which tests the H₀: there are r co-integrating vectors against the H₁: there are r + 1 co-integrating vectors and is calculated as follows: λ_max(r, r + 1) = –Tln(1 – λ_r+1). Results of the Johansen co-integration tests are displayed in Table 2.

Table 2 presents the result for co-integration tests between output and indices of infrastructure. The results based on both trace test statistics and maximum eigenvalue statistics do not point toward the existence of co-integrating equation between PCNDP and the three infrastructure indices (the same is also true for sectoral outputs). In this scenario, Granger causality tests were undertaken for the non-stationary variables using the Toda-Yamamoto procedure.

Granger Causality

Here we have empirically tested, for a given Y (output) and X (infrastructure index), Y (output) is said to be Granger caused by X (infra) if X (infra) helps in the prediction of Y (output), that is, if lagged values of X (infra) are statistically significant.

Δ y_{t} = a + b_{1} Δ y_{t - 1} + \dots + b_{1} Δ y_{t - 1} + c_{1} Δ {(infr a)}_{t - 1} + \dots + c_{1} Δ {(infra)}_{t - 1}

Δ (infra) t = a + b_{1} Δ {(infra)}_{t - 1} + \dots + b_{1} Δ {(infra)}_{t - 1} + c_{1} Δ y_{t - 1} + \dots + c_{1} Δ y_{t - 1}

As the different co-integration tests showed no evidence of a long-run relationship for different infrastructure indices and economic output, an error correction model-based causality test is not appropriate (Toda & Phillips, 1993). Thus, upon determining the maximum order of integration for the group of variables to be tested to be 1 (all variables are I(1) in nature), we set up a VAR model in levels of the data and determine the appropriate maximum lag length (p) for the variables in VAR using AIC and Schwarz information criterion which in this case was 1. VAR thus constructed was tested for autocorrelation. With this VAR in place, we add 1 (the maximum order of integration) additional lag to the equation. Upon doing this step, we test for the hypothesis that the coefficients of first p lagged values are zero using the standard Wald test. Then we do the same thing for the coefficients of the lagged values of the other variables entering the VAR. If we follow this procedure, then the Wald test statistics will be asymptotically chi-square distributed. Rejection of the null hypothesis implies rejection of Granger non-causality, that is, causality exists.

Table 2.

Tests for Co-integration: Output and Infrastructure Index

	No. of Co-integrating Equations	Trace Statistics	Max Eigenvalue Stat	Outcome
Between LnPCNDP and LnII	None	11.3(0.36)	11.0(0.31)	Both tests indicate NO co-integrating eqn
Between LnPCNDP and LnII	At most 1	0.29(0.58)	0.29 (0.58)	Both tests indicate NO co-integrating eqn
Between LnPCNDP and LnIUI	Zero	9.37(0.54)	9.07(17.1)	Both tests indicate NO co-integrating eqn
Between LnPCNDP and LnIUI	At most 1	0.30(0.58)	0.30(0.58)	Both tests indicate NO co-integrating eqn
Between LnPCNDP, LnII, LnGFCF, and LnLaborw	Zero	47.5(0.20)	25.9(0.17)	Both tests indicate NO co-integrating eqn
Between LnPCNDP, LnII, LnGFCF, and LnLaborw	At most 1	21.6(0.60)	12.8(0.69)	Both tests indicate NO co-integrating eqn

Source: Authors’ calculations.

Note: Figures in parentheses are the probability values.

The reported F-statistic is the Wald statistic which tests the joint significance of hypothesis: b₁=…= b_i = 0 are presented in Table 3. In all cases m = 1, that is, the series are integrated of Order 1. Table 3 presents causality results between infrastructure variables—electricity consumption, road density, rail density, tele-density as well as with infrastructure index separately with aggregate output. The results indicate that for the Indian economy, there has been a unidirectional impact from infrastructure on output. This is important considering it helps determine the direction of relationship and appropriate econometric methodology can be applied to measure the impact of infrastructure in India.

Upon establishing the direction of relationship, in order to see the impact of infrastructure growth on output growth, OLS regression was estimated. As a result, impact of growth of infrastructure on output growth was estimated and the results are presented in the next section.

Empirical Results

In this section, we begin by presenting the results from growth regressions wherein the dependent variable which are in first differences are stationary in nature and represent growth figures. The growth figures for infrastructure variables were found to be uncorrelated thus multi-collinearity will not be a concern.

Estimating the impact of selected infrastructure indicators’ growth rate on output growth rate is indicative toward the impact that growth in infrastructure has had on India’s growth rate. In this case, use has been made of three indicators of physical infrastructure sector—electricity generation (per capita)2

At all-India level it makes more sense to include electricity generation as infrastructure indicator for power infrastructure sector as it gives better indication toward the total electricity that is generated in the country and the state of power infrastructure.

, surfaced road density, and tele-density—most directly productive activities use electricity, telecommunication, and transport services as intermediate inputs (World Bank, 1994)—In Table 4, column 1 includes only the infrastructure indicators as explanatory variables, whereas column 2 includes labor, capital, and infrastructure indicators. This exercise is repeated with infrastructure indices—II and IUI (columns 3 and 4). For robustness check, we have also estimated these indices using fixed weight methodology, assigning equal weights to each indicator and the results for the same are also included (columns 5 and 6).

Table 3.

Results of Granger Causality Tests using Toda–Yamamoto (1995) Procedure

Causalty Running from	Causality to	Lag Length	Test Statistic and Significance
LnII	LnPCNDP	2	3.38***(uni)
LnIUI	LnPCNDP	2	3.49***(uni)
LnElec	LnPCNDP	2	2.81**(bi)
LnRoad	LnPCNDP	2	1.15 (no)
LnTele	LnPCNDP	2	6.52*** (uni)
LnPCCAP	LnPCNDP	1	2.8*(uni)
LnLabor	LnPCNDP	1	6.7***(uni)

Source: Authors’ calculations.

Notes: Data used for the period from 1970–1971 to 2011–2012. The causality test results as presented are for causality running from explanatory variables toward output. Cases where bidirectional causality was also found have been mentioned. *, **, and *** indicate the 10, 5, and 1 percent level of significance.

From columns 1 and 2, it is observed that electricity and telecommunication growth rate have a significant and positive impact on output growth. This holds true even after including capital and labor in the regression estimate. The coefficient for electricity generation is 0.35 and is significant at 5 percent level. And that of tele-density growth rate is 0.15 and is significant at 1 percent level. The coefficient on tele-density growth rate could be small because our time period starts from 1971 to 1972 and in India telecommunication growth rate picked up in the late 1990s and it is only from then that telecommunication became important infrastructure indicator. Even after including labor and capital, electricity growth rate and tele-density growth rate are found to have a significant and positive impact on output growth rate in the country.

Impact of growth rate of surfaced road density on output growth rate was not significant. This does not mean that road infrastructure is not an important infrastructure. Where electricity and telecom services are used directly in the process of production, the same cannot be said for road or transport infrastructure. The manner in which building more roads can impact output growth is more indirect. There are positive externalities from the same which it is possible get reflected in terms of capital productivity and labor productivity. This will have to be explored further by looking at the impact of roads on productivity directly.

Further, we observe a negative and significant relationship between output growth and employment growth in all the cases. The long-term trend of a decline in the rate of employment growth has been a reality in India. However, this decline has been accompanied by an acceleration in the rate of economic growth. Looking at the data, when India’s output grew at 4.7 percent between 1972–1973 and 1983, employment growth was at 2.4 percent; during 1983 to 1993–1994, employment growth was 2 percent whereas output growth was 5 percent; between 1993–1994 and 2004–2005, employment growth further declined to 1.8 percent and output growth accelerate to 6.3 percent and finally, during 2004–2005 to 2009–2010, output growth was at 9 percent, whereas employment growth was just 0.22 percent (Papola & Sahu, 2012).

Additionally, the impact of other capital on output growth has been positive and significant, with 1 percent increase on capital, output growth increased by roughly 0.15 percent for all the cases.

Looking at the results where infrastructure indices are used as explanatory variables, we observe a positive and significant contribution of infrastructure growth on output growth. Looking at column 3, it can be seen that with an increase of 1 percent in infrastructure index that includes all the indicators of infrastructure, output was found to increase by 0.23 percent (the corresponding figure for fixed weight index was much higher at 0.41 percent). The impact on infrastructure index that was constructed including only the usage variables was found to be smaller at 0.12 percent (corresponding figure for IUI fixed weight was 0.29). Hence we find that if labor and capital were kept constant and infrastructure index was increased by 1 percent, India’s output growth would increase by 0.36 percent. This is a substantial amount and justifies investing in infrastructure by the Indian government in its endeavor to increase output growth.

Table 4.

Estimation Results for Growth Rate of Output and Infrastructure Variables and Infrastructure Indices: 1971–1972 to 2011–2012

	GrPCNDP(1)	GrPCNDP(2)	GrPCNDP(3)	GrPCNDP(4)	GrPCNDP(5)	GrPCNDP(6)
Constant	–1.30(–0.8)	2.18(0.75)	7.67(1.96)***	7.9(4.03)***	6.59(3.2)***	6.49(3.15)***
GrElecGen	0.35(1.9)**	0.31(1.77)*
GrTele	0.15(4.69)***	0.09(2.09)***
GrSurfRoad	0.08(0.64)	0.09(0.78)
GrCapital		0.15(1.9)**	0.16(1.85)**	0.16(1.88)**	0.15(1.99)**	0.14(1.78)**
GrLabor		–1.3(–1.46)***	–2.11(–2.8)***	–2.20(–3.03)***	–2.2(–3.12)***	–1.87(–2.54)***
GrII(PCA)			0.23(1.9)**
GrIUI(PCA)				0.12(1.8)**
GrII(FW)					0.41(1.7)*
GrIUI(FW)						0.29(1.75)*
R square	0.40	0.47	0.37	0.4	0.41	0.41
F Stat	7.92	6.21	7.44	7.38	8.46	8.5
DW Stat	2.67	2.73	2.62	2.6	2.7	2.7
JarqueBera test	1.50[0.47]	0.53[0.76]	7.26[0.03]	7.07[0.03]	7.38[0.02]	9.4[0.10]
Heterosked test	0.62[0.54]	2.04[0.16]	0.25[0.61]	0.23[0.63]	0.37[0.54]	0.36[0.54]
n	41	41	41	41	41	41

Source: Authors’ calculations.

Note: Figures inside “()” are the t-statistics and figures inside “[ ]” are p values.

Conclusion

The discussion on the contribution of infrastructure to output growth has been ongoing for several decades. There are theoretical arguments both in favor and against the short and long term impact of infrastructure on economic growth and hence it remained as an important empirical question. It is in this context that this article engages in estimating the relationship between infrastructure and output growth specifically for India. It tried to address the various concerns raised by researchers—such as using stock of infrastructure instead of investment in infrastructure as explanatory variables. This is because in developing countries like India, investment figures do not necessarily translate into actual creation of infrastructure. Additionally, care was taken to use appropriate econometric methodology when dealing with time series data and various tests were performed to ensure the validity of results. Considering that the data show unidirectional causality running from infrastructure to output, the impact of infrastructure on output was then measured.

Since the data were time series in nature, it was important to test for unit root in the data otherwise the estimation results would be spurious in nature. The order of integration of all infrastructure indicators, various infrastructure indices, and other variables of interest was determined and presence of unit root confirmed. It was found that all the series were I(1). Upon establishing lack of co-integration between output and infrastructure indices in a bivariate framework as well as after including variables like capital and labor use was made of VAR analysis for determining direction of causality and the nature of relationship between output growth and infrastructure index and individual infrastructure variables.

Following this, a simple OLS regression was estimated for output and individual infrastructure indicators—road, electricity, and telecommunication sector as well as for infrastructure indices. This was done in first differences in order to take care of unit roots and multicollinearity issue. Upon considering the impact of growth rate of various infrastructure indicators as independent variables (all series are I(0) in growth rate), it is found that electricity and telecommunication growth rate has had significant and positive impact on output growth. The coefficient for electricity generation is 0.35 and is significant at 5 percent level. And that of tele-density growth rate is 0.15 and is significant at 1 percent level.

In India’s development path the role and significance of infrastructure sectors is thus found to have played a major role in enhancing economic growth. Sectors like telecommunication have played a major role in influencing growth of the services sector and, similarly, electricity has been found as a major player in determining growth. Thus, constraints in smooth supply of these infrastructure services will result in putting strain on India’s economic growth and care should thus be taken to ensure a seamless delivery of these services. This will not just help in maintaining the health of the economy but also further enhance the growth potential.

Footnotes

Author’s Biography

Dr. Sumedha Bajar is a Post-Doctoral Associate at National Institute of Advanced Studies, Indian Institute of Science, Bangalore. She obtained her PhD from Institute for Social and Economic Change, Bengaluru in 2016. She has worked as Assistant Director at National Institute of Labour Economics Research and Development, under NITI AAyog (formerly, Planning Commission) and interned at several places like United Nations Conference on Trade and Development (UNCTAD) and National Council of Applied Economic Research (NCAER), Delhi. Her principal research interest lies in the field of infrastructure sector and its impact upon the Indian economy.

Dr. Meenakshi Rajeev is the Reserve Bank of India Chair Professor in the Institute for Social and Economic Change, Bangalore, India. She graduated from IIT Kanpur and received a Phd degree from Indian Statistical Institute, Kolkata. She has published on a variety of research topics from both theoretical as well as empirical perspectives in national and international journals. Her recent publications include her book titled “Emerging issues in economic development” for Oxford University Press.

References

Aghion

Howitt

(1992). A model of growth through creative destruction. Econometrica , 60(2), 323–351.

Arun

Kumar A.V.

Upendranadh

(1993). Infrastructure development in India: An alternative approach to measurement. Productivity , 34(3).

Aschauer

(1989). Is public expenditure productive? Journal of Monetary Economics, 23(2), 177–200.

Basu

Chakraborty

Reagle

(2003). Liberalization, FDI, and growth in developing countries: A panel cointegration approach. Economic Inquiry , 41(3), 510–516.

Brander

Dowrick

(1994). The role of fertility and population in economic growth: Empirical results from aggregate cross-national data. Journal of Population Economics , 7(1), 1–25.

Brunetti

(1997). Political variables in cross-country growth analysis. Journal of Economic Surveys , 11(2), 163–190.

Demetriades

P.O.

Mamuneas

T.P.

(2000). Intertemporal output and employment effects of public infrastructure capital: Evidence from 12 OECD economies. Economic Journal , 110(465), 687–712.

Duggal

V.G.

Saltzman

Klein

L.R.

(1999). Infrastructure and productivity: A nonlinear approach. Journal of Econometrics , 92(1), 47–74.

Elhance

A.P.

Lakshamana

T.R.

(1988). Infrastructure-production system dynamics in national and regional systems: An economic study of the Indian economy. Regional Science and Urban Economies , 18(2), 511–531.

10.

Engle

R.F.

Granger

C.W.J.

(1987). Cointegration and error correction: Representation, estimation, and testing. Econometrica , 55(2), 251–276.

11.

Government of India. (2015). National Accounts Statistics. Central Statistical Organisation, Ministry of Statistical and Programme Implementation, Delhi.

12.

Gramlich

Edward M.

(1994). Infrastructure investment: A review essay. Journal of Economic Literature , 32(3), 1176–1196.

13.

Gulati

S.C.

(1977). Dimensions of inter-district disparities. Indian Journal of Regional Science , 9(2), 196–206.

14.

Holtz-Eakin

Schwartz

A.E.

(1995). Infrastructure in a structural model of economic growth. Regional Science and Urban Economics , 25(2), 131–151.

15.

Johansen

Søren

(1991). Estimation and hypothesis testing of cointegration vectors in Gaussian Vector Autoregressive Models. Econometrica , 59(6), 1551–1580.

16.

Knack

Keefer

(1997). Does social capital have an economic payoff? A cross-country investigation. The Quarterly journal of economics , 112(4), 1251–1288.

17.

Lall

S.V.

(1999). Role of public infrastructure investments in regional development: Experience of Indian states. Economic and Political Weekly, 34(12), 717–725.

18.

North

(1990). Institutions, institutional change and economic performance. Cambridge: Cambridge University Press.

19.

Papola

T.S.

Sahu

P.P.

(2012). Growth and structure of employment in India: Long-term and post reform performance and the emerging challenge; A study prepared as part of research programme structural changes, industry and employment in the Indian economy. Indian Council of Social Science Research.

20.

Phillips

P.C.B.

Perron

(1988). Testing for a unit root in time series regression. Biometrika , 75(2), 335–346.

21.

Prakash

(1977). Regional inequalities and economic development with special reference to infrastructural facilities in India. Indian Journal of Regional Science , 9(2), 172.

22.

Rakesh

Kapur

(2014). Secular stagnation: Can India buck the trend? THINK TANK 20: Growth, convergence and income distribution: The Road from the Brisbane G-20 Summit.

23.

Rao

, (1977). Identification of backward regions and the trends in regional disparities in India. ArthaVijana , 9(2), 93–112.

24.

Romer

P. M.

(1997). Endogenous Technological Change. Journal of Political Economy , 98(5), S71–S102. International Library Of Critical Writings In Economics, 77, 250–281.

25.

Shah

(1970). Overall Summary: Infrastructure for the Indian Economy. In Dagli

Vadilal

(Eds), Infrastructure for the Indian Economy (pp. 32–40).

26.

Solow

R.M.

(1956). A contribution to the theory of economic growth. Quarterly Journal of Economics , 70(1), 65–94.

27.

Tewari

R.T.

(1983). Inter-regional Pattern of Industrialization in India: 1971–1981. In R.T.

Tewari

(Ed.), Development and Change in India . New Delhi: Asish Publishing House.

28.

Tewari

R.T.

(1984). Economic infrastructure and regional development in India. Man and Development , 6(4).

29.

Toda

Phillips

P.C.B.

(1993). Vector auto-regressions and causality. Econometrica , 61, 1367–1393.

30.

Toda

H.Y.

Yamamoto

(1995). Statistical inferences in vector autoregressions with possibly integrated processes. Journal of Econometrics , 66, 225–250.

31.

World Bank (1994). World development report 1994: Infrastructure for development. New York: Oxford University Press.