The determinants of electricity theft: An empirical analysis of Indian states

Energy Policy 93 (2016) 127–136

Contents lists available at ScienceDirect

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

The determinants of electricity theft: An empirical analysis of Indian states Vasundhara Gaur, Eshita Gupta n Department of Policy Studies, TERI University, India

H I G H L I G H T S








Over 20% of total electricity generated in India is lost to thefts. The study attempts to identify the determinants of electricity theft in India. Use of panel data from 2005 to 2009 for 28 Indian states. FGLS and OLS regression results are compared. The determinants of power theft are both governmental and socio-economic in nature.

art ic l e i nf o

a b s t r a c t

Article history: Received 1 October 2015 Received in revised form 23 February 2016 Accepted 25 February 2016

More than 20% of the electricity generated in India is lost to rampant thefts. Drawing data from 28 states of India over a time span of five years (2005–2009), this paper examines the role played by socio-economic and governance factors in determining the extent of electricity thefts in Indian states. Results from the Feasible Generalised Least Squares (FGLS) model demonstrate that lesser corruption, higher state tax to GDP ratio, greater collection efficiency of electricity bills by state utilities, higher share of private installed capacity, lesser poverty, greater literacy and greater income are closely associated with lesser power thefts. A better understanding of the key determinants of thefts in electricity distribution is vital for policy makers for designing policies. & 2016 Elsevier Ltd. All rights reserved.

Keywords: Electricity theft T&D losses Indian states

1. Introduction India's demand for electricity is growing fast and the supply is unable to keep pace with the growth in demand resulting in the demand–supply gap of 3.6% in 2014–15 (Load Generation Balance Report, 2015–16). On July 31st, 2012, India experienced the largest power outage in the history (so far), when the entire northern grid collapsed and plunged most of north India in darkness, affecting around 700 million people. Why does India have these massive power shortages? It is increasingly being recognised by academicians and practitioners that the lack of domestic resources, poor infrastructure and poor governance are the key factors responsible for the continued problem. More than 30% of total electricity generated in India is lost due to thefts and inefficiencies in transmission and distribution (Sekhar, 2014). A better understanding of the key determinants of thefts in electricity n

Corresponding author. E-mail addresses: [email protected] (V. Gaur), [email protected] (E. Gupta). http://dx.doi.org/10.1016/j.enpol.2016.02.048 0301-4215/& 2016 Elsevier Ltd. All rights reserved.

distribution is vital for policy makers to address the challenge of reducing the size of the existing power shortages. Electricity can be stolen through several techniques. In India, it is not uncommon to find consumers drawing electricity by hooking a wire to utility poles. Since the wire is not connected to a metre and the consumption is unrecorded, this constitutes to theft. Even when electricity metres are in place, consumers often commit fraud and tamper with them using magnets to show consumption being less than it would be otherwise. Finally, consumers either default completely on their bills or bribe the utility employees to record the metre at a lower value than it shows, both of which lead to billing irregularities the ultimate effect of which is power thefts. (Smith, 2004; Depuru et al., 2011) Several studies analysing the relationship between electricity theft and its determinants have used transmission and distribution (T&D) losses as an indicator of power thefts for a given economy (Smith, 2004; Min and Golden, 2014). The T&D losses have two components: technical losses and commercial (or non-technical) losses. The technical component comprises of inevitable losses due to energy dissipation when electricity is transmitted across long distances but they are intrinsic to the system and can be brought

128

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136 35

30

T&D Losses (% of output)

25

20

15

10

5

0

Brazil

China

India

Low & middle income

Mexico

Russia

Fig. 1. T&D loss comparison across other nations (World Development Indicators, World Bank, 2014).

down to optimal levels. However, it is the component of commercial losses (including electricity theft through pilferage and other fraudulent practices) that constitutes a major part of the total T&D losses and which goes unaccounted for (Smith, 2004). Unlike in developed economies, where the technical losses range between 2% and 4% of the total electricity output, in developing countries these losses can reach up to as high as 10–12% (Smith, 2004). However, developing countries report T&D losses to be far higher than this figure. In India itself, even if we assume the supply system to be of the least efficient kind, the average T&D losses are almost two times than that, implying that 20% of these losses are on account of power thefts. As can be seen from Fig. 1, even when compared to other low and middle income countries, India's T&D losses have historically been very high. For the year 2009, the state-wise aggregate T&D losses and profits (as a percentage of revenue) of power utilities are displayed in Fig. 2. While the all India aggregate T&D losses are 32.23%, there are states with T&D losses as high as twice that amount. It is clearly visible that the states with the highest T&D losses (specifically Bihar, Arunachal Pradesh, Mizoram, Jammu and Kashmir, Nagaland and Manipur) are also the ones whose power utilities are under the most financial turmoil with the lowest profits (highest losses). When consumers steal electricity, they are in effect drawing more than the allocated power which causes an overload at the generation unit. This causes the electricity supply to not only be sporadic but also fraught with frequent voltage fluctuations. The distribution companies (DISCOMs), unable to generate enough revenues as a result of this theft, face financial imbalances, which further reduces their capability to invest in better infrastructure and manpower to check pilferage. Thus begins the vicious theft-loss spiral. According to a study by World Bank, power theft reduces India's Gross Domestic Product (GDP) by around 1.5% (Bhatia and Gulati, 2004). Kochhar et al. (2006) also found a strong, negative association existing between the economic growth of Indian states and their respective power sector's transmission and distribution losses. The primary query of this paper centres on the role played by different factors in determining the extent of electricity thefts in Indian states. There have been a few empirical studies to identify the key determinants and evaluate their impact on electricity thefts. A significant part of the literature points out that poor governance factors are the key determinants of power thefts in

addition to socio-economic factors (Min and Golden, 2014; Smith 2004; Steadman, 2011). In an attempt to find a relation between populism and power theft, Min and Golden (2014) undertook a study in the state of Uttar Pradesh in India and found a concrete relation between electricity theft and electoral cycles by demonstrating how theft rates increased just prior to a new election term. This paper contributes to the existing literature significantly. This is the first study on power thefts that has been conducted at the all-India level. The underlying motivation of the study stems from the fact that there exists a considerable variation among the states in terms of electricity thefts. The degree to which this discrepancy is explained by the differences in socio-economic and governance indicators of the states needs to be explored. One would envisage that as socio-economic and governance factors improve the extent of electricity thefts decrease. How far this phenomenon is reflected in the Indian scenario needs to be empirically tested. Results from the Feasible Generalized Least Squares (FGLS) model demonstrate that lesser corruption, higher state tax to GDP ratio, greater collection efficiency of electricity bills by state utilities, higher share of private installed capacity, lesser poverty, greater literacy and greater income are closely associated with lesser power thefts. These results are also compared with the estimates obtained from an OLS cluster-robust regression. The next section is dedicated to the discussion of existing literature and the methodological approach taken to study the determinants of power thefts in India. The results of the same are analysed in depth in the third section. The fourth and final section summarizes these findings and concludes with policy recommendations that may play a role in tackling this issue.

2. What factors determine power thefts? There has been a considerable amount of research examining the relationship been electricity thefts and its key determinants. Studies have found corruption to be a crucial determinant of power theft in developing countries like India and Pakistan (Jamil and Ahmad, 2013; Katiyar, 2013). Furthermore, poor governance was linked to corruption through political instability, lack of accountability and bribery (Smith, 2004; Steadman, 2011). Similarly, good governance was associated to a state's enforcement capacity

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

Puducherry

129

-7.23

11.84

Goa

16.99

1.83

West Bengal

18.33

2.94

Andhra Pradesh

18.37

2.43

Tamil Nadu

18.41

Chattisgarh

18.62

Karnataka

18.76

1.57

Kerala

19.59

3.75

Himachal Pradesh

20.52

-59.95 -5.74

-5.01

Delhi

22.09

Jharkhand

22.24

Gujarat

22.77

Punjab

23.39

7.95 -36.50 1.61 -14.36

Maharashtra

25.16

Uttarakhand

25.27

-2.21 -26.71

Rajasthan

29.99

Haryana

31.00

-8.34 -16.00

Assam

32.82

Uttar Pradesh

33.15

-21.68 -58.66

Tripura

35.55

Orissa

37.00

Madhya Pradesh

38.32

Sikkim

39.01

-12.94

Meghalaya

39.06

-11.90

Bihar

0.85 -9.02 -46.52

43.58

Arunachal Pradesh

-72.19

48.04

Mizoram

53.80

Manipur

54.66

Nagaland

-270.44 -206.29 -92.87

56.91

Jammu & Kashmir

67.35 T&D Losses (% of output)

-119.13 -234.85 Profits (as % of Revenue)

Fig. 2. State-wise profits and T&D losses of utilities for 2009 (All India Electricity Statistics, 2011; Report on the Performance of State Power Utilities for the years 2009–10 to 2011–12, Power Finance Corporation (PFC)).

by Gümüşdere (2004) through the tax to GDP ratio. Tax to GDP ratio is a good indicator of governance to the extent that it would indicate people's honesty with respect to paying their taxes and also the effectiveness of the respective state governments in collecting these taxes through proper enforcement mechanisms for ensuring public compliance. While Marangoz (2013) finds no evidence that links the illegal use of electricity to political variables like political party in Turkey, studies in India have found a positive relation between populism and power theft (Min and Golden, 2014; Kochhar et al., 2006).

Some studies also attempted to explore the effect of socioeconomic variables like household income, unemployment, poverty, illiteracy and high rates of murder on power thefts (Gümüşdere, 2004; Steadman, 2011; Marangoz, 2013). Further research in this context revealed a link between electricity theft and economic sectors, particularly agriculture. Golden and Min (2012) in their study in Uttar Pradesh observed electricity theft to be greater in zones with more agricultural users, mostly on account of the elite farmers who easily had their electricity bills reduced by influencing politicians. However, in the other sectors

130

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

(domestic, commercial and industry), they found the trend of line losses to be either flat or decreasing. Viswanathan (2014) attributes the higher theft rates in the agricultural sector to the unmetered supply of electricity to the consumers in those areas. The electricity tariff in the agricultural sector is based on the power of the pump sets that the consumers declare that they are using (measured in horsepower per unit). This declaration also decides the power load that will be sanctioned, which is generally low based on the disclosures. However, after this proclamation, the consumers are free to increase the connected load far above the level that had been authorised. In effect, the consumers draw more power than had been licensed to them which eventually show up as T&D losses. The impact of electricity prices on the theft of power has also been extensively studied in various contexts and a positive relation between the two was observed (Jamil and Ahmad, 2013; Katiyar, 2013). Thus, this study also examines the association between power theft and socio-economic indicators like poverty, unemployment, literacy rate, per-capita income, electricity prices and share of agricultural and industrial sectors. While modelling electricity theft in Turkey, Gümüşdere (2004) additionally considered the distribution utilities′ managerial variables and physical variables like transformer utilisation ratio and distribution line length as an indicator of investment towards infrastructure. The better the infrastructure, the lesser will be the loss of power during transmission and distribution. A summary of the literature discussed is presented in Appendix.

3. Data and methodology The study comprises of 28 Indian states. The state of Sikkim and all union territories were excluded due to data limitations. Our data is annual and spans over five years starting from 2005 to 2009, thus giving us five observations per state. The dependent variable is electricity thefts, measured by transmission and distribution losses, as obtained from Central Electricity Authority (CEA). While power theft cannot be calculated, it can certainly be estimated. In India, the most commonly used measure of power losses is Transmission and Distribution (T&D) losses which is a measure of losses incurred while transmitting electricity between sources of supply and points of distribution and in the distribution to consumers, including pilferage.1 However, this measure is not an indicator for thefts alone, since a component of it includes genuine technical losses which inevitably occur while transmitting electricity over long distances. So far, no simple method exists that can be used to reliably decompose the losses into its components of technical losses and stolen electricity. However, Golden and Min (2012) still use T&D losses as a measure of power theft as the line losses stemming from India's inefficient systems was about 12 percent and the component of theft was greater than fifty percent of total T&D losses. Therefore, following the previous literature this study will also use T&D losses as an indicator of power theft in various states of India. The explanatory variables are grouped in to two categories: 1) Governance variables and 2) Socio-economic variables. Following the literature, we have first considered governance indicators in order to capture factors such as each state's enforcement capacity, their effectiveness in collecting taxes and bills and the rule of law. Tax-GDP ratio is considered as an indicator of enforcement, the data for which was taken from the Planning Commission. This study additionally considers the collection efficiency of power 1 The losses are formally calculated by subtracting the energy billed to the consumer from the total energy input.

utilities as an important indicator of the governance and management at the utility level. The better the utility is at collecting revenue, lesser will be the rate at which consumers default on their bill payment which will consequently lead to lower power loss. This collection efficiency is simply the percentage of the electricity bill that the utilities are able to recover. Since this data is available at the level of utilities, an average value for all the utilities in a particular state was considered to arrive at the statelevel measure. In order to capture the effect of governance through corruption, the number of cases registered under the prevention of corruption and related acts was taken into account using data from the National Crime Records Bureau (NCRB). Given India's history of populism and the findings of Min and Golden (2014), a dummy variable was also taken into consideration as an indicator of whether the political party at the centre was the same as the one at the state level. In order to remain in power in a particular state and with no opposition from the centre, it is expected that politicians will be more susceptible to being bribed for cheaper electricity in exchange for votes than politicians from other parties who are concerned about scrutiny from the centre (Parsai, 2015; Zargar, 2015). The socio-economic indicators considered for the model are electricity prices, per-capita income, urbanisation, poverty, literacy rate, rate of urban unemployment, structure of the economy, infrastructural investment and total population. State population is included as a control variable. Electricity prices were accounted for in this study using the study of Gupta (2016) as a source of data for state-wise aggregate electricity prices.2 In our study, we take electricity prices as being exogenously determined since electricity prices in India are regulated and not determined through the supply and demand mechanism. While higher electricity thefts can lead to higher electricity prices, it is not so in the case of India since electricity prices are not market determined. Urbanisation is taken as the share of urban population to the total state population. Literacy, urbanisation and population data is available in the census of India while the data for urban unemployment rate is taken from the National Sample Survey Organisation (NSSO).3 The data for per-capita income was taken from the Central Statistics Organisation (CSO) while the data on poverty4 was extracted from the Planning Commission of the Government of India. The study also considers the impact of the structure of the economy on the level of power thefts. The number of energised pump sets in a state is taken as an indicator for the dependence of a state on its agricultural sector since the primary use of electricity in agriculture is to power tube wells (or pump sets) for the purpose of irrigation. The extent of industrialisation of a state's economy is captured by the industrial share which is the percentage of state GDP that is constituted by the industrial sector. The data for industry share and the number of energised pump sets was obtained from the Reserve Bank of India (RBI) and the General Review of the CEA5 respectively. The level of investment in infrastructure was incorporated into the study through variables like the average cost of electricity supply, capital employed, private generation, total capacity, households electrified, line length and number of transformers. While the average cost of electricity supply would be an excellent 2 Gupta (2016) calculated state-wise aggregate electricity prices by finding the average electricity prices in the domestic, commercial, agricultural and industrial sector, using the share of the sale of electricity in each sector as weights. 3 The NSSO describes urban unemployment as the number of people actively looking for a job in urban areas. 4 Poverty is described here as the percentage of population below the poverty line. 5 From CEA's publication: All India Electricity Statistics (General Review).

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

indicator of the financial health of the power sector, the amount of capital employed, number of transformers and total line length would capture the level of investments made towards providing better power infrastructures and facilities. The total installed capacity in each state is also considered as a measure of electricity generation capability. Finally, in order to capture how effective the private sector is in curbing power theft, private generation capacity is taken into account by the electricity generation capacity by private players as a percentage of the total power generation capacity in a state. The data on average cost of supply and capital employed was obtained from the PFC report. The data on the number of distribution6 transformers, total line length, total installed capacity (in megawatts) and private generation capacity was obtained from the General Review of the CEA. The number of electrified households was considered as a reflection of the reach of electricity to the respective state's population. The lesser the households with electricity, the more will their neighbours attempt to steal from them in order to get electricity for themselves. This data was extracted from the census of India.7 3.1. Dataset The summary statistics for states as a whole for the estimation period is reported in Table 1. The average value of T&D losses all over India between the years 2005–2009 is 34.07% which is extremely high even by the standards of low and middle income countries. In 2009, the country-wide average T&D loss was 32.23%, which was the lowest among all years considered in the sample. The average T&D losses have displayed a decline from their peak of 37.28% in 2005 to 34.30% in 2006, 33.16% in 2007 and 33.37% in 2008, finally reaching their trough in 2009. A high standard deviation of 13.12 is reflective of immense variation across states. While the minimum value of 5.89% is comparable to some developed countries, the maximum value of 67.35% is extremely worrisome at the least. With regard to T&D losses, across states the worst performer is Jammu and Kashmir followed by North Eastern states such as Nagaland and Manipur. On the other hand, Kerala emerged as the best performer. The governance indicators similarly display extreme variations across states. The tax-to GDP ratio varies from lower than 2% in Manipur and Nagaland to over 11% in Karnataka and Uttaranchal. It has remained mostly stable in all states across years, increasing or decreasing by only a few points every year. The north-eastern states like state Arunachal Pradesh, Assam, Manipur, Meghalaya and Nagaland along with Goa also report the least cases of corruption while Tamil Nadu and Maharashtra report the highest number. Collection efficiency is the worst at 28.91% in Bihar and the best at 100% in many of the smaller states including Goa, Delhi, Mizoram and Tripura. Fig. 3 plots the governance indicators across T&D losses of 30 regions (29 states and Pondicherry) for the year 2009. The socio-economic indicators effectively capture the cultural, economic and social diversity of India with almost all the population literate in Kerala to barely half in Bihar. Bihar also performs poorly in other indicators of development with the lowest per capita income, over half of its population living below the poverty line and having only around a tenth of its households electrified. The average price of electricity in rupees per unit displays high 6 Distribution transformers have been considered since they are the ones that distribute power to the users (industrial and domestic) at lower voltages and end user connectivity. Power transformers were not considered since they are used for transmission purposes at high voltages, with no thefts occurring at that level. 7 The data is available as the percentage of houses using electricity as their primary source of lighting in a state.

131

variation across states on account of different policies regarding energy, regulation and agricultural subsidies. Goa is the only state that can be credited to have the entirety of its electricity generation under private players. It was also observed that the states with no private generation capacity like Jammu and Kashmir, Bihar, Manipur, Meghalaya, Mizoram and Nagaland are also the ones which have the highest T&D losses. Gujarat has the lowest rate of urban unemployment (less than 2%) whereas Tripura has over a quarter of its urban population unemployed. Some socio-economic indicators are plotted across 30 regions (29 states and Pondicherry) for the year 2009 in Fig. 4.

4. Results and discussion This study estimates the Feasible Generalized Least Squares (FGLS) regression model. While the Fixed Effects (FE) is an ideal model to use when the study is done across particular countries or the states within countries (as each has its own peculiarities and idiosyncrasies that is unique to them), it was dropped from consideration as it disregards the impact of time-invariant variables. Therefore, any variables that remain constant or are slow to change across time will be dropped from the model. This study contains several variables such as poverty rate, literacy rate, urbanisation and tax-GDP ratio that are almost constant over the five years of the study period such as. Dropping so many variables would be infeasible and hence, the FE model was not considered for this study and the FGLS method was used. Also, unlike the fixed effects method that uses an intercept term to capture unobserved heterogeneity, the FGLS method assumes it to be captured in the errors. We have estimated two FGLS regression models – model A and model B. Model A includes all the explanatory variables. Model B looks at the effect of dropping poverty from model A. Results from both the FGLS models are displayed in Table 2 with all governance indicators and most socio-economic variables turning out to be significant. The reported Wald chi-square statistic is an indicator of the overall goodness of fit of both the models. The p-value of 0 implies that all the coefficients taken together in the model are statistically significant. The regression results of the FGLS model A and model B are compared with the results obtained from a pooled OLS regression with cluster-robust standard errors (to account for heteroskedasticity) in model C. The R-squared for the OLS regression is 0.80. Of the governance indicators (in model A and B), the coefficient of tax to GDP ratio indicated a very substantial impact of the tax to GDP ratio on T&D losses. On an average, an increase in the tax to GDP ratio by 1% decreases T&D losses by about 1.2%. This is indicative of the importance of the citizens' honesty in revealing their true incomes and paying their obligations as well as the power of the government to enforce compliance and encourage such behaviour of honesty and cooperation. Gümüşdere (2004) also found the tax to GDP ratio to be an important determinant of power thefts in Turkey. The positive and significant coefficient on corruption implies that corruption has a positive impact on power thefts. This finding is consistent not only with the results of Katiyar (2013), Min and Golden (2014) and Smith (2004) but also the common perception and theory. Corruption here is not limited to populism and the pandering of politicians for votes but extends to the employees of utilities who accept bribes that lead to non-payment of electricity bills and as a result, higher losses. The size of the corruption coefficient is much smaller than the size of the tax to GDP ratio indicating that one unit lower corruption has relatively lesser positive impact on power thefts as compared to the one unit higher tax to GDP ratio. Also, the results from the pooled OLS

132

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

Table 1 State-level summary statistics. Variables

Observations

Mean

Standard deviation

Minimum

Maximum

T&D loss (%) Electricity price (Rs./kWh) Per capita income (Rs.) Poverty (%) Literacy (%) Population Unemployment (%) Urbanization (%) Industry share (%) Energised pumpsets Tax-GDP ratio ACS (Rs./kWh) Collection efficiency (%) Private capacity (MW) Total capacity (MW) Capital employed (Rs. crore) Corruption Households Electrified (%) Line length (circuit kms.) Total transformers INC

140 140 140 140 140 140 140 140 140 140 140 140 138 136 140 140 140 140 140 140 140

33.68 2.80 35,757.63 27.78 73.28 4.04×107 5.40 31.37 15.39 556,135 5.87 2.67 93.32 15.57 3,395.22 4,001.69 116.89 68.83 3.02 137,127.80 0.46

13 0.72 20,387.28 12 8.82 4.23×107 4.32 16.93 8.49 803,322 2.25 0.87 12.81 22.25 3,873.77 3,580.57 137.69 22.24 4.39 174,355.10 0.50

13.16 1.54 7,914 8.70 53.73 9.65×105 1.20 10.02 2.48 0 1.49 0.85 28.91 0 0 129 0 12.74 0.17 114 0

67.35 4.67 135,966 54.22 93.30 1.92×108 28 96.79 38.50 3,115,630 11.62 5.44 137.27 99.94 16,061.81 14,794 498 97.86 28.77 729,967 1

regression indicate corruption to be insignificant (in model C). The dummy INC indicates a significant, positive and a surprisingly substantial impact on power thefts. The coefficient of about 3.5 in both models (A and B) suggests that on an average, states where the ruling political party is the same as the one that is at the centre have T&D losses that are higher by nearly 3.5 units than those states where the ruling party is different. This is a clear cut indication of the breeding populism so rampant in India with corrupt and venal politicians encouraging the theft of electricity or even the diversion of power to their own constituencies. By indicating what percent of the bill the electricity utilities are able to recover, collection efficiency captures governance at the utility level. The results indicate that an increase in the collection efficiency by 1% will reduce T&D losses by about 0.2 units. Thus, greater the effectiveness of the utilities in collecting bills, lesser will be the thefts. It follows that in the presence of high vigilance and strictness in collections, losses can be reduced. Regarding socio-economic indicators, the coefficient of poverty in model A implies that as the percentage of population living below the poverty line increases by 1 unit, T&D losses will increase by nearly 0.33 units. However, since poverty and per capita income are related, the results may be erroneous on account of a high correlation between the two variables. Indeed, when poverty was dropped, per capita income became significant (at 10%) with a negative sign (as shown in model B). This is indicative of the fact that as the states become richer, power theft decreases. Similarly, the variable for households electrified was significant at 5%. However, the coefficient for households electrified was insignificant in model B. This may be due to the high correlation between poverty and households electrified. Intuitively, it is expected that if a state is poor, it will have a greater percentage of the population living without electricity. The coefficient on industrial share turns out to be significant and negative. It implies that everything else remaining constant, if the state economy becomes 1% more industrialised, T&D losses will decrease by nearly 0.61 units on an average. However, the negative sign of energised pumpsets indicated that power thefts should be less as an economy became more agrarian, which is contrary to the results obtained by Golden and Min (2012). This unexpected result may be on account of an unforeseen correlation between the variables. Indeed, when poverty was omitted (model B), the variable became insignificant.

The coefficient of electricity price turns out to be insignificant. An increase in the literacy rate or urbanisation rate by 1% would decrease T&D losses by 0.6 units. The insignificance of unemployment is in line with the results of the study conducted by Marangoz (2013) in Turkey and by Steadman (2011) in Jamaica. It is however, significant (at 10%) and positive in model B (in which poverty was omitted), indicating a correlation between poverty and unemployment. It can be inferred (from model B), that as urban unemployment increases by 1%, electricity theft also increases by 0.23 units, thus implying that joblessness also has an impact on power theft. Among the infrastructural variables, the significant variables (at 1%) were private capacity and line length. At the same time, capital employed, total capacity and average cost of supply remained robustly insignificant. The significance of the variable for private capacity (as an indicator of the presence of private players) is reflective of its importance as a determinant of power thefts. Since the private sector is driven by the profit motive, it will always attempt to cut losses as much as possible, wherever it is possible. According to 2003–04 estimates (BSES Limited First Half Results – FY2003-04, 2003) by Reliance Energy, decreasing T&D losses only in the Mumbai Supply area by just 1% will increase their profits by nearly Rs. 28 crore per annum. The prospect of capturing such high profits is probably what drives the private players to tackle high T&D losses with fervour. The variable for transmission line length is more of an indicator of overall line losses during transmission and distribution than power thefts. Its significance implies that as the line length increases by one circuit kilometre (ckt. km.), T&D losses decrease by 2 units on an average. The length of distribution lines affects power loss in two opposite ways. First, as the length of lines increases, it not only increases the distance electricity has to be travel to reach consumers but also the resistance it faces in the wire. As a result of this resistance, the electric wires get heated up and some power is lost as heat energy, thus increasing power losses with an increase in line length. On the other hand, an increase in the number of lines would cause the current to be distributed across these many lines and at the same time, reduce the burden on any one line to provide to the population, leading to lesser power losses. Since the deterring effect of resistance is linear while that of the current is

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

Pondicherry

12

7

Goa 4

7

Tamil Nadu

8

Chhatisgarh

7

Karnataka Kerala

8 5

Delhi

6

6

Punjab

6

Maharashtra

7

Rajasthan

6

Haryana

6

Nagaland Jammu and Kashmir

45

163

95

27

100

23

99

25 2

128 96

31 3

96 48

33

88

1

96

100 86

38 8 0

69 77

44

94 7

48

100

54 0

100

55

2

9 57

1

67

6

T&D Loss (% of output)

Tax-GDP Ratio

129

98

6

2

375

96

39

2

336

95

39

5

414

95

30

37

4

162

94

36

4

183

99

23

8

Bihar

Manipur

22

6

Meghalaya

198

98

3

Sikkim

Mizoram

20

33

Madhya Pradesh

395

91

6

Orissa

100

19

4

Tripura

Arunachal Pradesh

49

25

5

Uttar Pradesh

498

96

22

Gujarat

322

98 18

95

4

Assam

94

18

21

Himachal Pradesh

Uttaranchal

100

0

19 9

94

0

18

Andhra Pradesh

Jharkhand

4 17

7

West Bengal

133

83 61 76

Corruption

Collection Efficiency

Fig. 3. Government indicators for Indian states for 2009 (All India Electricity Statistics, 2011; Report on the Performance of State Power Utilities for the years 2009–10 to 2011–12, Power Finance Corporation; Databook for PC, 2014).

quadratic,8 the latter effect dominates the former one, leading to decreased losses with increased line length overall (Gümüşdere, 2004). Of all the variables considered, most were found to be significant and important determinants of power thefts in India. Among those variables, there were some like literacy rate, urbanisation, share of industry, percentage of private capacity, INC and governance indicators like tax-GDP ratio and collection efficiency 8

By definition, the power (P) in an electrical circuit is generated by a source of (V) volts carrying current (I), giving the relation P ¼VI. From Ohm's Law we get V ¼ IR, where I is the current in a circuit, R is the resistance of the conductor and V is the potential difference across the ends of the circuit. Substituting the Ohm's Law expression into the first identity, we get P ¼I2R.

that were robustly significant (with expected and unchanging signs) irrespective of the change in variables in the model and even across the two models (FGLS and OLS). On the other hand, the coefficients on ACS, population and unemployment remained robustly insignificant in all models.

5. Conclusion and policy implications The objective of the study is to find out the state level determinants of power thefts in India. This objective is achieved through conducting an econometric study which led to results that were both expected and surprising. We find significant negative impact of good governance

134

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

10

Jammu and Kashmir Nagaland

3

57 7

Manipur Mizoram

3 9

Arunachal Pradesh Bihar

67

12

Sikkim

18

Madhya Pradesh

17

Orissa

18

Tripura

0

13

Rajasthan

17

39

0

38

216

52 4294 2546

0

148

0

5484

863

21 23

Gujarat

23

Jharkhand

22

400

28

16 19

Chhatisgarh

19

Tamil Nadu

16 18 14 11

Goa

27

360

12

1684

0

58

1056

386 196

2351 9118

2893 30

1800

3719 5945

1362

18 33 36

T&D loss (% of output)

11632 10818

3129

18

17

11416

5514

20

Karnataka

16062

5139

58

19 21 8

1785 4731

22

Himachal Pradesh

5733

25

Punjab

6

3580 1031

30

21

Maharashtra

25 8

31

21 25

Uttaranchal

Industry share

189

471

18

Pondicherry

585

0 0

33

Haryana

West Bengal

83

0

33

12

Andhra Pradesh

80

39

36

Assam

Kerala

54

37

5

Uttar Pradesh

Delhi

51 0

44

Meghalaya

31

0

55

48

6

1093

0

7375

78 78

33

0

Total Capacity

Private Capacity

Fig. 4. Socio-economic indicators for Indian states for 2009 (All India Electricity Statistics, 2011; Handbook of Statistics on Indian Economy, Reserve Bank of India).

indicators on power thefts. Given the high relevance of the indicators of management at the utility level and of governance indicators at a broader level, the necessity of having policies aimed at achieving lower corruption, high compliance and an overall better management reveals itself. In this regard, regular auditing and increased or surprise inspections may go a long way in ensuring a degree of transparency and honesty so crucial to avoiding losses and improving collective efficiency. In Gujarat in 2005, power shortages were tackled by helping local DISCOMs repair their finances (Wilkes, 2014). Doing so may enable utilities to use

their funds for a better and more efficient management of the distribution system. The results of this study also indicate T&D losses to be decreasing in the presence of private players. Thus, a push towards privatisation of electricity transmission and distribution is offered as a suggestion in order to achieve lower levels of line losses in the system. The deregulation of the electricity sector reduced the distribution losses by half in seven years in Chile and only three years in Argentina (Rudnick and Zolezzi, 2001). Beyond the selfish aim of private players of reducing losses by maximising their own

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

Table 2 Regression results for Feasible Generalized Least Squares Modela. Dependent variable

Model A

Model B

Model C

Electricity price Per capita income (PCY) Poverty Literacy Population

0.046 (0.731)  0.00003 (0.00005) 0.325*** (0.040)  0.590*** (0.114)  1.25  10  8 (2.14  10  8) 0.099 (0.123) 0.557*** (0.063)  0.608*** (0.107)  3.56  10  6* (1.91  10  6)  1.166** (0.467)  0.744 (0.752)  0.150*** (0.021)

 0.379 (0.576)  0.0001* (0.0005) –  0.631*** (0.120)  4.51  10  9 (2.24  10  8) 0.232 (0.126) 0.576*** (0.070)  0.333** (0.109)  2.42  10  6 (2.10  10  6)  1.189** (0.494)  0.922 (0.885)  0.165*** (0.052)

 0.254 (1.498)  7.46  10  5 (0.0001) 0.324*** (0.084)  0.744*** (0.187)  1.57  10  8 (  3.18  10  8) 0.269 (0.222) 0.511*** (0.103)  0.538** (0.216) 3.19  10  6 (4.50  10  6)  0.949 (0.830)  0.160 (1.531)  0.142** (0.065)

 0.160*** (0.035)  0.0002 (0.0003)  0.0008 (0.0002) 0.008* (0.005) 0.114** (0.049)

 0.143*** (0.037)  0.0003 (0.0004)  0.0002 (0.0002) 0.011** (0.005) 0.042 (0.049)

 0.217** (0.073)  0.0003 (0.0009)

Unemployment Urbanisation Industry share Energised pumpsets Tax to GDP ratio ACS Collective efficiency Private capacity Total capacity Capital employed Corruption Households electrified Line length Total transformers

 0.0003 (0.0004) 0.009 (0.011) 0.221** (0.078)

INC Intercept

 1.928*** (0.453)  1.871*** (0.486)  1.746** (0.734)  5.33  10  6  9.52  10  6*  1.41  10  5 (4.96  10  6) (4.98  10  6) (9.82  10  6) 3.626*** (0.814) 3.362*** (1.024) 3.840* (2.181) 83.286*** (7.076) 101.226*** (8.980) 82.621** (12.541)

Wald chi2 (20) Wald chi2 (19) R-squared Prob4chi2 Prob4F Observations

1766.75 – – 0.0000 – 134

– 704.96 – 0.0000 – 134

– – 0.8016 – 0.0000 134

134 observations as against 140 due to missing data. a

Standard errors in parenthesis. Po 0.10. ** Po 0.05. *** P o0.01. *

135

profits, privatisation may have an additional benefit of breaking the link between pandering politicians and corruption in the power sector. Central to the strategy for reducing the theft of electricity are the electricity prices. What is required, are power tariffs that are rationalised and tailored according to the socioeconomic conditions and the economic structure of specific regions. The importance of socio-economic factors in determining power thefts, and by association, the overall health of the power sector has already been discussed at length. Being literate is not enough-the citizens need to be educated about the disadvantages of thieving power and the enormous toll it takes on the economy. Above all, greater efforts must be made to brand power thefts as a crime and punishable offence rather than a socially acceptable norm. For instance, K-Electric (Formerly: Karachi Electric Supply Company Ltd.) obtained a fatwa (or decree) from several Islamic scholars declaring electricity theft to be a sin after deducing it costs several millions of dollars annually (Aziz, 2009). Improvements in the already existing electricity infrastructure along with the use of smart metres would ensure an increased and a better quality of electricity supply that might be successful in discouraging consumers from stealing if they can be promised a continuous supply. The Accelerated Power Development and Reforms Programme (APDRP) scheme that was initiated in 2002–03 involved, among other things, the renovation and modernisation of electricity distribution systems. As a result, it was able to bring down the average line losses in India from their peak value of 28% (See Fig. 1). These results point at mainly top-down approaches to curbing power thefts while literature abounds with conclusions that point to more bottom-up solutions. This may be due to the broader scope of this study that has been undertaken for an entire country, using its various states and union territories as the smallest crosssection. A study focused at a more micro-level involving individual preferences may be more successful in this direction. Throughout the study, there is one overarching implication that echoes on all accounts: that the issue of power theft is more complicated and multifaceted than it may appear at first blush. The interdependence and correlation of its determinants may be a

Table A1 Summary of key studies on power thefts. Study

Variables

Methodology

Key results

Smith (2004)

T&D Losses and government indicators

102 countries 1980–2000.

Gümüşdere (2004)

Theft-loss ratio (TLR), Theft Loss per capita (TLPC), economic variables, variables reflecting enforcement capacity and reach of state, state and authority related variables, distribution utility's managerial variables and physical variables. Baseline losses

81 provinces of Turkey. 1994–200. 1Panel data analysis using Feasible Generalised Least Squares (FGLS).

Electricity theft is highly correlated with low governance estimates and corruption. TLR and TLPC are different although they are correlated.Some variables that affect TLR and TLPC are factors that can be controlled by the distribution utilities, whereas some others are not.

Katiyar (2005)

Steadman (2010)

Distribution loss, proportion of male population, unemployment rate, GDP growth rate.

Golden and Min (2012)

Line losses, election variables, criminal charges, population and HDI

Marangoz (2013)

Illegal Electricity Use, per-capita income, illiteracy rate, unemployment rate, population size, political party, geographic region and the occurrence of terrorist events.

Jamil and Ahmad (2013) (working paper)

Distribution losses, per-capita income, electricity prices, probability of detection, fine per incidence, load shedding and temperature

South Rajasthan. Mapping of two 11 kV feeders.

Power theft mainly due to high entry costs, collusion and an environment where stealing electricity is considered the norm. Jamaica. Time-series data between Electricity Theft is not associated with economic 1998–2008. factors such as unemployment and economic growth. Uttar Pradesh (India). 2000–2009. Electricity theft is highly correlated with electoral OLS Estimation. cycles. It is greater in zones with more agricultural users. 67 provinces in Turkey. 2009. OLS Illegal electricity usage is not linked to economic Estimation. factors such as unemployment rate and income. Illiteracy and terrorist attacks are positively related to illegal electric usage. No evidence which proves the relationship between electricity theft and political party. 9 electricity distribution companies Electricity price has a positive impact on power in Pakistan. Dynamic panel between thefts while per-capita income and fine on convic1988–2010 using fixed effects model tion have a negative impact.

136

V. Gaur, E. Gupta / Energy Policy 93 (2016) 127–136

contributing factor for the same. What is therefore required is a more comprehensive, integrated and cohesive intervention system that is as multi-faceted as the problem it wishes to tackle.

Appendix A See Table A1.

References All India Electricity Statistics, 2011. Central Electricity Authority, Ministry of Power. Government of India, New Delhi. Aziz, F., 2009. Pakistani clerics condemn theft of electricity. Reuters. [Online]. Available: 〈http://in.reuters.com/article/2009/07/13/pakistan-power-decreeidINISL13847820090713〉. Bhatia, B., Gulati, M., 2004. Reforming the power sector: controlling electricity theft and improving revenue. The World Bank, Washington, DC [Online]. Available: 〈http://rru.worldbank.org/documents/publicpolicyjournal/272bhatia_Gulati. pdfS〉. BSES Limited First Half Results – FY2003–04, 2003. Presentation. [Online]. Available: 〈http://www.rinfra.com/newsmedia/pdf/BSESH1Presentation15Oct03. pdf〉. Databook for PC, 2014. The Planning Commission, New Delhi. [Online]. Available: 〈http://planningcommission.nic.in/data/datatable/0814/table_166.pdf〉. Depuru, S.S.S.R., Wang, L., Devabhaktuni, V., 2011. Electricity theft: overview, issues, prevention and a smart meter based approach to control theft. Energy Policy 39, 1007–1015. http://dx.doi.org/10.1016/j.enpol.2010.11.037. Golden, M., Min., B., 2012. Theft and Loss of Electricity in an Indian State. Working Paper, Version 20. Gümüşdere, E., 2004. Theft and Losses in Turkish Electricity Sector: Empirical Analysis and Implications for Tariff Design. Sabanci University, Istanbul. Gupta, E., 2016. The Effect of Development on the Climate Sensitivity of Electricity Demand in India. Climate Change Economics, Vol.7, No. 2. World Scientific Publishing Company. http://dx.doi.org/10.1142/S2010007816500032. Handbook of Statistics on Indian Economy, Reserve Bank of India. [Online]. Available: 〈https://rbi.org.in/Scripts/AnnualPublications.aspx? head ¼Handbook þ of þStatistics þon þ Indian þ Economy〉. Jamil, F., Ahmad, E., 2013. An empirical study of electricity theft from electricity distribution companies in Pakistan. 29th annual general meeting and

conference, Pakistan Society of Development Economists, PIDE, Islamabad. Katiyar, S.K., 2013. Political economy of electricity theft in rural areas theft. Econ. Polit. Wkly. 40, 644–648. Kochhar, K., Kumar, U., Rajan, R., Subramanian, A., Tokatlidis, I., 2006. India's pattern of development: what happened, what follows? [Online]. Available: 〈https:// www.imf.org/external/pubs/ft/wp/2006/wp0622.pdf〉 doi:10.1016/j. jmoneco.2006.05.007. Load Generation Balance Report 2015–16. Central Electricity Authority, Government of India [Online]. Available: 〈http://www.cea.nic.in/reports/yearly/lgbr_ report.pdf〉. Marangoz, C., 2013. Illegal electricity use in Turkey: causes and policy implications. Vanderbilt University, Nashville. Min, B., Golden, M., 2014. Electoral cycles in electricity losses in India. Energy Policy 65, 619–625. http://dx.doi.org/10.1016/j.enpol.2013.09.060. Parsai, G., 2015. Kejriwal opposes same party rule in state, centre. The Hindu. [Online]. Available: 〈http://in.reuters.com/article/2009/07/13/pakistan-powerdecree-idINISL13847820090713〉. Report on the Performance of State Power Utilities for the years 2009–10 to 2011– 12, Power Finance Corporation, Ministry of Power. [Online]. Available: 〈http:// www.pfcindia.com/writereaddata/userfiles/file/ResearchReport/Report%20on% 20the%20Performance%20of%20State%20Power%20Utilites%20for%20the% 20years%202009-10%20to%202011-12.pdf〉. Rudnick, H., Zolezzi, J., 2001. Electric sector deregulation and restructuring in Latin America: lessons to be learnt and possible ways forward. IEE Proc. – Gener. Transm. Distrib. 148, 180. http://dx.doi.org/10.1049/ip-gtd:20010230. Sekhar, P., 2014. Secured Techno-Economic Growth of India: Unleashing Hidden Growth Potential. Micro Media Marketing Pvt. Ltd., Mumbai. Smith, T.B., 2004. Electricity theft: a comparative analysis. Energy Policy 32, 2067–2076. http://dx.doi.org/10.1016/S0301-4215(03)00182-4. Steadman, K.U., 2011. Essays on Electricity Theft. State University of New York, Binghamton. Viswanathan, P.K., 2014. Trends and determinants of agricultural power consumption and distribution losses: an analysis of agricultural feeders across power distribution companies in Gujarat. Ahmedabad. Wilkes, T., 2014. India to invest $4 billion to tackle power theft. Reuters. [Online]. Available: 〈http://in.reuters.com/article/2014/11/21/india-electricityidINKCN0J50F620141121〉. World Development Indicators, World Bank, 2014. Electric power transmission and distribution losses (% of output). [Online]. Available: 〈http://data.worldbank. org/indicator/EG.ELC.LOSS.ZS〉. Zargar, S., 2015. Survey Finds AAP Government Less Corrupt Than Departments Under Centre In Delhi. [Online]. Available: 〈http://www.scoopwhoop.com/ news/delhi-corruption-study/〉.