Académique Documents
Professionnel Documents
Culture Documents
a r t i c l e
i n f o
Article history:
Received 8 August 2011
Revised 23 December 2011
Accepted 23 December 2011
Available online 26 January 2012
Keywords:
Kerosene
Developing countries
Climate change mitigation
India
Energy subsidies
Distributional impacts
a b s t r a c t
Kerosene subsidies intended for Indian households have been known for their poor targeting and high scal
costs. However, the distributional benets to the 160 million households that use kerosene are not well understood. In this paper, the kerosene subsidy is formally assessed as an instrument of income redistribution.
The subsidy incidence, progressivity and efcacy of the kerosene subsidy are calculated for the state of Maharashtra, under actual and ideal implementation conditions.
The analysis shows that kerosene subsidies are regressive and of minimal nancial value to poor rural households. This is in part because household quotas are based on cooking needs, but kerosene is used predominantly for lighting. In urban areas, subsidies are progressive, and provide benets of up to 5 to 10% of
household expenditure among poorer households which lack affordable access to LPG and biomass. Overall,
only 26% of the total subsidy value directly reaches households. This analysis suggests that subsidies targeted
only to kerosene-dependent urban areas would have a higher efcacy than broad-based subsidies.
2012 International Energy Initiative. Published by Elsevier Inc. All rights reserved.
Introduction
Energy consumption subsidies have attracted renewed attention
with the urgency of climate change (UNEP, 2008). Globally, annual
consumption subsidies for electricity and fossil fuels outside the
OECD were estimated at about a quarter trillion dollars in 2005, or
0.7% of world GDP (IEA, 2007). These subsidies are often justied as
instruments of redistribution in many developing countries, in part
because of the lack of broad-based institutions that enable direct
cash transfers (Piketty and Qian, 2009). However, these subsidies in
many developing countries incur high costs, encourage wasteful consumption and increase greenhouse gas emissions without necessarily
providing intended benets to the poor. The International Energy
Agency (IEA) estimates that the cost to economic growth from consumption subsidies in just the eight largest non-OECD countries is
257 billion dollars (IEA, 1999). Among household cooking fuels,
while LPG subsidies typically benet higher income households, the
distribution of benets from kerosene subsidies are more ambiguous
(Heltberg et al., 2003). This is part due to kerosene's usage among
urban low-income groups, such as in India, Nepal, Vietnam and
South Africa.
According to the IEA, India is among the highest of the non-OECD
subsidizers of energy consumption, with subsidies of over $10 billion
per year, despite undertaking price reform of fossil fuels in the last decade. With the recent run-up in world crude oil prices since 2005, the
cost of subsidies more than doubled (Shenoy, 2010). Kerosene use also
0973-0826/$ see front matter 2012 International Energy Initiative. Published by Elsevier Inc. All rights reserved.
doi:10.1016/j.esd.2011.12.007
36
provide income benets to the poor, and that kerosene use tends to
rst increase and then decrease with income.
The high scal and environmental costs of household fuel subsidies in India have been well documented in the above cited studies.
Economists estimate that kerosene subsidies, in particular, cost
the government $46 billion per year, and carry efciency losses 5 of
$12 billion per year (IEA, 1999; Morris et al., 2006). It is well
known that suppliers and distributors divert 4066% of kerosene allocated for states to other lucrative markets such as transportation, 6
forcing some households to purchase kerosene in the black market
(NCAER, 2005; Shelar et al., 2007). But the nancial impacts of
the subsidies that do reach households are not formally estimated.
One study of fuel taxation in India nds that the direct income benets from household kerosene use are progressive (Datta, 2010). However, the study offers limited detail on underlying regional differences
and causes. Further, progressivity alone may not justify public spending on kerosene subsidies. Subsidies may be regressive but still
provide substantial income relief to the poor. Conversely, even if kerosene subsidies are progressive, other less progressive redistributive
policies may have higher impacts on poverty and cost less to deliver.
The efcacy measure used in this study provides a metric for such a
comparison across policies.
Many studies illustrate the inappropriateness of encouraging kerosene use for lighting, but no study critiques the subsidy for cooking
use. Many studies have pointed out that kerosene lamps are the
most expensive and inefcient form of lighting (Dutt and Mills,
1994; Mahapatra et al., 2009; Reddy, 1981). Rehman, Malhotra et al.
(2005) have pointed out that rural kerosene quotas are based on
cooking but used only for lighting. However, subsidy quotas may be
better suited to urban kerosene use for cooking. In this study, I assess
the contribution of design and implementation failures to targeting
efcacy of kerosene subsidies in urban and rural areas.
This study relies on data from the National Sample Survey of India
for Consumption Expenditure, 200405 (NSSO0405) and draws qualitative insights from a primary survey I had conducted of 450 households in urban and peri-urban parts of Maharashtra in 2009/2010
(primary survey).
In Section 2, I describe the various uses for kerosene in different
household in Maharashtra, and how these functions inuence the
subsidy's relative benets. In Section 3, I discuss the measurement
approach and the results of the subsidy performance analysis. In
Section 4 I discuss the subsidy performance under ideal implementation conditions. I close with some policy recommendations in
Section 5.
Previous literature
See Maharashtra Cracks Down on Oil Maa, Hindu Business Line, January 28, 2011.
Kerosene that is diverted to automobile and other markets have income distributional impacts as well, such as the rents earned by middlemen, or the marginal prots
earned by beneciaries of adulterated fuel. Data on these diversions are unavailable.
4
5
Efciency losses refer to reduced aggregate income from the misallocation of investments in response to price signals that do not reect actual costs. In economic jargon, this is calculated as the dead weight losses that result from pricing below
marginal cost.
6
In transportation, kerosene is used to adulterate gasoline and diesel. In construction, kerosene is used in making tar for paving roads.
37
a) Rural
Fig. 1. Variation in urban household kerosene use Mumbai, Maharashtra and India
200405.
b) Urban
10
See Appendix A for how the household quota was inferred from sample survey
data.
11
Theoretically, one would expect that black market prices would be set by the prices
of fuels that they substitute in other markets, namely diesel. In reality, black market
prices are lower, and vary by region, due to market imperfections. Data on black market prices are available in NSSO0405 in household surveys as the price of other kerosene purchases made outside the PDS.
38
Table 1
Kerosene consumption by market: Maharashtra 200405.
Table 2
Household kerosene use by function, region and priority in Maharashtra.
Monthly,
million liters
HH quota
total
HH PDS
purchase
HH black market
purchase
Diversions to
other markets
Rural
Urban
Totals
80
38
118
25
15
40
12
16
28
42
7
50
Source for PDS and black market purchases: National Sample Survey Consumption
Expenditure, 20045.
See Appendix A for details on quota estimation.
Numbers don't add due to rounding.
a) Subsidy Prices
Source: National Sample Survey 2004-05. Sales prices shown for all retail sellers in the state.
Million users
Lighting
Cooking
Primary
Secondarya
kerosene than lighting users, but a subset of them might also pay a different replacement price in the absence of subsidies, namely LPG prices.
The distribution across income groups of kerosene use as a
cooking fuel also varies between urban and rural areas (Fig. 4).
These patterns are explained by their relative reliance on LPG,
which is more closely tied to kerosene usage. In rural areas, kerosene
use increases with income, since LPG use increases with income.
Among poorer households, kerosene is used as a backup to wood,
albeit to a lesser extent, when wood is unavailable (such as during
monsoon), or too expensive (where purchased).
In contrast, in urban areas, such secondary use is highest among
lower and middle income groups, but then decreases with income
at the highest income levels, as more households have two LPG cylinders or piped gas supply.
39
Fig. 5. Kerosene vs. LPG delivered fuel cost comparison (200405 prices).
Note: Cooking use assessed as household use above 4 liters per month.
Source: National Sample Survey of Consumption Expenditure, 2004-05
month for a family of seven. 16 This is less than a half a percent of poor
households monthly expenditure. Such households may overestimate these savings, or consider even such small savings worthwhile.
Households that cook with black-market kerosene pay approximately the same fuel price as LPG (about Rs 22/kg on average in
urban Maharashtra), which implies they pay more on an energy basis
than they do to cook with LPG (since kerosene stoves are less efcient).
This suggests that LPG users who buy black market kerosene for cooking
must also be Last Resort Users who lack reliable access to LPG.
In summary, Economy Users may represent a minority of kerosene
users. However, the nancial risk to them of removing subsidies is
signicantly less than others, given the thin margins of savings from
using subsidized kerosene over LPG. To the extent possible, differentiating these households would be important in calculating the distribution of subsidy benets.
Assessment of subsidy performance
In this section, I discuss the merits of kerosene subsidies as a redistributive policy based on three measures: materiality; progressivity;
and efcacy. I rst present the measurement approach and metrics.
I then discuss the results and the drivers of low efcacy.
Measurement approach
I rst present the method for calculating individual household
benets, in terms of budget share, based on the categorization of
kerosene users discussed in the previous section. Based on these
budget shares, I estimate progressivity and efcacy.
Several approaches have been used to assess the distributional
benets of subsidies to households in other developing countries. A
common approach is to compare energy budgets with and without
the subsidy (Dube, 2003; Kebede, 2006) for different types of households. I use a similar approach to estimating materiality. Dube (2003)
compares households' willingness to pay to the price they would pay
without subsidies. However, this carries a hypothetical bias. Instead I
use market prices for substitutes to evaluate budget impacts. Saboohi
(2001) calculates Lorenz curves for energy subsidy benets to assess
progressivity. Olivia and Gibson (2008) use an aggregate welfare
measure that incorporates an inequality aversion parameter. Both
approaches capture equity in terms of inequality, but not necessarily
16
This estimate is based on each family member bathing every day, each using 10 liters of hot water heated from 10 degrees C to 70 degrees C, and with savings of Rs 0.25
per megajoule of delivered energy from using kerosene instead of LPG.
40
Subsidy progressivity
Progressivity is an important metric for policies that have a wide
impact. A poverty index, for example, would not provide an indication of the relative benets between the poor and non-poor. Thus, a
policy that reduces the poverty gap can be regressive if it benets
middle income groups to a greater extent. Such a policy could be an
expensive instrument of redistribution.
I adapt a measure of pro-poor growth used in the development
economics literature, which uses a Growth Incidence Curve (GIC)
to measure the extent to which income growth accrues to the poor
relative to what accrues to those above poverty, as dened by some
poverty threshold (Grosse et al., 2008).
Specically, a variation of the GIC is constructed with the average
income percentage change for every population centile in order of increasing income. If the slope of this curve is increasing (decreasing),
the subsidy provides greater (lesser) benets to higher income
groups on an individual household basis. In aggregate, the average
percentage change in income for all centiles (H) below the threshold
(p) is compared to the average percentage change () for the entire
population (n). The former would be higher for a progressive policy.
H
1X
p
I =I
H i1 i i
1X
i
I =I
n n i i
I
HH
CP Subsidy :Q AggQuota
Materiality
Among those spending less than $2/day in rural and urban areas, the
average income relief from the kerosene subsidy amounts to 0.3% and
0.5% of household expenditure respectively. However, the benets
vary widely by income centiles, from 0 to 1.3% of household expenditure. Further, almost 24% and 47% of the rural and urban keroseneusing population respectively receives benets that exceed 1% of total
expenditure (Fig. 6). With an elasticity of 0.25, these gures drop to
10 and 33%. Savings are highest in metropolitan urban areas, such as
in Mumbai, and in remote districts with limited LPG supply.
For instance, for the poor in District 1 a remote district with 1.3
million people the savings on average are 1.7% of their monthly
expenses, and over 5% for those earning less than $1/day. Considering
that the poorest urban households spend over 10% of their income on
energy (World Bank, 2003), this could double their energy budgets in
the worst case.
Among LPG (higher income) users, Last Resort users have an average savings of 0.5%, while Economy Users have savings of 0.4%, with
the reduction being higher at upper middle income levels.
18
The studies use cross-sectional data (and therefore estimate long-term elasticity)
for the entire country, do not differentiate cooking from lighting, and capture the effects of substitution, rather than just conservation.
19
Figures show median and one standard deviation, for studies from 31 countries
and 57 separate estimates of elasticities for electricity and from 18 countries and 155
separate estimates of elasticities for water.
41
a) Urban
b) Rural
This nding has important implications for the relationship between
LPG and kerosene policies. The more liquid and reliable the LPG market,
the less upper income households would use kerosene. This would improve the progressivity of kerosene subsidies, simply by altering market
incentives. Note that this nding is different from the conventional wisdom that LPG pricing alone would inuence kerosene use.
Progressivity
Fig. 7 below shows the cumulative average savings as a share of
household expenditure. Several observations stem from the gures.
First, in rural areas, the subsidy benets are regressive. The cumulative average savings rate (Eqs. (3) and (4)) is 0.37% for those who
earn under $1/day, 0.41% for those who earn below $2/day.
However, in urban areas, the subsidy is consistently progressive
for the entire population, using the same thresholds. 20 The average
income relief is 0.58% for lower income groups, and 0.53% for middle
income groups.
Policy efcacy
The efcacy of the kerosene subsidy (Eq. (5)) from the perspective
of all households is at best ~26.5%. That is, for every 100 Rupees of
subsidy, only Rs 26.5 of income relief is delivered to households
directly. The efcacy for delivering income relief to those earning
under $2/day is 17%. Thus, for every rupee of income transferred to
these poor, six rupees has to be spent by the government.
The low efcacy reects the fact that only 31.3% of the kerosene picked up by wholesalers was delivered to households through the PDS in
200405, based on NSSO0405. The difference of ~5 percentage points,
between the quantity (31.3) and benet (26.5) shortfalls, is attributable
to price discrimination. That is, households pay prices that include actual transport costs and rents in addition to the wholesale subsidy price.
Kerosene subsidies and ideal implementation
The shortfall in the subsidy value that reaches households can be attributed to design failures and implementation failures. For policy reformists, this distinction would be important. Even under ideal
implementation conditions, a kerosene subsidy can only benet
kerosene users up to the use of their quota. The mismatch between demand and the quota reects the limits of the value of the subsidy even
under ideal implementation conditions (Fig. 8). The correlation between
household quotas and their usage was found to be only 22% in rural
areas, and 58% in urban areas, indicating that the quota design is better
suited to urban kerosene needs. Implementation failures, on the other
hand, are reected in the fact that households purchase part of their entitled kerosene quotas in the black market, or that they fail to obtain the
intended subsidy price for the part of the quota they do obtain. Indeed,
20
Note that prices in urban areas are signicantly higher, so the same cutoff reects
greater poverty than in rural areas. Data were unavailable to create price-adjusted poverty thresholds.
Design failures
Most poor households do not claim their entitled share of subsidized kerosene, as reected in the gap between the average quota
and total usage (Fig. 8). This is the case in rural areas across all income
groups. As mentioned earlier, most rural households cook with wood.
Since the government allocates quotas based on cooking needs, but
rural households purchase PDS kerosene mostly for lighting, rural
households forego most of their quota. In urban households and particularly metropolitan areas on the other hand, kerosene's use for
cooking aligns with the quota, except for where low-income households have access to wood, as discussed earlier (Note that in Maharashtra I estimate cooking demand to account for only 21% of
kerosene demand in rural areas, but 61% in urban areas, which is
slightly higher than the national averages shown earlier).
At the same time, about a fth of households in the lowest three
deciles in urban areas buy kerosene in excess of their quotas
(Fig. 9). This heterogeneity among the urban poor stems from varying
access to wood, as reected in the fact that those who forego their
quotas consume on average about four times the quantity of wood
as those that consumer above their quotas. In rural areas, because
kerosene use increases with income, the share of households whose
kerosene purchases exceed their quota also increases with income.
Thus, the discrepancy between allocated quotas and demand
reinforces the difference in kerosene benets between urban and
rural areas. The progressivity of subsidies would likely increase in
urban areas from better targeted, and potentially higher, quotas for kerosene subsidies. In rural areas, on the other hand, the extent of unused
quotas only provides incentives for their diversion to other sectors.
42
Policy implications
Note: Data show households that purchase any amount of kerosene in excess of their quotas.
Results summary
LPG/ non-LPG
Area
1.5 to 2
2
Persons
1 Person
2 Persons
3 Persons
4 and more persons
MumbaiThane 1 Person
rationing area
2 Persons
3 Persons
4 Persons
5 Persons
6 Persons
7 and more persons
LPG connection All areas
1 Cylinder
holder
2 Cylinder
Non LPG
connection
holder
Rural
Urban
Municipal
corporations
& A class
municipalities
Quantum
[liters per person]
Maximum
(l)
15
15
24
Liters
3
8
10
3 per person
5
24
12l
15
18
21
23
24
4
Nil
Source: Maharashtra Food, Civil Supplies and Consumer Protection Department. (http://www.maharashtra.gov.in/english/food/
schemesKerosene.php).
Methodology for Determining Household Quotas from NSSO Data
Households' entitlements were calculated based on fuel use data in
the NSSO0405. Households that did not use LPG were identied by the
absence of LPG consumption in the survey. The NSSO0405 does not indicate how many cylinders households own. Households with two
LPG cylinders were assumed to be those households that did not use
any backup fuel. All other households that used LPG were assumed
to have one cylinder. Data on household size and location were used
to determine the nal kerosene entitlement (as dened above).
References
Chaturvedi BK. Report of the High Powered Committee on Financial Position of Oil
Companies. Petroleum Ministry, Government of India; 2008.
Datta A. The incidence of fuel taxation in India. Energy Econ 2010;32:S2633.
Dube I. Impact of energy subsidies on energy consumption and supply in Zimbabwe.
Do the urban poor really benet? Energy Policy 2003;31:163545.
Dutt GS, Mills E. Illumination and sustainable development Part II: Implementing lighting efciency programs. Energy Sustain Dev 1994;1:1727.
Gangopadyaya S, Ramaswami B, Wadhwa W. Reducing subsidies on household fuels in
India: how will it affect the Poor? Energy Policy 2005;33:232636.
Grosse M, Harttgen K, Klasen S. Measuring pro-poor growth in non-income dimensions. World Dev 2008;36:102147.
Gundimeda H, Khlin G. Fuel demand elasticities for energy and environmental
policies: Indian sample survey evidence. Energy Econ 2008;30:51746.
Gupta G, Khlin G. Preferences for domestic fuel: Analysis with socio-economic factors
and rankings in Kolkata, India. Ecol Econ 2006;57:10721.
Heltberg, R. Household energy use in developing countries: a multicountry study. In:
ESMAP Technical Paper, editor. Washington DC: The World Bank; 2003.
Hosier RH, Kipondya W. Urban household energy use in Tanzania: Prices, substitutes
and poverty. Energy Policy 1993;21:45473.
IEA. Looking at energy subsidies: Getting the prices right. World Energy Outlook.
Washington DC: International Energy Agency; 1999.
43