The effect of regulatory governance on efficiency of thermal power generation in India: A stochastic frontier analysis

Energy Policy 89 (2016) 11–24

Contents lists available at ScienceDirect

Energy Policy journal homepage: www.elsevier.com/locate/enpol

The effect of regulatory governance on efficiency of thermal power generation in India: A stochastic frontier analysis Ranjan Ghosh a,n, Vinish Kathuria b,1 a b

Department of Economics, Swedish University of Agricultural Sciences (SLU) Ulls hus, Ulls väg 27, 756 51 Uppsala, Sweden SJM School of Management, Indian Institute of Technology Bombay, Mumbai 400076, India

H I G H L I G H T S








The impact of regulatory governance on Indian generation efficiency is investigated. Stochastic frontier analysis (SFA) on a panel dataset covering pre and post reform era. Index of state-wise variation in regulation to explain inefficiency effects. Results show improved but not very high technical efficiencies. State-level regulation has positively impacted power plant performance.

art ic l e i nf o

a b s t r a c t

Article history: Received 23 February 2015 Received in revised form 7 October 2015 Accepted 9 November 2015

This paper investigates the impact of institutional quality – typified as regulatory governance – on the performance of thermal power plants in India. The Indian power sector was reformed in the early 1990s. However, reforms are effective only as much as the regulators are committed in ensuring that they are implemented. We hypothesize that higher the quality of regulation in a federal Indian state, higher is the efficiency of electric generation utilities. A translog stochastic frontier model is estimated using index of state-level independent regulation as one of the determinants of inefficiency. The dataset comprises a panel of 77 coal-based thermal power plants during the reform period covering over 70% of installed electricity generation capacity. The mean technical efficiency of 76.7% indicates there is wide scope for efficiency improvement in the sector. Results are robust to various model specifications and show that state-level regulators have positively impacted plant performance. Technical efficiency is sensitive to both unbundling of state utilities, and regulatory experience. The policy implication is that further reforms which empower independent regulators will have far reaching impacts on power sector performance. & 2015 Elsevier Ltd. All rights reserved.

Keywords: Indian thermal efficiency Regulatory governance Stochastic frontier analysis

1. Introduction The impact of regulatory reforms on utility level efficiency in developing countries remains under investigated. One probable reason is that in several of these countries reforms have been enacted in piecemeal manner (Erdogdu, 2013), thereby testing for their effectiveness is inconclusive. Still a key inference from the process of reforms in these countries is that reforms are effective only as much as the regulators are effective in ensuring that they are implemented, which is about how committed they are (Dubash, 2008). This is an institutional issue and in recent times there n

Corresponding author. E-mail addresses: [email protected] (R. Ghosh), [email protected] (V. Kathuria). 1 Tel.: þ91 22 25767863. http://dx.doi.org/10.1016/j.enpol.2015.11.011 0301-4215/& 2015 Elsevier Ltd. All rights reserved.

has been an increasing interest on, and acceptance of, the role of institutional quality in determining the performance in the utilities sector (Cubbin and Stern, 2006; Erdogdu, 2013). While the earlier emphasis of the regulatory literature has been on incentives (Loeb and Magat, 1979; Laffont and Tirole, 1993), the new institutional economics is concerned with governance (Spiller and Tommasi, 2005). In distinguishing between incentives and governance, Levy and Spiller (1994) refer to incentives as the rules related to utility pricing, subsidies etc. and governance as the ways in which high credible commitments are generated. Unless there is a commitment against expropriation of rents, investments in high asset-specific electricity infrastructure does not take place (Ghosh and Kathuria, 2015). In liberalized electricity sectors this role of credible commitments is delegated to independent regulators. It is therefore important that the role of regulators in performance enhancement be tested empirically.

12

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

India is a perfect laboratory for such analysis since it has a federal structure with all the states having some flexibility and individual responsibility for electricity reforms (Sen and Jamasb, 2010). In India a system of independent regulation in power sector began in 1998 with the passage of Electricity Regulatory Commissions Act, 1998. The Act resulted in constitution of the Central Electricity Regulatory Commission (CERC) which lays down the major guidelines and has jurisdiction over both-centrally-owned utilities and inter-state transmission and trade issues. Additionally, each federal state was mandated to have its own regulatory authority known as State Electricity Regulatory Commissions (SERCs) with regulatory oversight at the state-level. The establishment of independent regulation was followed by another set of regulatory reforms in the form of the Electricity Act, 2003 (Bhattacharya, 2005). Thus, compared to the regulated pre-reform era of 1990s by the middle of the decade of 2000s the institutional foundations for a liberalized power sector were clearly laid down. Some of the key changes brought in were the de-licensing of thermal generation and captive production allowing for private participation, licensefree generation and distribution in rural areas, non-discriminatory open-access in transmission, a road-map for open access in distribution, provision for power trading and setting up of multi-year tariff principles (Singh, 2006). These were meant to usher in favorable environment for a competitive market and induce power plants to operate at higher efficiency levels. Additionally there were special measures to ensure that there is competition in the generation segment. Private investors were assured a fair rate-ofreturn and states were advised to unbundle their electricity boards into separate generation, transmission and distribution functions within the electricity supply chain (Dubash and Rao, 2008). There were clear regulatory guidelines instituted to compensate power plants based on their scheduled generation and operating heat rate. Provisions were made to compensate for the fixed costs (like interest on loans and return on equity) based on plant availability 2 (load factor). On the input side, tariffs on coal imports were also reduced (Chikkatur et al., 2007; Malik et al., 2011). However, there is no conclusive evidence so far to suggest that regulators have been successful in implementing these regulatory changes thereby impacting the performance of the power sector. This paper fills the gap by investigating the impact of regulators on the technical efficiency of thermal power plants (which is almost entirely coal-based) in India.3 We estimate a single stage inefficiency effects model, which accounts for both the technical change and time-varying inefficiency effects in a single equation and avoids the problem of inconsistency often encountered in the two-stage approach (Battese and Coelli, 1995; See and Coelli, 2012). Dynamic technical efficiency is calculated using a stochastic frontier method for a panel of 77 coal-based power plants in India for the period 1994–95 to 2010–11. An index of state-level regulation is computed to be then used as a variable explaining the thermal power plant’s inefficiency. The index captures governance and is an aggregation of sub-indexes like tariff setting, transmission and distributions gains, age of regulatory commission, unbundling and regulatory commission composition. It covers 11 year period from 2000–01 to 2010–11 and is constructed for 14 major Indian states. It is hypothesized that the efficiency level of a generation utility in a state is positively impacted by the quality of regulation in that federal state. The paper improves on existing literature in several ways: first, 2 An availability based tariff (ABT) controls power supply to the grid where a tariff is imposed upon any deviation from scheduled generation metered by system frequency. 3 Our focus is on coal based power production as it is the backbone of Indian electricity system comprising nearly 70% of total electricity generated (Shrivastava et al., 2012).

an index of state-level regulation is constructed which helps measure the impact of how regulators govern the sector. The index includes institutional aspects which have been often neglected in the literature dealing with utility level performance. The use of the index also reflects a sense of reality in terms of the relative policy importance of the indicators and facilitates a sensitivity analysis. Second, the study covers the period from 2000 to 01 onwards when regulatory reforms were initiated in India and regulatory agencies were established. This gives the advantage of using federal state level regulatory variances to explain technical inefficiencies. Besides, a recently developed method of a single stage inefficiency effects model is applied. Lastly, instead of the standard Cobb–Douglas production function as used in the studies before, a flexible translog production function is used. The rest of the paper is structured as follows: Section 2 reviews the literature on efficiency analysis of Indian thermal power plants, the determinants of technical inefficiency and the role of regulatory factors on utility performance. Section 3 describes the stochastic frontier analysis (SFA) method that incorporates exogenous influences on efficiency. Section 4 details the empirical strategy by specifying the stochastic frontier model and describing the data. The section explains how the composite regulatory index is constructed and the data sources for the key indicators. It then presents the results from the estimation of model parameters. Section 5 provides explanations for the observed results on the determinants of technical inefficiency, especially the regulatory factors. Section 6 summarizes the paper and concludes with potential policy outcomes.

2. Literature review The literature on impacts of regulation on utility level efficiency, specifically in the Indian context, is classified into three strands in this section. Since the focus is on Indian power plant performance, we begin by discussing the relevant literature available on Indian thermal efficiency. As will be revealed subsequently, there is not much information on the possible factors behind inefficiency in the Indian context. There has been some on plant level factors but hardly any on the regulatory determinants. Therefore, the literature on how plant level factors affect utility performance is discussed followed by the regulatory determinants of inefficiency in the international context. The purpose of this review is to help situate the study around the developments in this field and also to locate the gaps which need to be filled. Moreover, the discussion on regulatory determinants lays down the theoretical frame which emphasizes the need to assess the role of the institutional quality of regulation through the governance aspect. The earliest study of Indian thermal power efficiency was conducted by Singh (1991). It used plant level data for the year 1986–87 to estimate a deterministic frontier production function and calculate technical efficiencies. The results showed that the efficiency of a power plant was positively influenced by its size and capacity utilization. However, efficiency was not associated with its location. Khanna and Zilberman (1999) measured the efficiency of 63 coal-based power plants prior to the reforms i.e., from the period 1987-1988 to 1990–91. They found that efficiency improved with ‘high-heat content’ coal, private ownership and better management practices. Inefficient operation, lack of coal washing facilities and high imported coal tariffs reduced efficiency of thermal plants. Khanna et al. (1999) estimated a stochastic frontier cost function for 66 thermal power plants in India for the period 1987–88 to 1990–91. They found that publicly owned power plants were less efficient than private plants. Plant age did not have a significant effect on efficiency while capacity utilization

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

positively impacted it. The average inefficiency of plants was 48%, whereas there was a 300% difference between the most efficient and the least efficient plant. More recently, Shanmugam and Kulshreshtha (2005) used a stochastic frontier method to calculate the technical efficiency (TE) of 56 coal-based thermal power plants for the period 1994–2001. They used a Cobb–Douglas (CD) function for estimation. The output variable was annual power generated in giga-watt hours (GWh) and the input variables were specific coal consumption, auxiliary consumption and secondary oil consumption. They found the mean technical efficiency of Indian thermal power plants to be 73%. The most efficient plant had a TE score of 96% while the lowest plant had a score of 46%. The western region of India was technically more efficient than the other regions and younger plants performed better than the older ones. Coal input and capital were the most significant determinants of plant productivity. Malik et al. (2011) analyzed the effects of Indian electricity sector restructuring on the operational efficiency of state-owned thermal power plants using a difference-in-differences (d-i-d) method. They used specific coal consumption, operating heat rate, deviation of operating heat rate from design heat rate, plant availability, plant load factor and auxiliary consumption as independent variables. The d-i-d method helped capture the variation in timing of reforms across different states of India. The unbalanced panel consisted of 83 thermal power plants across 17 different federal Indian states for the years 1988–2009. Their key results suggest that while unbundling has improved annual plant availability (PLF) due to a reduction in forced outages, there has been no effect on operating heat rate. They also found that the biggest improvements took place in states which unbundled before passing of Electricity Act, 2003. Shrivastava et al. (2012) estimated the TE of thermal power plants in India using DEA method. The output variable was electricity generated and input variables were specific coal consumption, secondary oil consumption and auxiliary power consumption. Their results suggest that 31.67% of power plants were good performers, 35% were moderate performers, and 23.33% were laggards whereas 10% were poor performers based on a ranking of their variable returns to scale (VRS). Smaller power plants were on average less efficient than medium and large plants. State-owned power plants had lower performance as compared to central and privately-owned power plants. Internationally, studies have shown that privately-owned utilities have higher efficiency than publicly-owned ones (Bagdadioglu et al., 1996; Hiebert, 2002; Berg et al., 2005). Yet there is no clear consensus as some studies have also shown that public ownership leads to higher efficiency (Färe et al., 1986; Khanna et al., 1999). Some have even argued that both private and publicly owned utilities can perform equally (Färe et al., 1985; Pollitt, 1995). Plant level factors like plant age, plant size, fuel type and market share have also been used to explain plant level efficiency (See and Coelli, 2012). Generally higher capacity utilization and lower plant age is associated with higher technical efficiency (Khanna et al., 1999; Hiebert, 2002). The type of fuel used (i.e. coal or natural gas) and the type of plant (base-load or peak-load) has some impact on the efficiency levels, although there are conflicting results about which fuel type or plant type is better (Färe et al., 1986; Pollitt, 1995; Diewart and Nakamura, 1999). Several studies have supported a positive relation between bigger plant size and greater efficiency (Joskow and Schmalensee, 1987; Meibodi, 1998) whereas some have argued that this relationship may not necessarily hold (Sarica and Or, 2007). On the role of market share, Olatubi and Dismukes (2000) in their analysis of competitive effects on generation efficiency in US have shown that greater the share in total generation of a utility, lower is the level of inefficiency. It has also been found that the thermal efficiency of power plants is low in developing countries as compared to the

13

developed countries (Maruyama and Eckelman, 2009). This is not surprising as in most of the developing countries context the power sector had been under strict governmental control with hardly any competitive pressures or incentives to improve performance. However, with liberalization this has been changing and there is a move towards competition. India is one such case being one of the largest countries to be in this transition phase. Yet, there is a very limited set of studies on the trends in efficiency of thermal power plants in India. When it comes to the role of the operational environment of a utility, institutional factors like regulation are the most important. Regulation, either through incentive instruments or through ensuring that there are consistent returns to investment, affects the level of utility performance. But this relationship has not been very highly emphasized in the literature. So even as there is substantial literature on the measurements and comparisons of efficiencies (Hiebert, 2002; Abbott, 2006; Granderson, 2006; Graus et al., 2007; Goto and Tsutsui, 2008; Sueyoshi and Goto, 2011; Jaraitė and Maria, 2012; Jung and Lee, 2014), only a handful of studies have analyzed the regulatory determinants. Some relevant ones are as follows: Lam and Shiu (2004) found that in the Chinese electric generation sector, environmental factors and autonomy from state control influenced efficiency scores. In a study of US electric generation utilities using the non-parametric data envelopment analysis (DEA) method, Olatubi and Dismukes (2000) found that regulatory expenditures and incentive regulation increased generation efficiency. Knittel (2002) tested the impact of alternative regulatory mechanisms on firm efficiency in the US electric generation plants using a stochastic frontier method and found that heat rate and plant availability programs positively impacted efficiency. When it comes to the role of institutional quality (typified as regulatory governance) in the electricity sector, the literature is narrower (Cubbin and Stern, 2006; Zhang et al., 2008; Andres et al., 2008). The regulatory governance literature actually stems from the seminal work of Levy and Spiller (1994) and related works (Noll, 1989; Stern and Holder, 1999; Dubash, 2008). It builds on ‘economics of governance’ (North, 1990; Williamson, 2005) referring to those set of devices which are used to bring order in transactions through well-defined property rights, contracts and other enforcement mechanisms (Dixit, 1996). The fundamental problem is about how to bind players into agreements or how to credibly commit to enable complex contracting (North, 1993). In the utilities sector, the core argument is that higher the quality of institutions more sustained will be the investment flow. Assets in utilities are not only very capital intensive but most of the times are either specific in nature (with very low opportunity costs outside the particular use) or are characterized by sunk investments. This mandates an effective institutional framework which includes an independent regulatory agency. It is then incumbent on the agency to provide the required institutional quality for efficient functioning of the utility sector. Therefore, where the institutional design includes an independent regulator, the institutional quality will be determined by regulatory governance (Cubbin and Stern, 2006). In an inter-country comparison based on a sample of 28 countries, Cubbin and Stern (2006) use per-capita generation capacity (by country and year) as the dependent variable. The key independent variables include dummies indicating presence of electricity (or energy) regulatory law, presence of an autonomous or ministry regulator, license fee or government budget regulatory funding, and civil service pay scales for regulatory staff. The study found that regulatory governance was a consistent factor in improved per-capita generation capacity. The review in this section has shown that studies attempting to find the impact of institutional quality on utility performance are not many. Specifically for India, there is not much on the impacts

14

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

70

under which production takes place and which is beyond the control of production managers. Then a cross-sectional stochastic frontier could be specified as in Eq. (1) below:

60

lnyi = lnf (xi , zi; β)+vi − ui

50

40

30 1995

2000

2005

2010

Year of observation Annual Generation in '0000 GWh

Annual PLF (%)

Fig. 1. All India trends – generation and PLF Source: Own compilation.

of independent regulation on plant-level performance. Moreover, while annual generation in India has been rising, plant utilization measured through the plant load factor (PLF) has been falling especially in the last few years as shown in see Fig. 1. Up to 1995, PLF was at low levels of 55–57% but then it began to improve in the next decade and peaked in 2006 when it went above 70%. But there has been a decline ever since and the latest estimates for 2014–15 suggest that it has fallen to the lowest levels in last 15 years4. Therefore, a greater understanding about efficiency of Indian thermal power plants and the role of exogenous influences are necessary. The next section explains briefly the method employed in this study.

3. Method: Stochastic frontier analysis In this section, we present the method used to estimate technical efficiency of thermal power plants and the effects of regulation on it. The hypothesis that higher quality of regulation by a regulator in a federal state leads to higher generation efficiency is tested using a stochastic frontier analysis (SFA) method5 that incorporates exogenous influences on efficiency. Exogenous variables are those factors that are parts of neither the input nor the output processes, yet may affect plant performance by characterizing the production environment (Kumbhakar and Lovell, 2000). They may influence the input to output conversion process through impact on the technology structure or may impact the efficiency of the conversion process. We specifically choose a stochastic frontier production function (Eq. (2)) that helps model technical inefficiency effects i.e. the exogenous influences, in single stage (Battese and Coelli, 1995). This avoids the problem of inconsistency which is possible in the two-stage approach. The problem in two-stage estimation follows from some of the initial approaches to incorporate exogenous influences (Simar et al., 1994; Pitt and Lee, 1981; Mester, 1993, 1997).When an input vector x i is used to produce a scalar output yi then zi could be a vector of exogenous variables that captures the environment 4 http://economictimes.indiatimes.com/industry/energy/power/india-sees-low est-plant-load-factor-in-15-years-power-capacities-operating-at-65/articleshow/ 47463610.cms accessed on 27.09.15. 5 Stochastic frontier approach has an advantage over deterministic frontier approach because it allows for random deviations from the production frontier due to factors beyond the control of producers. It allows for a separation of the error term into a two-sided random noise component, Vit in Eq. (2), and a nonnegative inefficiency component, Uit in Eq. (2) (Kumbhakar and Lovell, 2000; Seo and Shin, 2011)

(1)

vi captures the random noise and ui captures the technical inefficiency. An important assumption here is that vi and ui are independently distributed of each other as well as the input regressors. As compared to a stochastic frontier which does not include exogenous influence, the vector of parameters, β, now contains technological as well as exogenous influences. This way zi directly influences output yi , not by influencing inefficiencies but the structure of the production frontier which bounds the relation between production inputs and output (Kumbhakar and Lovell, 2000; pp. 263). In two-stage approach, variation in exogenous influences zi is associated directly with variation in estimated efficiency. In the first stage a stochastic frontier (such as Eq. (1)) is estimated (without the inclusion of zi ) from which efficiencies are estimated. These efficiencies are then regressed in a second stage against the exogenous variables (ibid.). This brings us to the problems of inconsistency in the two-stage process mentioned earlier. First, it is a requirement that elements of zi are not correlated with those of x i . If they are correlated then we get biased estimates of β due to omitted variables. This makes the estimated efficiencies biased too and hence interpretation of second-stage results problematic. Another problem is that while in the first stage ui is assumed to be identically distributed, in the second stage a function relationship is formulated with exogenous influences zi,again causing inconsistency (Kumbhakar and Lovell, 2000). This problem is resolved in Kumbhakar et al. (1991), Reifschneider and Stevenson (1991), Huang and Liu (1994) and extended to panel data estimations by Battese and Coelli (1995). To see how, first Battese and Coelli (1995) frontier for panel data is specified as follows:

yi =exp (xit β + vit − uit )

(2)

where, yit is output of the ith plant in tth year x it is a vector of the production inputs of the ith plant in tth year; β is the vector of unknown parameters; vit s are the random errors assumed to be independently distributed of uit s; uit s are non-negative random variables associated with the technical inefficiency effects. These are assumed to be a function of a set of explanatory variables given by zit and an unknown vector of coefficients δ . Eq. (3) specifies this as follows:

uit = zit δ + wit

(3)

wit is the random component of inefficiency, uit For a single stage estimation the frontier can be arrived at by substituting (3) in (2):

yi =exp (xit β − zit δ + vit − wit )

(4)

For maximum likelihood estimation (MLE) estimation of Eq. (4) it is a requirement that uit = zit δ + wit ≥ 0. Looking at Eq. (3), this is possible when wit is truncated from below such that wit ≥ − zit δ and by assigning a distribution to wit such that wit ~N (0,σw2 ) (Huang and Liu, 1994). It means that if wit cannot take a value lesser than −zit δ then uit will always be positive. Thus by truncating a normal distribution with zero mode from below at a variable truncating point this specification allows estimation of Eq. (4) by MLE. Essentially what this variable truncation does is to allow the mode to be a function of exogenous variables zit . Since inefficiency depends on the mode of truncated normal distribution, inefficiency now depends on exogenous variables and is expressed as:

TEit = exp( − uit )=exp (−zit δ − wit )

(5)

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

15

4. Empirical estimation

4.1. Production function

This section explains the empirical strategy, the data for estimation of production function, data for construction of the index, the reasoning behind the choice of indicators and the results. For empirical analysis the frontier model stated in Section 3 needs to be specified with the production function input–output variables as well as the exogenous variables which explain some of the inefficiency. This will serve the two objectives of the study: get efficiency estimates of Indian thermal power plants for the periods before and after regulatory institutions were introduced; and find out what impact state-level regulation has on technical efficiency variations. Following Battese and Coelli (1995), a translog function is specified because of its flexibility. As compared to a Cobb– Douglas form, a translog function does not require perfect substitution between inputs or perfect competition in their markets. This makes it more realistic for the Indian case where factor markets operate under several restrains. Also this allows for nonlinearity in relationship between inputs and outputs which makes the tests more relevant for a power sector which has been reformed only recently. The stochastic frontier function to be estimated is as follows:

Data for the plant level input and output variables were collected from the annually published Performance Review of Thermal Power Stations by the Central Electricity Authority (CEA) of India. The final dataset used is a balanced panel containing 77 coal-fired (base-load) power plants and covering a 17 year time period from 1994–1995 to 2010–2011 across 14 major states in India. These states fall in all five regions of India: North (Rajasthan, Punjab, Haryana, Delhi), Central (Uttar Pradesh, Madhya Pradesh/ Chhattisgarh), East (Bihar/Jharkhand, West Bengal, Orissa), West (Maharashtra, Gujarat) and South (Tamil Nadu, Karnataka and Andhra Pradesh). The CEA reports contain data for 13 more thermal plants (some became operational after 1995 and some ceased to operate in recent years). These could not be included given the objective was to observe the change in technical efficiencies of plants which were operational before and after the regulatory reform period. Power plants from those states which had less than 1% of the total share of thermal generation were excluded. Power plants in the sample constitute roughly 77% of the total coal-fired installed capacity and roughly 60% of the total thermal capacity in India.6 89% of the plants in the sample are government (central and state sector) owned whereas 11% are privately owned. Table 1 presents descriptive statistics of the output and input variables used in estimation of the production function. Annual power generated (in Gigawatt-Hour, GWh) is the output variable as has been often used in the literature (Kopp and Smith, 1980; Shanmugam and Kulshreshtha, 2005; Goto and Tsutsui, 2008; Du et al., 2009; See and Coelli, 2012). The average power generation is 4774.22 GWh with the smallest plant generating 77.26 GWh and the largest generating approximately 27586 GWh. The input variables7 based on previous important studies on Indian thermal power plants (Singh, 1991; Shanmugam and Kulshreshtha 2005; Shrivastava et al., 2012) are capital in GWh, specific coal consumption per unit of generation in gram/KWh, specific secondary oil consumption per unit of generation in liter/GWh and auxiliary power consumption in percentage8. Coal is the most important fuel input in thermal generation in India and the average consumption is 756.85 g/KWh. The mean secondary oil consumption – additional fuel other than coal – is 5395.34 l/GWh. The mean auxiliary power consumption-self-consumption of energy for operation of a power plant – is 9.77%. While the other three input variables are straightforward, capital input variable is constructed following previous literature (Dhrymes and Kurz, 1964; Singh, 1991; Shanmugam and Kulshreshtha, 2005) as:

lnPowerit = β0 + β1ln(Capitalit ) + β2ln(Coalit ) + β3ln(Secondary Oil it ) + β4ln(Auxiliaryit ) + β5(Yearit ) + 0. 5β ln(Capitalit )2 + 0. 5β7ln(Coalit )2 6

+ 0. 5β8ln(Secondary Oil it )2 +0. 5β9ln(Auxiliaryit )2 + 0. 5β10(Yearit )2 +β ln(Capitalit )ln(Coalit ) 11

+ β12ln(Capital it )ln(Secondary Oilit ) + β13ln(Capitalit ) ln(Auxiliaryit ) + β14ln(Coalit )ln(Secondary Oilit ) + β15ln(Coal it )ln(Auxiliaryit ) + β16ln(Secondary Oil it )ln(Auxiliaryit ) + β17ln(Capitalit )(Yearit )+β18ln(Secondary Oilit )(Yearit ) +β19ln (Coal it )(Yearit ) + β20ln(Auxiliaryit )(Yearit )+v

it

− uit

(6)

where, vit is the random error term, uit is the technical inefficiency effects and can be explained by:

uit = π 0 + π1RegulatoryIndex it + π PlantCapacityit + π3PlantAgeit 2

+ π4Yearit

CAPITAL= (7)

The time variable Yearit in Eq. (6) accounts for Hicks-neutral technological change. The time variable Yearit in Eq. (7) models changes in technical efficiency as time increases. This identification of both the technical change and time-varying inefficiency effects is made possible because of the distributional assumptions on the inefficiency effects (Battese and Coelli, 1995; See and Coelli, 2012). The time variable interacted with the input variables in Eq. (6) is to allow for non-neutral technical change. The time squared variable is to allow for non-monotonic technical change. Though it is common to use Hicks-neutral technical change, in the present context where Indian thermal power generation industry is often considered to have excess capital, the change in regulatory regime may not be Hicks neutral. Also, the change may not be monotonic, it may vary with increase in the strength of regulatory index. Hence, we model non-monotonic and non-neutral technical change.

S* T *(1 − PM) 1000

(8)

where, S is the installed plant capacity (in mega watts, MW); T is the number of hours in a year and PM is the planned 6

Own calculations based on CEA reports. Labor is not used as an explanatory variable although it is an important input. This is primarily for two reasons: firstly, there is no reliable data available for manpower used in Indian power plants (including the centralized source of CEA). This explains why no previous estimation of thermal efficiency in India has ever used labor as an input. There is only one exception of (Khanna et al., 1999) where expenses on labor have been used to estimate a cost function for Indian power plants but the data was for the years 1987–88 to 1990–91. That was the pre-reform era. After early 1990s CEA has stopped publishing plant-wise manpower related information. Additionally it has also been argued in the previous literature that labor is non-substitutable for fuel and capital (Komiya, 1962; Kopp and Smith, 1980; Singh, 1991; Shanmugam and Kulshreshtha, 2005) and electricity output can be treated as dependent only on fuel and capital. 8 The term ‘specific’ coal and ‘specific’ secondary oil are used by the Central Electricity Authority, Government of India in its reports based on international conventions. Henceforth for simplicity we will use the terms Coal and Secondary Oil to signify these variables. 7

16

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

Table 1 Summary statistics of production-function variables.Source: Own compilation

Power generated (GWh) Capital (Gwh) Coal (gm/KWh) Secondary oil (liter/Gwh) Auxiliary (%) Plant age (Years) Plant capacity (MW) Regulatory index

Obs

Mean

Std. Dev.

Min

Max

1280 1280 1273 1278 1280 1280 1280 847

4774.22 6131.16 756.85 5395.34 9.77 21.05 764.19 0.44

4569.7 4918.38 156.71 8085.35 2.03 9.68 592.8 0.31

77.26 147.28 110 50 0.71 1 30 0

27585.85 27432.43 2000 81890 19.72 49 3260 1

Note: The observations for Regulatory Index is truncated because the index is calculated only for the time period 2000–2010 whereas the full production model is for the time period 1994–2010

13.6

13.4

13.2

13

12.8

Table 2 Summary statistics of state-level regulatory indicators.Source: Own estimations Variable

Obs.

Mean

Std. Dev.

Min.

Max.

Tariff (Rs./KWh) T&D gains (%) Regulator age (years) Unbundling Regulator composition

154 154 154 154 154

3.01 68.49 5.83 0.66 0.25

0.64 9.56 3.36 0.47 0.43

1.54 42.91 0 0 0

4.68 86.53 13 1 1

2

2.5

3

Y - Log of Annual Generation in ‘0000 GWh

3.5

4

X – All India Average Tariffs in Rs/KWh

Fig. A3. Relation between all India average tariffs (2000–2010) and thermal generation Source: Own compilation.

Fig. A4. All India AT&C gains Source: Own compilation from data sources (CEA and Indiastat).

Fig. A1. State-wise and All-India average tariffs for the last decade Source: Own compilation based on SERC tariff orders.

Fig. A5. State-wise age of regulatory commissions Source: Central Electricity Regulatory Commission (CERC), India.

Table 3 Results of hypothesis testing.Source: Own compilation Fig. A2. All-India trend in average tariff Source: Own compilation based on SERC tariff orders.

maintenance with (1-PM) term indicating what percentage of time the plant was operational. The mean capital consumption is 6131.16 GWh with values for different power plants ranging from a minimum of 147.28 GWh to a maximum of 27,432.43 GWh

Test

Null hypothesis (H0)

Test statistic

Critical value

Cobb–Douglas

β6 − 20=0

215.72

18.31

Reject H0

Inefficiency Effects

δ1 − 3=0

11.02

7.81

Reject H0

Decision

( χ 2α=0.05 )

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

17

Table 4 Estimated parameters of Battese and Coelli (1995) (BC95) inefficiency effects model. IE: Inefficiency Effects

BC95_NORG (Model 1) IE: Plant age þ installed capacity

BC95_RG (Model 2) IE: Plant age þ installed capacity þ RG index

Capital

1.7637*** (0.2536) 0.8974 (1.0051) 1.3365*** (0.2148)  0.3624 (1.4698)  0.0317 (0.0429) 0.0009 (0.0078) 0.0291 (0.0460)  0.0198*** (0.0047)  0.1268*** (0.0271) 0.0001 (0.0002)  0.1168*** (0.0434) 0.0185** (0.0088)  0.1092** (0.0465)  0.1545*** (0.0329) 0.3606 (0.2227)  0.1277*** (0.0396) 0.0049*** (0.0016)  0.0022 (0.0013)  0.0043 (0.0063) 0.0150** (0.0075)  9.3996** (4.7415) 2.8466*** (0.9980)  2.3571*** (0.7022)

1.6280*** (0.2164) 2.7005*** (0.9214) 0.2741 (0.2523) 4.5150*** (1.6740)  0.0335 (0.0704) 0.0089 (0.0071)  0.0889** (0.0405)  0.0149*** (0.0046)  0.1646*** (0.0337)  0.0003 (0.0005)  0.1022** (0.0404) 0.0141 (0.0089)  0.1038** (0.0419) 0.0063 (0.0388)  0.3736 (0.2453)  0.1206*** (0.0342)  0.0008 (0.0022)  0.0037** (0.0015) 0.0054 (0.0102) 0.0187* (0.0103)  16.5484*** (4.4921) 5.1321** (2.0152)  2.8232*** (0.9389)  2.7951* (1.6066)

Coal Secondary oil Auxiliary Year Capital squared Coal squared Secondary oil squared Auxiliary squared Year squared Capital x coal Capital x Secondary oil Capital x auxiliary Coal x secondary oil Coal x auxiliary Secondary oil x auxiliary Capital x year Secondary oil x year Coal x Year Auxiliary x Year Constant Plant age Plant capacity Regulatory index

BC95_RG_AHW (Model 3) IE: Plant age þ installed capacity þ RG index with Higher weights for age of regulators 1.5690*** (0.2211) 2.2732** (0.9522) 0.3129 (0.2537) 3.9439** (1.6797)  0.0287 (0.0708) 0.0081 (0.0070)  0.0755 (0.0460)  0.0144*** (0.0045)  0.1568*** (0.0336)  0.0003 (0.0005)  0.0908** (0.0405) 0.0134 (0.0089)  0.1030** (0.0419) 0.0018 (0.0388)  0.2854 (0.2454)  0.1260*** (0.0339)  0.0008 (0.0022)  0.0036** (0.0015) 0.0048 (0.0102) 0.0179* (0.0102)  14.4342*** (4.5714) 5.7825*** (1.6810)  3.4550*** (0.8059)

BC95_RG_UBHW (Model 4) IE: Plant age þ inst. cap. þ RG index with higher weights for unbundling 1.5524*** (0.2108) 2.1471** (0.8895) 0.3363 (0.2498) 3.8165** (1.6403)  0.0386 (0.0707) 0.0027 (0.0068)  0.0944** (0.0388)  0.0135*** (0.0044)  0.1607*** (0.0325)  0.0003 (0.0005)  0.0683* (0.0393) 0.0048 (0.0087)  0.0961** (0.0389) 0.0133 (0.0375)  0.2593 (0.2394)  0.1377*** (0.0326)  0.0002 (0.0021)  0.0040*** (0.0015) 0.0065 (0.0101) 0.0169* (0.0098)  13.7657*** (4.3601) 6.3675** (2.7616)  4.1448** (1.6674)

 3.1878**

Regulatory index with higher weight for age of regulators

(1.5519)  10.5299**

Regulatory index with higher weight for unbundling N pseudo R2 sigma_u sigma_v Log lik. Chi  squared

1269

830

830

(4.4618) 830

1.46 0.1012 241.01 30557.55

1.67 0.0524 260.90 36423.85

1.79 0.0522 261.76 37602.43

1.96 0.0487 277.80 40905.55

Notes: Standard errors in parentheses *

p o 0.10, po 0.05, *** p o 0.01. **

(Table 1). Table A1 (Appendix) gives the correlation matrix of the production function input variables. Table 1 also presents summary statistics for the explanatory variables of technical inefficiency as identified in Eq. (7) above.

This includes the plant level factors – age and capacity – and the index of state-level regulation (or regulatory index). Plant age is calculated as the difference between the particular year of operation and the year the plant was commissioned. The mean plant

18

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

Table 5 Indicator weights in different models for sensitivity analysis.Source: Own compilation Tariff AT&C Gains Model 2 0.4 Model 3 0.3 Model 4 0.3

0.4 0.3 0.3

Age of Regulators

Unbundling Regulatory Composition

0.1 0.3 0.1

0.05 0.05 0.25

0.05 0.05 0.05

age is 21.05 years. The lowest plant age of 1 year means that there are power plants which were commissioned in the year 1993–94 and therefore when our period of observation begins in 1994–95, they were 1 year old. The oldest plant in our data-set is 49 years old. Plant capacity is the installed capacity of a power plant measured in MW with a mean of 764.19 MW and ranging within the extreme values of 30–3260 MW. The regulatory index takes values from 0 to 1 with a mean of 0.44. The next sub-section provides a detailed explanation of the various components of the index, the data and the method used to construct it. 4.2. Index of state-level regulation To capture the status of state level regulation, regulatory index is calculated as a composite index of the weighted average of several components which are individual indicators of regulatory performance. This follows a standard method of index construction similar to the ones used for UNDP's Human Development Index (Hicks, 1997). Since the parameters (see Table 2) have different units, the first step in the index construction is to normalize them. This is done using a min-max criterion which is based on the ‘distance from the ideal’ approach (Klugman et al., 2011). Minimum and maximum observed values are set in order to convert the indicators into sub-indices between 0 and 1. For each state for each year the minimum and maximum values are taken. Using the following identity, the dimension or sub-index is calculated as shown in Eq. (9):

θis=(βis,actual − βis,min)/(βis,max − βis,min)

(9)

where, βis ¼ value of parameter i for a federal state s; θis ¼value of score of each parameter i for a federal state s This is individually done for all the parameters. Then all the parameter wise sub-indices are aggregated for each state, year wise. This aggregate index is further normalized using the minmax criteria to arrive at the final composite index as shown in Eq. (10). n θ i = 1 it, actual

ρst =( ∑

−

n

n

n

∑i = 1 θit, min)/( ∑i = 1 θit, max − ∑i = 1 θit, min)

(10)

Where, ρst ¼ value of score of federal state s for time period t. The indicators include tariff setting, transmission and distributions gains, age of regulatory commission, unbundling and regulatory commission composition. Data is collected for 14 states

Table 7 Summary statistics of technical inefficiency (TIE).Source: Own compilation Variable

Obs

Mean

Std. Dev.

Min

Max

TIE_BC95_NORG (Model 1) TIE_BC95_RG (Model 2) TIE_BC95_RG_AHW (Model 3) TIE_BC95_RG_UBHW (Model 4)

1269 830 830 830

.2112 .2330 .2316 .2331

.2338 .2929 .2927 .2967

.0110 .0108 .0106 .0109

2.3105 2.3642 2.3631 2.3616

Table 8 State wise technical inefficiencies.Source: Own compilation

Rank of performance

State wise TIE State name

Region

Mean RTIE

1 2 3 4 5 6 7 8 9 10 11 12 13 14

Rajasthan MP/Chhattisgarh Maharashtra Tamil Nadu Punjab Karnataka Gujarat Haryana Andhra Pradesh Delhi Orissa Uttar Pradesh West Bengal Bihar/Jharkhand

North Central West South North South West North South North East Central East East

0.09 0.14 0.14 0.14 0.15 0.15 0.16 0.18 0.19 0.20 0.20 0.22 0.37 0.44

in India where thermal power plants are located. The index covers the period from year 2000–01 to 2010–11 because the Electricity Regulatory Commissions Act was passed in 1998 and most of the states established their independent regulatory agencies around the year 2000. Apart from Cubbin and Stern (2006) there is no study which uses quantitative estimates of regulatory indicators on the electricity sector. However, their indicators were mostly dummies indicating the presence/absence of electricity (or energy) regulatory law, presence/absence of an autonomous or ministry regulator, procedures for license fee or government budget regulatory funding and civil service pay scales for regulatory staff. These are useful for cross country estimation as there can be variation by country but are restricted in terms of temporal variation. For example, once an electricity law is passed or an autonomous regulator is established the dummy variable will not vary for the subsequent years. For a within-country study like the present one, these are not effective as India has an electricity law at federal level (Electricity Act, 2003) and every state has an autonomous regulator. Because the state regulators were established by legislation, they are funded through government budgets and the staffs being government employees are paid civil service pay scales, hence will bear no variation in case a dummy is used for this. Some other India specific studies (Nakhooda et al., 2007; Teri,

Table 6 Spearman rank correlations of technical inefficiency (TIE) scores.Source: Own compilation

TIE_BC95_NORG TIE_BC95_RG TIE_BC95_RG_AHW TIE_BC95_RG_UBHW

TIE_BC95_NORG (Model 1)

TIE_BC95_RG (Model 2)

TIE_BC95_RG_AHW (Model 3)

TIE_BC95_RG_UBHW (Model 4)

1.0000 0.9631 0.9632 0.9573

1.0000 0.9993 0.9976

1.0000 0.9980

1.0000

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

19

Table A1 Correlation matrix of input variables.Source: Own estimations

Plant Load (PLF) Capital (Gwh) Coal (gm/KWh) Secondary oil (liter/Gwh) Auxiliary (%) *

Plant Load (PLF)

Capital

Coal

Secondary Oil

Auxiliary

1.000 0.4241*  0.4674*  0.6760*  0.5198*

1.000  0.2235*  0.3239*  0.4848*

1.000 0.3942* 0.3819*

1.000 0.3625*

1.000

Regulatory Index

significance at.05 level.

Table A2 Correlations of regulatory indicators.Source: Own estimations

Tariff (Rs./KWh) T&D Gains (%) Regulator age (years) Unbundling Regulator composition *

Tariff

T&D Gains

Regulator age

Unbundling

Regulator composition

1.0000 0.1668 0.6087* 0.2112*  0.1004

1.0000 0.3592*  0.1016 0.1156

1.0000 0.3840*  0.0834

1.0000  0.0773

1.0000

significance at.05 level.

2007; Sen and Jamasb, 2010) throw interesting options but could not be included due to data unavailability.9 In order to make up for the lack of spatial and temporal data on qualitative indicators, the objective outcomes of regulatory process have been used as fair approximation of state-level regulation. The literature agrees that regulatory outcomes are valid indication of how well the regulatory set-up has performed (Teri, 2007). If tariff rationalization and reducing technical losses were the two most important objectives of regulation, then performance on these two counts would be useful indicators of how regulators are governed. Additionally, institutional factors like experience of regulatory commissions, the stability of regulatory commissions, and electricity sector unbundling have been included in the index. It is expected that as time passes, with increase in experiences, regulatory commissions will mature and function better (Cubbin and Stern, 2006; Andres et al., 2008). Stability in regulatory composition ensures there is higher accountability, learning and autonomy and hence would yield better performance. Whether the state has an unbundled electricity sector (with different generation and distribution companies) or is still under state monopoly will determine competitive pressures, role of private entities and the general regulatory environment. If there is state monopoly it may mean a state agency only regulating another state agency, known as idiosyncratic regulation (Dubash, 2008) and hence lower quality of governance. Further details of the five key indicators used for index computation are as follows: a) Annual Power Tariffs (in Rs./KWh): Tariff rationalization has been a key objective of regulatory reforms and a key indicator of regulatory effectiveness (Teri, 2007). Before the reform period, tariffs were much deflated and any loss made by the generation unit was to be financed through fiscal adjustments. Moreover, some segments like households and agriculture 9 Nakhooda et al. (2007), for example come up with qualitative indicators like, (a) the institutional structure of regulatory decision-making i.e. whether decisions are made through executive orders or through independent commissions, (b) how authorized the regulatory body is in seeking information, investigating, penalizing and enforcing, (c) how robust and transparent are the selection criteria for members of regulatory commission, (d) appeal mechanisms, (e) procedural certainty, (f) pro-activeness, (g) public participation, (h) consumer service, and (i) conflict resolution. However there is no available data displaying spatial and temporal variability in these.

were highly subsidized. This meant very high tariffs for industrial (and commercial) users which also caused many of them to set up their own captive power units (Ghosh and Kathuria, 2014). This resulted in even lower revenues for the power plants. With reforms it is expected that regulators will regularize and rationalize the tariff determination process which in turn will incentivize the power plants to increase efficiency in order to gain from higher prices for their power. Data was, therefore, collected for year wise average tariffs of the main consumer categories-domestic, commercial, industry and agriculture-for the 14 states from the year 2000–01 to 2010–11. These came primarily from the annual tariff orders10 of the SERCs. Fig. A1 (Appendix) gives the average tariffs for each state over the period 2000–01 to 2010–11 and also the all-India average. The states of Delhi, Gujarat, Haryana, Maharashtra and West Bengal offered tariff rates higher than the national average. Fig. A2 (Appendix) gives the all-India trend of average tariff for the years 2000–01 to 2010–11 and Fig. A3 (Appendix) plots its relation to the trend in thermal generation which indicate a positive correlation. b) Transmission and Distribution Gains: Transmission and Distribution (now known as AT&C11) losses are those which take place in the overall system due to lack of investments in system improvement, extensions of distribution lines, overloading of transformers and conductors, low metering efficiency, theft and pilferages. Loss figures have been traditionally very high in India till the reforms began and it reached the peak of 32.86% in the year of 2000–01.12 It was expected that with the introduction of independent regulation, there will be reductions in these losses. Data on AT&C loss percentages were collected from government data sources13 and then subtracted from hundred to get AT&C gain figures. Fig. A4 (Appendix) gives the all-India trend of AT&C gains in the last decade which has shown an increase from roughly 67% in year 2000 to around 72% in the year 2010. c) Age, Unbundling and Regulatory Composition: The age of

10 Around 140 tariff orders were analyzed to get the tariff rates where every order ran into hundreds of pages. 11 Aggregate Technical and Commercial Losses. 12 Source: http://powermin.nic.in/distribution/distribution_overview.htm accessed on 03.10.2013. 13 Source: http://www.indiastat.com/default.aspx accessed on 10.10.2013.

20

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

Table A3 Translog SFA models.Source: Own estimations

Table A3 (continued )

TL_FELS_TV TL_FE_TIV TL: Translog; TV: (2) Time Variant; TIV: (1) Time Invariant; FE: Fixed Effects; IE: Inefficiency Effects

TL_FECSS_TV (3)

TL_IEBC95_TV (4)

0.6240 (0.4509) 2.3607** (0.9820) 0.4039 (0.2482) 2.1620* (1.1583)  5.741e þ 11* (3.220e þ 11) 0.0628*** (0.0161)  0.0340 (0.0462)  0.0163***

1.7171*** (0.2413)  0.4488 (0.9008) 1.4838*** (0.2142)  1.4373 (1.4298)  0.0057 (0.0428)  0.0131** (0.0055) 0.0434 (0.0491)  0.0268***

Capital

1.1226

Coal

0.4506 (2.5945) 0.2154 (1.8882)  1.4543 (0.0000) 0.0131 (0.1737)  0.0014 (0.0319)  0.0905 (0.2413)  0.0167

1.5342*** (0.4455) 5.0546*** (1.0519) 0.5003** (0.2382) 2.0058 (1.3432) 0.0908** (0.0407) 0.0453*** (0.0137)  0.1447*** (0.0480)  0.0139***

(0.0191)  0.0483 (0.0941) 0.0001 (0.0016)  0.0202 (0.1357) 0.0025

(0.0048)  0.0590** (0.0257) 0.0001 (0.0003)  0.2362*** (0.0525) 0.0131

(0.0043)  0.0494** (0.0208) 4.172e þ 10* (2.374e þ10)  0.1112** (0.0487)  0.0043

(0.0046)  0.1148*** (0.0260) 0.0002 (0.0003)  0.0483 (0.0341) 0.0079

(0.0365)  0.0354 (0.1981) 0.0002

(0.0084)  0.0219 (0.0366)  0.0805**

(0.0088)  0.0338 (0.0312)  0.0364

(0.0070)  0.1356*** (0.0377)  0.1396***

(0.2205) 0.3211 (0.4428)  0.0373

(0.0367)  0.2682 (0.1919) 0.0165

(0.0353)  0.2792* (0.1684) 0.0189

(0.0337) 0.5647*** (0.2013)  0.1426***

(0.1880) 0.0014 (0.0068)  0.0009

(0.0253) 0.0023 (0.0016)  0.0008

(0.0205)  0.0244*** (0.0056) 0.0005

(0.0359) 0.0039** (0.0015)  0.0039***

(0.0059)  0.0014 (0.0235)  0.0030 (0.0298)

(0.0014) (0.0018)  0.0139**  0.0106 (0.0061) (0.0075) -0.0037  0.0036 (0.0074) (0.0072)  21.9875*** (6.0506)

Secondary oil Auxiliary Year Capital squared Coal squared Secondary oil squared Auxiliary squared Year squared Capital x coal Capital x secondary oil Capital x auxiliary Coal x secondary oil Coal x auxiliary Secondary oil x auxiliary Capital x year Secondary oil x year Coal x year Auxiliary x year _cons Time dummies xi1 xi2 xi3 xi4 xi5 xi6 xi7 xi8 xi9 xi10 xi11 xi12

1.0000 (1.1252) 1.0571 (1.1934) 1.0889 (1.1075) 1.2711 (1.1267) 1.1379 (1.1622) 1.2151 (1.2846) 1.3579 (1.3351) 1.2247 (1.2278) 1.6200 1.4781 (1.3833) 1.5025 (1.3916) 1.5744

(0.0014)  0.0052 (0.0064) 0.0152* (0.0078)  4.2349 (4.3083)

TL_FELS_TV TL_FE_TIV TL: Translog; TV: (2) Time Variant; TIV: (1) Time Invariant; FE: Fixed Effects; IE: Inefficiency Effects

xi13 xi14 xi15 xi16 xi17 N pseudo R2 sigma_u sigma_v Log lik. Chi  squared

TL_FECSS_TV (3)

TL_IEBC95_TV (4)

(1.3669) 1.2195 (1.3120) 1.3009 (1.4036) 1.4600 (1.5435) 1.2663 (1.3082) 1.1998 (1.1558) 1269

1269

1269

1269

0.2427 0.1779

0.3465 0.1734

1.229e þ 12 0.1245

2680.3247

616.7329

1.5246 0.1059 203.1570 40166.0607

Standard errors in parentheses. *

p o0.10, p o 0.05, *** p o0.01. **

regulatory commission was calculated from the year each SERC was established. If an SERC was established after the year 2000, then the previous years were given zero values. Unbundling was given a value of 0 if in that particular year the state had not yet unbundled its electricity board. The change in regulatory composition is indicated by the change in political regime. This is because de-facto regulatory commissions are an extended wing of the government (Dubash, 2008) and any new state government eventually leads to change in the high ranking regulatory staff. Although this may not always be the case, it is a fair approximation given Indian political conditions. In the absence of any concrete data on the changes in regulatory composition, a dummy value of 0 is given for a year when there was no new government as compared to the previous year and a value of 1 if that year witnessed a new state government taking office. Please refer to Table 2 for description of data on these three indicators. In aggregation, higher weights are assigned to the tariff and AT&C sub-index as they are more important representation of regulatory performance (Berg et al., 2005; Jamasb and Pollitt, 2007; Nakhooda et al., 2007), whereas lower weights are attached to other parameters. In subsequent models, however, (see the results section) the weights of the ‘Age of Regulators’ and ‘Unbundling’ were changed and a sensitivity analysis was performed. The reason for performing a sensitivity analysis for these two indicators is because there is keen interest in the literature on them (Cubbin and Stern, 2006; Sen and Jamasb, 2010). It is generally hypothesized that the opening up of the electricity sector through dismantling natural monopoly i.e. unbundling will improve the efficiency of all the segments in the electricity supply chain. It is also believed that with time the experience of regulators will have a positive impact as they will be able to get more information about the utility’s production function and perform overall better coordination. Currently membership of the regulatory commission is part of the governmental civil service limiting ‘regulatory culture to a particular set of experiences, one that arguably lacks practical knowledge of business and consumers and favors a public sector mindset’ (Dubash and Rao, 2008; p. 326). With passing time

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

Table A4 Relative technical efficiency (RTE) scores (in %).Source: Own estimations Plant name

State

RTE Scores

Satpura North Madras Kota Korba II (E) Parli Bhusawal Tuticorin Neyveli Khaperkheda II Singrauli STPS Vijaywada Vindhyachal STPS Korba STPS Suratgarh Nasik Mettur Rajghat Wanakbori Panipat Ramagundem STPS Rayalseema Korba West TPS Koradi Kothagudem Rihand STPS Gandhi Nagar Unchachar Titagarh Anpara Sabarmati/Torrent Ukai Thermal Ropar Raichur Lehra Mohabbat (GHTP) Badarpur (NTPC) Southern REPL Chandarpur Dadri (NCTPP) Sanjay Gandhi Dhanu (BSES) Bhatinda (GNDTP) Kolaghat Neyveli (M Cut) I B Valley Korba III (E) Kutch Lignite Mejia Paricha Sikka REPL Talcher Kahalgaon Durgapur Farakka STPS Talcher STPS OBRA Thermal Faridabad Etxn Ramagundem B A.E.Co. Budge Budge Tanda Ennore Bokaro B Amar Kantak Ext Panki IP Station Tenughat Paras Harduaganj B Durgapur DPL Bandel Patratu Nellore New Cossipore Muzaffarpur Barauni

MP/Chattisgarh Tamil Nadu Rajasthan MP/Chattisgarh Maharastra Maharastra Tamil Nadu Tamil Nadu Maharastra UP AP MP/Chattisgarh MP/Chattisgarh Rajasthan Maharastra Tamil Nadu Delhi Gujarat Haryana AP AP MP/Chattisgarh Maharastra AP UP Gujarat UP West Bengal UP Gujarat Gujarat Punjab Karnataka Punjab Delhi West Bengal Maharastra UP MP/Chattisgarh Maharastra Punjab West Bengal Tamil Nadu Orissa MP/Chattisgarh Gujarat West Bengal UP Gujarat Orissa Bihar/Jharkhand West Bengal West Bengal Orissa UP Haryana AP Gujarat West Bengal UP Tamil Nadu Bihar/Jharkhand MP/Chattisgarh UP Delhi Bihar/Jharkhand Maharastra UP West Bengal West Bengal Bihar/Jharkhand AP West Bengal Bihar/Jharkhand Bihar/Jharkhand

92.9 92.6 91.3 91.2 91.0 90.8 90.5 90.4 90.4 90.0 90.0 89.9 89.8 89.6 89.3 89.2 89.1 88.9 88.9 88.7 88.4 88.3 88.2 88.0 87.8 87.7 87.6 87.3 87.1 86.8 86.2 85.7 85.4 85.2 85.0 84.7 84.7 83.9 83.7 83.4 83.0 82.9 82.7 82.4 81.5 81.5 80.7 80.5 79.9 79.6 78.8 77.1 77.1 76.9 74.7 74.7 74.5 74.2 73.8 72.3 71.7 68.1 66.2 66.1 66.1 63.9 63.3 53.3 52.4 52.2 50.7 43.0 35.8 30.1 27.7

21

Table A4 (continued ) Plant name

State

RTE Scores

Chandarpura (WB) Santaldih

West Bengal West Bengal

25.1 23.0

Table A5 Distribution of plants by TE (%) values.Source: Own estimations Range of TE (%)

Nr. of Plants

Percentage

Above 90 80–90 70–80 60–70 50–60 40–50 30–40 20–30 10–20 Below 10 Total

11 37 13 6 4 1 2 3 0 0 77

14.29 48.05 16.88 7.79 5.19 1.30 2.60 3.90 0.00 0.00 100.00

Table A6 Regional comparisons (1994–2010).Source: Own calculations from CEA (Govt. of India) data

MP/Chhattisgarh Uttar Pradesh Central Bihar/Jharkhand Orissa West Bengal East Delhi Haryana Punjab Rajasthan North Andhra Pradesh Karnataka Tamil Nadu South Gujarat Maharashtra West

Region

Mean Installed Capacity (MW)

Mean PLF (%)

Mean Power Generation (GWh)

RTIE RTE

Central Central

963.04 940.20 951.62 572.53 887.73 590.61 683.62 367.21 556.47 749.57 964.67 659.48 912.50 1198.24 835.78 982.17 533.15 863.27 698.21

71.46 64.33 67.90 33.34 72.89 55.86 54.03 60.80 57.62 75.20 84.68 69.57 78.55 78.64 70.79 75.99 65.39 71.52 68.46

6666.09 6016.78 6341.44 1936.00 5925.78 2883.71 3581.83 2200.10 3175.60 4943.93 7027.64 4336.82 6791.34 7831.61 5446.55 6689.83 3208.09 5388.59 4298.34

0.14 0.22 0.18 0.44 0.20 0.37 0.34 0.20 0.18 0.15 0.09 0.16 0.19 0.15 0.14 0.16 0.16 0.14 0.15

East East East North North North North South South South West West

0.86 0.78 0.82 0.56 0.80 0.63 0.66 0.80 0.82 0.85 0.91 0.84 0.81 0.85 0.86 0.84 0.84 0.86 0.85

the regulators are supposed to mature up and gain more expertize. Table 2 gives the descriptive statistics of the regulatory indicators and Table A2 (Appendix) shows their correlations. The mean tariff for the period 2000–01 to 2010–11 is 3.01 Rs./KWh with the minimum being 1.54 Rs./KWh in the state of Orissa and maximum being 4.68 Rs./KWh in the state of Delhi. The mean AT&C gains among the states has been 68.49% for the observed period with a minimum of 42.91% in the state of Orissa and maximum of 86.53% in Tamil Nadu. The average age of regulatory commissions from all states covered is 5.83 with the oldest being 13 years in the state of Orissa. Fig. A5 (Appendix) shows the age of regulatory commissions for different states in India. The indicator for unbundling takes the value 0 or 1 depending on whether a state has integrated or separate functions along the supply chain. Similarly, the indicator for regulatory composition takes the value 0 or 1 depending on whether there was a political shuffle in a given year.

22

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

4.3. Results The results from efficiency analysis are presented in this subsection. To start with, hypothesis tests were performed to find out the relative influence of explanatory factors. Wald tests were performed on the squared and interacted terms ( β6 − 20 ) in Eq. (6) to see if the translog specification is more appropriate than a Cobb– Douglas specification. Results, as can be seen in Table 3, confirm that the translog production function is more appropriate than the Cobb–Douglas production function. Wald test was also performed on the inefficiency effects explanatory variables (δ1 − 3). The result (see Table 3) confirms that inefficiency effects are associated with explanatory variables and hence justifies use of the translog specification. The SFPANEL function in STATA 11 (Belotti et al., 2012) was used to estimate the parameters of the specified translog production function (refer Eq. (6) and Eq. (7)). This allowed flexibility in robustness testing as the SFPANEL command allows for estimation of various fixed and random effects, time-varying and time-invariant models (Cornwell et al., 1990; Lee and Schmidt, 1993; Battese and Coelli, 1995). This also enables to see the robustness of the resulting coefficients. The results from the four models estimated are given in Table A3 (Appendix). For our analysis, we choose a maximum likelihood randomeffects time-varying inefficiency effects model, BC95, (Battese and Coelli, 1995) as it allows for a single stage estimation of the inefficiency parameters (See and Coelli, 2012). Variations of the BC95 model are estimated and the results are shown in Table 4. Model 1 has only plant age and installed capacity as determinants of the inefficiency effect and does not include regulatory index. The inclusion of regulatory index truncates the number of observations to 830 as index is computed for 2000–01 to 2010–11.14 Capital, Secondary Oil and Auxiliary Consumption have statistically significant coefficients and positive impact on generation. Coal does not appear to have significant impact on generation. In Model 2 when the index of state-level regulation was introduced, the coefficients for Coal also come out significant at 5% significance level. The coefficient for the regulatory index is significant at 10% level and the negative sign shows that better state-level regulation leads to a reduction in technical inefficiency. In the next models a sensitivity analysis was conducted to see the impacts of regulatory experience and unbundling. Table 5 shows weight assignment for different models. The base Model 2 has equal weights, 40% each, for tariff and AT&C gains. Ten percent weight is given to regulator’s age and 5% each for unbundling and regulatory composition. In Model 3 higher weight of 30% is assigned to the indicator ‘Age of Regulators’. The weights for tariffs and AT&C are 30% each. The coefficient for the regulatory index falls in value but is still significant. In Model 4 higher weight is given to the indicator ‘Unbundling’ in the index while keeping others same. The coefficient value falls further but remains highly significant. The result from Model 3 indicates that the experience of regulatory commission has a positive and significant influence on the performance of power plants. Model 4 indicates that if there is unbundling in a state where generation, transmission and distribution are separate functions, then also it has a very significant and positive influence on the performance of power plants. Plant capacity and plant age also turn out to be significant determinants of inefficiency. Increased plant capacity leads to reductions in inefficiencies suggesting that bigger plants are more efficient. There is an inverse relation with plant age which means as power plants 14 We also ran the full BC95 model without inclusion of regulatory index for the years before 2000. The results do not vary much and available from the authors upon request.

become older, their technical efficiency falls. The results for the Yearit variable (technical change) although has come up with the expected sign, is not significant in any of the models. Since Model 2 is the full model with the base regulatory index, its coefficients can be used for interpretation of results. The coefficients of Capital (1.628), Coal (2.7005), Secondary Oil (0.274) and Auxiliary (4.515) are greater than one showing increasing marginal returns. The sum of the four production coefficients (elasticities) is 9.1176 which show strong increasing returns to scale in electricity generation. Gamma (γ), the ratio of σv2 and σu2, the variance coefficients of random errors and inefficiency effects is 0.0241 which is greater than zero and indicates the presence of a random component of the technical inefficiency effects. As shown in Table 6 the Spearman correlations of the inefficiency scores for all the models are very high. This suggests that inclusion of the regulatory index although truncates the observations, does not affect the inefficiency rankings significantly. The summary statistics of technical inefficiency (TIE) scores is given in Table 7. Since the model estimations report the results of inefficiency, TIE, we need to calculate the technical efficiency (TE) score by subtracting the TIE from 1. Since Model 2 has been used for base results, the mean TIE is.23 which means the TE is 0.767 or 76.7%.15 Because Model 2 is timevarying for plant level inefficiency scores, the relative technical inefficiencies (RTIE) need to be calculated to get precise estimates for plant level performance. This is done by giving the most inefficient plant a score of 1 and normalizing the rest of the scores. To get the relative efficiency scores (RTE) the same procedure as above is adopted where the RTIE is subtracted from 1. Table A4 (Appendix) gives the plant level RTE scores. The next section discusses the results.

5. Discussion One objective of the present study has been to estimate the technical efficiencies of thermal power plants in India. Based on the SFA analysis, the most efficient plant is in Satpura in the state of Madhya Pradesh with an RTE of 92.9. The least efficient plant is Santaldih in West Bengal with RTE of 23.0. The distribution of thermal power plants by their technical efficiency values is given in Table A5 (Appendix). Around 14% of plants have efficiency scores higher than 90% whereas a majority 48% plants lie in the range of the score of 80-90%. As can be seen from Table 8 the best performing states in terms of thermal power efficiency are Rajasthan, Madhya Pradesh/Chhattisgarh16 and Maharashtra whereas the plants in states of Uttar Pradesh, West Bengal and Bihar/Jharkhand are the worst performing. The western region performs the best with a mean technical efficiency of 85% followed by the Southern and Northern regions at 84% (Table A6, Appendix). The mean efficiency of central region is 82% but the worst performing region is the Eastern region with a mean efficiency score of only 66%. Looking purely at the technical efficiency of Indian power plants, averaging at 76.7%, it is low by international standards and thus presents a substantial scope for efficiency improvements. The study also tries to make sense of the role of state-level independent regulation on the performance of power plants through its impact on the efficiency scores. The use of regulatory index, in fact, suggests that state-level regulators have positively 15 This is an improvement of approximately 4 percentage points from the mean technical efficiency of 72.66% from the figures of Shanmugam and Kulshreshtha (2005) which were up to the years 2001–02. 16 We have merged the states of Madhya Pradesh (MP) and Chhattisgarh together and Bihar and Jharkhand for our analysis. This is because Chhattisgarh and Jharkhand were carved out of MP and Bihar respectively in 2000.

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

impacted plant performance. This supports the hypothesis that better regulation ensures power plants get increased returns on investment through higher tariffs which lead to ramped up technical and managerial abilities. Reduction in technical losses also ensures higher revenue recovery and affects efficiency through increased marginal returns on machinery and base-load equipment investment. If regulatory composition changes frequently then it is an indicator of poor governance and will adversely affect all the regulatory outcomes like annual tariff determination. The sensitivity analysis with the indicators-unbundling and age of regulators-indicate that both have been effective in driving improvements in the generation sector. All of this implies that further reforms which empower independent regulators will have far reaching impacts on power sector performance.

6. Conclusions and policy implications Indian power sector has undergone substantive changes in the last two decades, most notable of which is the institution of independent regulation at the federal and state-level. It was expected to be a game changer as far as successful electricity provision is concerned. For the first time, its impact on an important indicator of performance – technical efficiency of thermal power plants – has been measured in this paper. This is done using an effective method, an updated and exhaustive data-set, and by construction of a unique index which captures the governance aspect of regulators. This provides the most current estimates of thermal power efficiency in India and the effect regulatory agencies have on it. Variations of a translog inefficiency effects model is estimated which allows single stage inclusion of factors influencing inefficiency. An index of state-level regulation is used as one of such factors. This is based on the premise of new institutional economics that without well-governed institutions, incentives would be ineffective. The index captures governance by regulators through indicators like tariff setting, reduction in technical losses, age, composition and the status of unbundling in the respective states. Results, based on the analysis of a balanced panel of 77 coal-based power plants from the period 1994–95 to 2010–11, indicate that the average technical efficiency of power plants is higher than those estimated by studies which considered a time period before independent regulation was introduced. Use of the composite index of state-level regulation to explain inefficiency suggests that regulators have a significant and positive influence on plant performance. There are some specific policy implications stemming from the results of efficiency analysis. The eastern region of India has been worst performing with an average RTE of 0.66 which means that the eastern states (especially Bihar and West Bengal) need to replace old capital through plant upgradation, move towards better quality coal use and speed up improvements in the transmission grid leading to lower distribution losses. Since regulatory index values also vary state-wise and regionally, specific efforts like faster unbundling are needed wherever state control in generation and distribution is still high. States which have lower performing generation utilities should grant more autonomy to the regulatory commissions through lesser political interference in the composition of members. Additionally, in line with the index used in our analysis, a system of benchmarking different regulatory bodies with respect to performance can be created. This way relative performance can be tracked and best practices can be emulated. The results from this analysis are encouraging because there has been an improvement in the performance in spite of all the bottlenecks and poor governance structure of Indian electricity regulation. Yet, there are significant gains to be achieved. This paper gives some more empirical insights into how institutional

23

quality, in the form of regulatory governance, is a significant determinant of utility performance. This is especially important to be studied in the context of developing and transition countries where the erstwhile state-controlled and inefficient sectors have been reformed in the last few decades.

Acknowledgments The authors would like to thank Jens Rommel, Christian Kimmich and two anonymous reviewers for useful comments. We also thank participants of the 19th Annual Conference of The International Society for New Institutional Economics (ISNIE), June 18–20, 2015 at Harvard Law School who gave valuable suggestions. We are thankful to Reema Priya for research assistance and grateful to Bitty Bathla and Sarwan Kumar for their guidance. Funding for this work came from the German Ministry of Education and Research (BMBF) and the German Academic Exchange (DAAD) under the 'Future Megacities' program.

Appendix A. Supplementary material Supplementary data associated with this article can be found in the online version at http://dx.doi.org/10.1016/j.enpol.2015.11.011.

Appendix See Tables A1–A6

References Abbott, M., 2006. The productivity and efficiency of the Australian electricity supply industry. Energy Econ. 28 (4), 444–454. Andres, L., Guasch, J.L., Azumendi, S.L., 2008.Regulatory governance and sector performance: methodology and evaluation for electricity distribution in Latin America. World Bank Policy Research Work Paper 4494. Bagdadioglu, N., Price, C.M.W., Weyman-jones, T.G., 1996. Efficiency and ownership in electricity distribution: a non-parametric model of the Turkish experience. Energy Econ. 18 (1), 1–23. Battese, G.E., Coelli, T.J., 1995. A Model for technical inefficiency effects in a stochastic frontier production function for panel data. Empir. Econ. 20 (2), 325–332. Belotti, F., Daidone, S., Ilardi, G., 2012. Stochastic frontier analysis using stata. Stata J. 55 (2), 1–39. Berg, S., Lin, C., Tsaplin, V., 2005. Regulation of state-owned and privatized utilities: ukraine electricity distribution company performance. J. Regul. Econ. 28 (3), 259–287. Bhattacharya, S.C., 2005. The electricity act 2003: will it transform the indian power sector? Util. Policy 13 (3), 260–272. Chikkatur, A.P., Sagar, A.D., Abhyankar, N., Sreekumar, N., 2007. Tariff-based incentives for improving coal-power-plant efficiencies in India. Energy Policy 35 (7), 3744–3758. Cornwell, C., Schmidt, P., Sickles, R.C., 1990. Production frontiers with cross-sectional and time-series variation in efficiency levels. J. Econ. 46 (1), 185–200. Cubbin, J., Stern, J., 2006. The impact of regulatory governance and privatization on electricity industry generation capacity in developing economies. World Bank. Econ. Rev. 20 (1), 115–141. Dhrymes, P.J., Kurz, M., 1964. Technology and scale in electricity generation. Econometrica: J. Econ. Soc. 32 (3), 287–315. Diewart, W.E., Nakamura, A.O., 1999. Benchmarking and the measurement of best practice efficiency: an electricity generation application. Can. J. Econ. 32 (2), 570–588. Dixit, A.K., 1996. The Making of Economic Policy. MIT Press, Cambridge, MA. Du, L., Mao, J., Shi, J., 2009. Assessing the impact of regulatory reforms on China’s electricity generation industry. Energy Policy 37 (2), 712–720. Dubash, N.K., 2008. Independent regulatory agencies: a theoretical review with reference to electricity and water in India. Econ. Polit. Wkly. 43 (40), 46–54. Dubash, N.K., Rao, D.N., 2008. Regulatory practice and politics: lessons from independent regulation in indian electricity. Util. Policy 16 (4), 321–331. Erdogdu, E., 2013. A cross-country analysis of electricity market reforms: potential contribution of new institutional economics. Energy Econ. 39, 239–251. Färe, R., Grosskopf, S., Logan, J., 1985. The relative performance of publicly-owned

24

R. Ghosh, V. Kathuria / Energy Policy 89 (2016) 11–24

and privately-owned electric utilities. J. Public Econ. 26 (1), 89–106. Färe, R., Grosskopf, S., Pasurka, C., 1986. Effects on relative efficiency in electric power generation due to environmental controls. Resour. Energy 8 (2), 167–184. Ghosh, R., Kathuria, V., 2014. The transaction costs driving captive power generation: evidence from India. Energy Policy 75, 179–188. Ghosh, R., Kathuria, V., 2015. Gas based power generation in India: lessons from a transaction cost analysis. Energy Strategy Rev. 8, 1–7. Goto, M., Tsutsui, M., 2008. Technical efficiency and impacts of deregulation: an analysis of three functions in us electric power utilities during the period from 1992 through 2000. Energy Econ. 30 (1), 15–38. Granderson, G., 2006. Externalities, efficiency, regulation, and productivity growth in the US electric utility industry. J. Product. Anal. 26 (3), 269–287. Graus, W.H.J., Voogt, M., Worrell., E., 2007. International comparison of energy efficiency of fossil power generation. Energy Policy 35 (7), 3936–3951. Hicks, D.A., 1997. The inequality-adjusted human development index: a constructive proposal. World Dev. 25, 1283–1298. Hiebert, L.D., 2002. The determinants of the cost efficiency of electric generating plants: a stochastic frontier approach. South. Econ. J. 68 (4), 935–946. Huang, C.J., Liu, J.-T., 1994. Estimation of a non-neutral stochastic frontier production function. J. Product. Anal. 5, 171–180. Jamasb, T., Pollitt, M., 2007. Incentive regulation of electricity distribution networks: lessons of experience from Britain. Energy Policy 35 (12), 6163–6187. Jaraitė, J., Maria, C.D., 2012. Efficiency, productivity and environmental policy: a case study of power generation in the EU. Energy Econ. 34 (5), 1557–1568. Joskow, P.L., Schmalensee, R., 1987. The performance of coal burning electric generating units in the United States: 1960–1980. J. Appl. Econom. 2 (2), 85–109. Jung, Y., Lee, S.-H., 2014. Electrification and productivity growth in Korean manufacturing plants. Energy Econ.45, 333–339. Khanna, M., Mundra, K., Ullah, A., 1999. Parametric and semi-parametric estimation of the effect of firm attributes on efficiency: the electricity generating industry in India. J. Int. Trade Econ. Dev.: Int. Comp. Rev. 8 (4), 419–430. Khanna, M., Zilberman, D., 1999. Barriers to energy-efficiency in electricity generation in India. Energy J. 20 (1), 25–41. Klugman, J., Rodríguez, F., Choi, H.-J., 2011. The HDI 2010: New controversies, old critiques. J. Econ. Inequal. 9 (2), 249–288. Knittel, C.R., 2002. Alternative regulatory methods and firm efficiency: stochastic frontier evidence from the US electricity industry. Rev. Econ. Stat. 84 (3), 530–540. Komiya, R., 1962. Technical change and production function in the united states steam power industry. Rev. Econ. Stat. 44, 156–161. Kopp, R.J., Smith, V.K., 1980. Frontier production function estimates for steam electric generation: a comparative analysis. South Econ. J. 46 (4), 1049–1059. Kumbhakar, S.C., Ghosh, S., Mcguckin, J.T., 1991. A generalized production frontier approach for estimating determinants of inefficiency in US dairy farms. J. Bus Econ. Stat. 9, 279–286. Kumbhakar, S.C., Lovell, C.A.K., 2000. Stochastic Frontier Analysis. Cambridge University Press, New York, USA. Laffont, J.J., Tirole, J., 1993. A Theory of Incentives in Procurement and Regulation. MIT Press, Cambridge, MA. Lam, P.-L., Shiu, A., 2004. Efficiency and productivity of China's thermal power generation. Rev. Ind. Organ. 24 (1), 73–93. Lee, Y.H., Schmidt, P., 1993. A Production Frontier Model with Flexible Temporal Variation in Technical Efficiency. The Measurement of Productive Efficiency: Techniques and Applications. Oxford University Press, United Kingdom. Levy, B., Spiller, P.T., 1994. The institutional foundations of regulatory commitment: a comparative analysis of telecommunications regulation. J. Law Econ. Organ. 10 (2), 201–246. Loeb, M., Magat, W.A., 1979. A decentralized method for utility regulation. J. Law Econ. 22, 399–404. Malik, K, Cropper, M.L., Limonov, A., Singh, A., 2011. Estimating the Impact of Restructuring on Electricity Generation Efficiency: The Case of the Indian Thermal Power Sector National Bureau of Economic Research (NBER) Working Paper No. 17383.

Maruyama, N., Eckelman, M.J., 2009. Long-term trends of electric efficiencies in electricity generation in developing countries. Energy Policy 37 (5), 1678–1686. Meibodi, A.E. 1998. Efficiency Considerations in the Electricity Supply Industry: The Case of Iran. Surrey Energy Economics Centre (SEEC) Discussion Paper 95, School of Economics, University of Surrey. Mester, L.J., 1993. Efficiency in the savings and loan industry. J. Bank. Financ. 17, 267–286. Mester, L.J., 1997. Measuring efficiency at US banks: accounting for heterogeneity is important. Eur. J. Oper. Res. 98, 230–242. Nakhooda, S., Dixit, S., Dubash, N.K., 2007. Empowering People: A Governance Analysis of Electricity – India, Indonesia, Philippines, Thailand. Washington, D. C. World Resources Institute & Prayas Energy Group. Noll, R.G., 1989. Economic Perspectives on the Politics of Regulation. In: SCHMALENSEE, R.A.W., ROBERT, D. (Eds.), Handbook of Industrial Organization II. North, D.C., 1990. Institutions, Institutional Change and Economic Performance. Cambridge University Press. North, D.C., 1993. Institutions and credible commitment. J. Institut. Theor. Econ. 149 (1), 11–23. Olatubi, W.O., Dismukes, D.E., 2000. A data envelopment analysis of the levels and determinants of coal-fired electric power generation performance. Util. Policy 9 (2), 47–59. Pitt, M.M., Lee, L.-F., 1981. The measurement and sources of technical inefficiency in the Indonesian weaving industry. J. Dev. Econ. 9, 43–64. Pollitt, M.G., 1995. Ownership and Performance in Electric Utilities: The International Evidence on Privatization and Efficiency. Oxford University Press, Oxford. Reifschneider, D., Stevenson, R., 1991. Systematic departures from the frontier: a framework for the analysis of firm inefficiency. Int. Econ. Rev., 715–723. Sarica, K., Or, I., 2007. Efficiency assessment of Turkish power plants using data envelopment analysis. Energy Policy 32 (8), 1484–1499. See, K.F., Coelli, T., 2012. An analysis of factors that influence the technical efficiency of malaysian thermal power plants. Energy Econ. 34 (3), 677–685. Sen, A., Jamasb, T., 2010. Diversity in unity: an empirical analysis of electricity deregulation in Indian States. Energy J. 33 (1), 83–130. Seo, D., Shin, J., 2011. The impact of incentive regulation on productivity in the US telecommunications industry: a stochastic frontier approach. Inf. Econ. Policy 23 (1), 3–11. Shanmugam, K.R., Kulshreshtha, P., 2005. Efficiency analysis of coal-based thermal power generation in india during post-reform era. Int. J. Global Energy Issues 23 (1), 15–28. Shrivastava, N., Sharma, S., Chauhan, K., 2012. Efficiency assessment and benchmarking of thermal power plants in India. Energy Policy 40, 159–176, January 2012. Simar, L., Lovell, C.K., Eeckaut, P.V., 1994. Stochastic frontiers incorporating exogenous influences on efficiency. Discuss. Pap., 9403. Singh, A., 2006. Power sector reform in India: current issues and prospects. Energy Policy 34 (16), 2480–2490. Singh, J., 1991. Plant size and technical efficiency in the thermal power industry. Indian Econ. Rev. 26 (2), 239–252. Spiller, P., Tommasi, M., 2005. The institutions of regulation: an application to public utilities. In: MENARD, C., SHIRLEY, M.M. (Eds.), Handbook of New Institutional Economics. Springer, Netherlands. Stern, J., Holder, S., 1999. Regulatory governance: criteria for assessing the performance of regulatory systems. an application to infrastructure industires in the developing countries of Asia. Util. Policy 8 (1), 33–50. Sueyoshi, T., Goto, M., 2011. DEA approach for unified efficiency measurement: assessment of Japanese fossil fuel power generation. Energy Econ. 33 (2), 292–303. Teri, 2007. Analysis and Compilation of Tariff Orders, Project Report No. 2006 RP23. TERI and Ministry of Power, Government of India. Williamson, O.E., 2005. The economics of governance. Am. Econ. Rev. 95 (2), 1–18. Zhang, Y.-F., Parker, D., Kirkpatrick, C., 2008. Electricity sector reform in developing countries: an econometric assessment of the effects of privatization, competition and regulation. J. Regul. Econ. 33 (2), 159–178.