Content uploaded by Anne Olivier
Author content
All content in this area was uploaded by Anne Olivier on Feb 29, 2020
Content may be subject to copyright.
I
Anne Olivier
Caterina Ruggeri Laderchi
GOOD PRACTICE NOTE 3
Analyzing the Incidence of Consumer
Price Subsidies and the Impact of Reform
on Households — Quantitative Analysis
HOUSEHOLDS
Public Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure AuthorizedPublic Disclosure Authorized
I
CONTENTS
Acknowledgments iii
About the Authors iv
Acronyms and Abbreviations v
1. Introduction 1
2. The Process of Conducting Distributional Analysis 5
3. A Quantitative Assessment of Distributional Impacts: Key Questions and How to
Approach Them 6
How Large Are Energy Subsidies and Who Is Benefiting from Them? 7
Who Is Going to Be Aected by the Removal of Energy Subsidies and—More Specifically—
Would Poverty Increase Significantly? 8
How Much Would It Cost to Compensate Vulnerable Groups? 10
4. Dierent Methods to Estimate Household-Level Welfare Impacts of Energy
Subsidy Reform 11
Partial Equilibrium Analysis 11
General Equilibrium Eects 12
5. Doing Distributional Analysis in Practice 13
Setting Up the Analysis and Complementary Data Needs 13
Assessing the Distributional Impact of the Direct Eects of Energy Subsidy Reform .....13
Assessing the Distributional Impact of the Indirect Eects of Energy Subsidy Reform . 16
Using Prepackaged Simulation Models ......................................17
Measurement Challenges 19
Limitations in the Energy Spending Variables ................................ 19
Challenges in Extrapolating Energy Quantities Consumed from Energy Spending Data .. 20
Methodological Choices in Constructing Key Variables other than Energy Consumption . 22
Annex A: Sample Outline for a Report on the Distributional Assessment of Energy
Price Increases 24
Annex B: Selected Studies from the World Bank and Their Main Methodological
Issues 29
Annex C: The Measurement of Energy Poverty 40
Endnotes 41
References 45
II
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
TABLES
Table 1: Direct and Indirect Eects on Households of Increases in the Price of a
Previously Subsidized Energy Source 3
Table 2: List of Additional Data Sources 23
Table B1: Welfare Eects of the 2014 Reform, Direct Eects, million Moroccan
dirhams 38
Table B2: Indirect Eects of 2014 Reform (% of total eects) 39
FIGURES
Figure 1: First Order Impact Based on the Share of Energy Expenditures 11
Figure 2: Overview of the Simulation Steps 13
Figure B1: Evolution of the Average Prices for Gas and Electricity for Residential
Customers Armenia 29
Figure B2: Evolution of the Total Compensation Budget 37
III
ACKNOWLEDGMENTS
ACKNOWLEDGMENTS
This is the third in the series of 10 good practice notes under the Energy Sector Reform
Assessment Framework (ESRAF), an initiative of the Energy Sector Management
Assistance Program (ESMAP) of the World Bank. ESRAF proposes a guide to analyzing
energy subsidies, the impacts of subsidies and their reforms, and the political context
for reform in developing countries.
This note is a product of a team from the Poverty Global Practice summarizing insights
from recent global, regional, and country-level work on energy subsidy reform. A
number of dierent World Bank sources has been quoted freely and adapted with
the permission of the authors. They are listed in the first part of the references
with an asterisk (*). The authors are very grateful to Marianne Fay, Thomas Flochel,
Sudarshan Gooptu, and Gabriela Inchauste for their overall guidance and support
in producing this note and to Ezgi Canpolat, Sophia Georgieva, Amr Moubarak and
Ruslan Yemtsov, for their very helpful comments and suggestions. Masami Kojima’s
contribution greatly improved the scope and clarity of the paper and of the arguments
here made. Zuzana Dobrotkova provided excellent research support.
The authors remain solely responsible for any remaining errors.
IV
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
ABOUT THE AUTHORS
Anne Olivier
Anne Olivier is a research associate at DIAL, a research unit from Paris-Dauphine
University and the Research Institute for Development (IRD), where she
specialized in impact evaluation of poverty reduction programs and distributional
impact of water tari reforms. For the last ten years she has been collaborating
extensively with the World Bank on issues related to water and energy reforms
in the Balkans and Central Asia, ranging from survey design, modeling of
distributional impacts and the direct provision of technical assistance. In 2013
she co-authored Balancing Act, a regional report on the distributional impacts
of tari reforms in the Eastern Europe and Central Asia region of the World
Bank. The book has become a reference on adopting a multi-sectoral approach
to tari reform in the region. Prior to be engaged in academic research, she
worked on water supply projects in many countries, with Action Contre la Faim, a
humanitarian organization and with Suez Environnement, a water utility. She holds
a Ph.D. in economics from the Paris School of Economics and a master’s degree in
economic demography from Sciences-Po Paris.
Caterina Ruggeri Laderchi
Caterina Ruggeri Laderchi is a senior economist in the World Bank’s Poverty
Global Practice. She joined the World Bank in 2003 and has experience working
on Latin America, East Africa, and Eastern Europe and East Asia. Her main areas
of work are poverty, inequality, and social inclusion, as well as distributional
impacts of dierent policies, particularly energy pricing. In 2013 she co-authored
Balancing Act, a regional report on the distributional impacts of tari reforms
in the Eastern Europe and Central Asia region of the World Bank. The book has
become a reference on adopting a multi-sectoral approach to tari reform in the
region. She holds a Ph.D. in economics from the University of Oxford, and worked
for the Human Development Report of UNDP and the European Commission
before joining the World Bank.
V
ACRONYMS AND ABBREVIATIONS
ACRONYMS AND ABBREVIATIONS
CNG compressed natural gas
CPI Consumer Price Index
DH Moroccan dirham
ECA Eastern Europe and Central Asia
EST energy subsidy reform
GDP gross domestic product
HUS Housing and Utilities Subsidies
kg kilogram
kWh kilowatt-hour
LPG liquefied petroleum gas
m3 cubic meter
MW megawatt
OECD Organisation for Economic Co-operation and Development
tcm thousand cubic meters
VAT value added tax
VDT volume-dierentiated tari
1GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
1. INTRODUCTION
This note aims to provide guidance on how to
assess the distributional implications of energy
subsidy reform (ESR) using quantitative
methods.1 It is intended for use by those familiar
with the basics of welfare measurement, ideally
part of a multi-disciplinary team. Ideally this
assessment would therefore be complemented
by insights from qualitative analysis (see
Good Practice Note 4) and by an analysis of
the eectiveness of feasible compensatory
measures (see Good Practice Note 5).
The note focuses on how to assess the
distributional implications of household level
impacts of ESR (as opposed to firm level,
discussed in Good Practice Note 6). Its scope
is confined to cases where ESRs lead to
higher prices paid by energy consumers. As
Good Practice Note 1 outlines, ESRs do not
necessarily lead to higher prices, and could
even decrease prices actually paid, such as
when producer subsidies in the form of price
support paid for by consumers are eliminated,
or when consumer price subsidies lead to
illegal diversion and out-smuggling, acute fuel
shortages, and prices that are even higher than
ocial prices on the black markets. The latter
is particularly important, because a lack of data
often forces the distributional analysis of ESRs
to take observed expenditures on subsidized
energy and scale them in proportion to the
calculated price gaps—the gap between the
unsubsidized price and the ocial price—to
estimate the incidence of subsidies, whereas
in practice consumers may be paying much
higher prices than the ocial prices. Further,
this note is not confined only to ESRs in that the
distributional eects of higher prices of fuels
used as feedstocks—such as natural gas used
in fertilizer manufacturer—are also captured.
In the above context, this note considers
only the impact of price increases. Other
impacts, such as on service quality, access,
or accountability, are therefore not explicitly
considered.2 In addition, while this note tries
to present a general approach, practical
pointers are provided that are relevant for
the analysis of dierent types of energy, the
prices of which are rising, and which are used
either directly or in the production of goods
and services widely in the economy. Overall,
therefore, the note discusses the analysis of
liquid fuels, gas, electricity and district heating
(a source of heating used primarily in Eastern
Europe). The word prices applies to all forms
of energy, while taris applies to schedules
of regulated prices that are applicable to
regulated electricity, gas, or district heating.
The central issue that motivates this note
is that while energy subsidies are generally
inecient as a measure to redistribute income
to the poor (whatever the rhetoric on their
implementation),3 their removal is likely to
aect lower-income households negatively.
The hardship that rising energy prices may
impose on lower-income households in most
cases would appear to have a relatively
limited eect on the incidence of poverty.4
The impacts on the depth and severity of
poverty, however, might be much pronounced.
And given the diculties poor people already
face in meeting their basic needs, cutting
further into their budgets can have serious
negative consequences. If not compensated
for, higher energy prices aect livelihoods,
particularly of the poor, through their impact
on general inflation, and through direct eects
on households and businesses, especially
energy-intensive industries.5
21. INTRODUCTION
For households—the focus of this paper—two
main channels of impacts can be identified,
relating respectively to consumption patterns
and income streams. Both consumption and
income can be aected directly by higher
prices for energy, or indirectly through other
price changes triggered by the changes in
energy prices (most notably through higher
transport costs caused by rises in gasoline and
diesel prices). These indirect eects, though
harder to quantify than direct eects, can
be significant for petroleum products. For
example, Coady, Flamini, and Sears (2015)
estimate that indirect eects would account
for about 55% of the potential impact of the
rise in fuel prices, with significant dierences
by region depending on the energy intensity
of household consumption.6 Other indirect
eects which go beyond the focus of this
paper but have important implication for the
recommendations one can make in terms of
compensation strategies, include increased
exposure to fuel price volatility and the health
and environmental impacts linked to a shift
back to biomass.7
Table 1 summarizes the relative vulnerability
of dierent groups of people to the various
eects of removing energy subsidies. The
table is no more than illustrative, as it makes
some important simplifying assumptions. In
each country, actual impacts will depend on
consumption patterns, the extent to which
consumers can adjust their consumption when
prices change,8 and the distribution and type
of income-generating activities, particularly
those in which the poor tend to engage.9 10
There can be significant dierences between
rural and urban areas, especially in low-income
countries, as rural areas are not only typically
poorer, but also less likely to be connected
to grid electricity. Table 1 also assumes that
consumers pay ocial unsubsidized prices
before the subsidy removal. In fact, it is not
uncommon for consumer price subsidies for
liquid fuels to result in acute fuel shortages
and for households to pay higher prices on
the black markets.11
3GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
TABLE 1: Direct and Indirect Eects on Households of Increases in the Price of a Previously
Subsidized Energy Source
Direct eects Indirect eects
Direct eects impact users of the previously
subsidized energy source, which costs more
following reform.
Indirect eects touch all households in the
economy through (a) increases in costs of goods
and services that depend on the energy source
for which the price rises, and (b) increases
in the costs of other energy sources through
substitution and generalized inflation.
Consumption
All households with access to the energy source
will be aected.
Impacts will vary depending on the share of
household resources spent on the energy source
and the price elasticity of demand (which in turn
depends partly on the potential for substitution,
other price elasticities, the degree of nonessential
consumption, and so on).
In the case of liquid fuels:
• Rural households, in particular the rural poor,
would be aected by kerosene price increases,
as kerosene is often used for lighting and
cooking.
• Urban households that have cars would be
aected by higher gasoline and diesel prices.
This eect may be especially relevant for poor
families in higher-income countries, especially
those with weak public transportation systems.
In the case of networked utilities (electricity, gas,
district heating):
• In better-o countries, consumption of energy
by the poor tends to be inelastic with respect
to price.
• The urban poor typically pay for fuels and are
also more likely to be connected to electricity
than their rural counterparts, making them
more vulnerable.
• Natural gas is generally not available in rural
areas. In low- and lower-middle-income
countries, rural households are also less likely to
be connected to electricity, or to consume LPG
as the primary fuel for cooking and heating.
Consumption
Strong indirect eects are to be expected for
higher prices of transport fuels, particularly diesel
(used by trucks).
Eects will vary depending on the consumption
baskets of poor households, on the price elasticity
of demand for dierent goods, and on the distance
between production and consumption centers.
In principle, one would expect the urban poor,
who are most dependent for their basic needs on
goods transported from somewhere else and on
public transportation for personal transport to be
most vulnerable to these eects, but this cannot be
ascertained a priori. There is evidence, for example,
that oil prices significantly aect food prices
(maize) in subnational markets.
Indirect eects are likely to be minor for LPG,
which is subsidized largely for use by households
(although illegal diversion to commercial
establishments is common), and for kerosene
(jet fuel, which is its main use, is typically not
subsidized, while about two-thirds of the rest is
estimated to be used by households).
41. INTRODUCTION
Broadly speaking one can distinguish three
types of analyses: (a) general equilibrium
analyses, incorporating both the direct and
the indirect welfare eects of the reforms;
(b) limited general equilibrium, incorporating
only a subset of the indirect eects; and (c)
partial equilibrium approaches focusing only
on the direct eect of reforms on prices and
household real incomes. The latter two are
commonly considered as the short-run impact
of reforms prior to household and producer
responses. They are also considered an upper
bound on longer-term adverse impacts, since
household responses (for example, switching
consumption away from goods that underwent
a price increase or toward subsidized goods)
tend to decrease adverse welfare impacts and
increase beneficial welfare impacts, although
they require time to materialize. This would
typically be through some eciency measure
cutting demand (for example, in the case of
heating insulation, higher eciency stove or
heaters) or switching to alternative energy
sources (for example, investing in dierent
fuel type heating equipment, or switching
to public transport to minimize expenses on
fuel for transportation).
Given this general context, this note is
structured as follows: Section 2 provides
an overview of the dierent issues that a
quantitative assessment of the distributional
impacts of price and tari increases would
seek to address. Sample TORs for conducting a
partial equilibrium assessment of distributional
impacts are included in Annex A. Section
3 provides a quick summary of the main
methodologies for estimating the welfare
impacts of price increases. Section 4 provides
an overview of practical issues related to
Direct eects Indirect eects
Income
All households that use the energy source for
income-generating activities will be aected. For
those engaged in commercial activities, the extent
of the impact depends critically on how much of
the additional input cost they can pass through to
final consumers.
Groups that have been found to be particularly
vulnerable include fisherfolk (who depend on diesel
fuel), farmers (who use diesel or electric pumps for
irrigation), and small and medium enterprises.
Income
All households involved in productive activities
are likely to be aected by increases in input costs
stemming from rising energy prices (such as higher
costs to transport inputs, higher costs of energy-
intensive inputs).
In some sectors, indirect eects can be particularly
strong. For example, agricultural households are
likely to be more aected by rising fertilizer costs
linked to the increase of specific feedstocks, such
as natural gas. Poorer farmers who are less likely
to adopt modern technology may still be aected
by substitution eects between energy sources.
For example, owing to higher prices of fossil fuels,
the biomass that is used for fertilizer in traditional
agriculture and as an energy source may become
scarcer and more expensive.
Other
Higher fuel prices can result in broader impacts on
the livelihoods of poor people and their communities.
Examples include the health, environmental, and
social impacts of greater reliance on traditional
biomass (often with a strong gender dimension in
the burden of collecting it), including exposure to
significantly high levels of unhealthy pollutants.
Source: Adapted from Ruggeri Laderchi (2015).
5GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
performing this type of analysis, focusing
in particular on data issues. Suggestions
for questions to be included in a household
survey (where such possibility might arise)
are included in Annex B. Section 5 concludes.
2. THE PROCESS OF CONDUCTING DISTRIBUTIONAL
ANALYSIS
Before looking at the technical aspects of the
distributional analysis of the impacts of ESR,
it is worth pointing out that while a number
of solid technical tools have been developed,
expert judgment relying on dierent areas
of expertise might be required to address
shortfalls in the information available. In
addition, the nature of the process is such that
several of the steps, if not most of the analysis,
might need to be conducted iteratively. In
practice, therefore, while the note tries to
describe an organized sequence of activities,
the analysis might be less straightforward
than it appears.
Examples of the complications which require
expert judgment include the data limitations
of standard household surveys when covering
energy issues, and the realities of energy
provision in low-income context where access
or metering might be limited, fuels might
be sold in dierent parts of a country with
dierent prices, multiple connections and
reselling of power might be practiced, and
rationing and smuggling might be rife.
As far as the process is concerned, providing
advice on how to conduct these reforms
requires collaboration across dierent areas
of expertise, often through repeated iterations.
An understanding of the policy contexts and of
the way dierent aspects of the reform interact
is therefore essential. A poverty economist
working in this area might start with the
simple brief of assessing the distributional
impact of existing subsidies and of how
“market parity” or “cost recovery” prices for
one or more energy sources would aect
the population. This, however, might already
entail familiarizing oneself with a number of
sectoral technical details. For example, the
unsubsidized price of energy might depend
on choices made by energy sector analysts
on the time horizon that is relevant (and
therefore whether investment plans should be
factored in or not), and estimates on which
would be considered reasonable levels of
eciency in that market. And these details
matter much beyond personal knowledge—
from the equity point of view, for example,
understanding whether existing estimates
would saddle consumers with the burden of
making energy sector’s providers’ budgets
square despite their various ineciencies (for
example, high distribution losses) is more than
a technical detail.
In addition, as the dialogue of a multisectoral
team progresses, the initial brief of the poverty
economist might expand. Dierent reform
options—for example dierent pathways of
price increases, or dierent combinations of
changes to the price structure (for example,
the introduction or refinement of lifeline
tariff levels and structures for networked
utilities) and price increases—might have
to be considered. Qualitative analysis
(discussed in Good Practice Note 4) may
63. A QUANTITATIVE ASSESSMENT OF DISTRIBUTIONAL IMPACTS: KEY QUESTIONS AND HOW TO APPROACH THEM
point to specific—for example, seasonal—
patterns of household energy spending,
dierent types of energy use, and nature of
household coping mechanisms that may reveal
higher vulnerabilities for specific groups. In
addition, technical inputs on the analysis of
existing safety nets conducted by the Social
Protection specialist might be easily produced
when household surveys include the required
information (as discussed in Good Practice
Note 5, this might not be the case and dierent
data sources might be required to conduct this
type of analysis, in addition to the program
readiness and institutional assessments).
And when compensation options are being
discussed the poverty economist might need
to collaborate closely with the rest of the
team to identify pros and cons of dierent
alternatives (such as compensation through
a new targeted social protection program;
compensation through a combination
of existing social protection programs;
compensation through a combination
of cuts in out-of-pocket expenditures on
health, education, and food) by evaluating the
distributional implications of each. Experience
shows that dierent teams find dierent ways
in practice of managing the complementarities
and potential overlap between the tasks of
the poverty economist and the other team
members, such as by relying on the same
research assistant, or working sequentially
on different parts while sharing program
files, or even by working independently and
collaborating in the packaging of the outputs
for a common audience. It is important,
however, to be aware of the complementarities
between these dierent tasks and how they
might evolve during the assessment process.
3. A QUANTITATIVE ASSESSMENT OF DISTRIBUTIONAL
IMPACTS: KEY QUESTIONS AND HOW TO APPROACH
THEM
A microeconomic analyst working in a team
supporting energy subsidy reform is typically
asked to help answering three interrelated
policy questions:
• How large are energy subsidies and who is
benefiting from them?
• Who is going to be aected by the removal
of energy subsidies and—more specifically—
would poverty increase significantly?
• How much would it cost to compensate
vulnerable groups?
Answering these apparently simple questions
requires pulling together a lot of dierent
sectoral expertise, and analyzing very dierent
kinds of data. In the context of the three
questions above, for example, a poverty
economist would be mostly focusing on the
second question, by looking at who is going
to be affected by energy subsidy reform
(ESR), how large the subsidies are relative
to household income and expenditures, and
what the estimated impacts of the removal
itself would be. However, he or she would also
have to provide an analysis of who benefits
from the energy subsidies,12 as well as the
incidence of the benefit, and might also be
called on to provide some insight from the
household survey to triangulate other sources
when estimating the absolute size of the
7GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
subsidy (something that is typically done by
an energy or macro specialist). Finally, the
poverty economist will also need to work
very closely with whoever is trying to assess
the costs of compensation, and might be
involved in modeling dierent compensatory
measures and their targeting performance.
In addition, tailoring the answers to the local
realities requires a close engagement with the
policy dialogue so that only relevant options
are analyzed.
This section provides a quick overview of the
role of the analysis of household survey data in
providing answers to the three key questions
highlighted above, where complementary
information and analysis is needed, and where
the distributional analysis, which is the main
focus of this note, would fit in.
How Large Are Energy Subsidies and
Who Is Benefiting from Them?
With respect to the first issue—how large
are energy subsidies and who is benefiting
from them—the analysis of a household
survey can mostly contribute to the second
part of the question: provide an estimate of
how dierent forms of subsidies that reduce
household expenditures on energy benefit
dierent households.13 While such an analysis
is limited, given that households often benefit
directly only from a small proportion of overall
subsidies, it can prove very helpful in policy
analysis, since subsidies are often justified on
the grounds that they make specific energy
sources aordable to poorer groups in society.
Such an analysis, for example, can highlight the
poor’s consumption of very limited quantities
of the energy sources subsidized and, as a
result, are very minor beneficiaries of the
subsidies. At the same time, caution should be
used in interpreting the results. While focusing
only on households, subsidies may appear
relatively progressive (more progressively
distributed than income). Even such a benign
assessment might prove incorrect once
consumption of subsidized energy sources
across the entire economy is considered. The
case was powerfully illustrated in the case of
the kerosene subsidy in India (World Bank
2003), where an assessment was reversed
(from not regressive to highly regressive)
once the finding that nearly half of subsidized
kerosene had been diverted, most likely to
the automotive diesel sector, was taken into
consideration.
Classifying subsidies may be done in a number
of ways. For example, Good Practice Note 1
discusses dierent mechanisms through which
subsidies are distributed. For the purpose of
this note, as we are focusing on households,
we adopt a simple distinction based on how
these subsidies reach households. The first is
consumer price subsidies, whereby consumers
pay less than in the absence of the subsidy.
The second is cash transfers or discounts,
which reduce the net expenditures on energy
by consumers.
While the analysis of the latter can rely on
well-established tools for the analysis of
social protection programs,14 the analysis
of consumer price subsidies presents some
additional complications. Households benefit
from price subsidies in proportion to their
consumption of the subsidized energy.
Since information on quantities consumed
by households is seldom, if ever, available
from household expenditure surveys or any
other surveys, quantities consumed must be
imputed. A common approach is to divide the
expenditures by an average price to arrive at
the quantity, and multiply by the dierence
between the estimated unsubsidized price and
the average price to arrive at the incremental
cost of subsidy removal for each household.
83. A QUANTITATIVE ASSESSMENT OF DISTRIBUTIONAL IMPACTS: KEY QUESTIONS AND HOW TO APPROACH THEM
The total amount of subsidy captured by
the households computed in this way can be
calculated and plotted with standard tools.
Where specific programs provide transfers
to various groups to help consume energy,
estimating the distribution of these benefits
is akin to estimating the distribution of
any other social protection program. The
analyst can therefore rely on such tools as a
concentration curve, which plots the share of
the overall expenditures spent on subsidies
that reaches the bottom x% of the distribution.
Concentration curves can be plotted alongside
Lorenz curves in order to assess the extent
to which they are progressive or regressive in
relative and absolute terms. Indicators such
as concentration coecients and Kakwani
indices can be computed, as well as standard
indicators of program performance, such as
the following:
• Coverage: Percentage of households where
at least one member receives subsidies
• Targeting or concentration shares: The
share of total transfers going to each group
(typically welfare quintile)
• Generosity or relative incidence: The
percentage of total household income and
expenditure constituted by the subsidy15
and, if there is a shared understanding
on whom the program should reach (for
example, the poor or the bottom 20%)
• Leakage: The proportion of those who
benefit from the subsidy who are not part
of the group they are meant to reach
Since subsidies are proportional to
consumption, they tend to be captured by the
better o, making them regressive (although
they may be relatively progressive if they
are more equally distributed than income).
However, their removal may harm the poor
disproportionately if they are able to purchase
subsidized energy and subsidies comprise a
large share of their household income and
expenditure. A household survey can be used
to analyze whether subsidies captured by
households are regressive by comparing the
concentration coecient of the subsidy with
the Gini. If the concentration coecient is
negative, the subsidy is progressive and pro-
poor. If it is positive, but lower than the Gini, it
is relatively progressive. If the concentration
coecient is higher than the Gini, subsidies are
regressive both in absolute and relative terms.
It is important to note that this comparison
does not allow drawing conclusions about the
overall progressiveness of the price subsidy
for that energy source. To assess the overall
progressiveness, data for all purchasers of the
subsidized energy source are needed, and
such data are never available in cases where
fuels are smuggled out of the country. One
exception is where the subsidy targets only
households. In such cases, a comparison of
the total amount allocated by the government
and the total amount consumed by households
will tell how much has been diverted, provided
that quantities consumed by households can
be reasonably estimated (see section 4 on the
diculties of doing so). A rare example of this
is India’s household expenditure survey, which
explicitly asks how much subsidized kerosene
and unsubsidized kerosene the household
consumes.
Who Is Going to Be Aected by the
Removal of Energy Subsidies and—More
Specifically—Would Poverty Increase
Significantly?
This is the central question that an analysis of
the distributional implications of ESR should
address. The frequently encountered lack of
information on the actual quantities consumed
9GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
poses a significant challenge to estimating
the impact on poverty, underscoring the
importance of highlighting the limitations in
discussing the findings.
Bearing the foregoing limitations in mind,
even in cases where policy attention is on
one single number (say, the poverty impact),
a comprehensive analysis of this issue would
entail the following steps:
1 | Understanding consumption patterns and
how they change by rural/urban location,
region, access to different sources of
energy, season when relevant, and any
other relevant characteristics (for example,
welfare group).16 Analysis of access,17
consumption and spending patterns, and
use of energy (for example, which source is
used for cooking, heating, lighting) allows
identification of categories of households
vulnerable to specific energy price
increases, since households’ vulnerability to
price increases crucially depends on their
demand for that energy source and the
availability of substitutes. If multiple data
points are available, analysis at dierent
points in time can help correlate past price
changes with changes in consumption
patterns, whether directly on energy or on
other items of interest (such as changes
in education or health spending resulting
from past price hikes). Beyond the main
indicators on consumption and spending
on dierent sources (as relevant to the
analysis), indicators of aordability—such
as energy expenditure shares, or others
that might be collected where relevant,
such as information on payment delays,
arrears on utilities, or the self-reported
ability to heat adequately—can also prove
very informative in this initial descriptive
step. Qualitative tools can also be used
to assess households’ level of access to
dierent types of energy sources, and
the purpose of their use, seasonality of
spending patterns, and differences in
consumption patterns by urban or rural
location (see Good Practice Note 4).
2 | Measuring the impact of the price increase
on household welfare. This is the core of
a distributional assessment, and its main
output is a simulated welfare distribution.
Additional variables, such as simulated
quantities consumed and simulated
energy shares, can also be produced.
Once household level eects have been
computed, they can be summarized and
graphed by income group, poor or non-
poor, and for other relevant categories
(for example, social assistance recipients,
pensioners, or other groups considered
“vulnerable”).
A crucial distinction in this analysis is the
extent to which the indirect welfare eects
that result once households and firms respond
by changing their demand for and supply of
goods and services and factors of production
following the price increase are incorporated
into the analysis.
3 | Summarizing the impacts of some key
indicators according to the focus of the
study. The real income or expenditure
distribution after energy price increases
can be employed to calculate poverty
incidence or other poverty indicators (using
commonly used national or international
poverty lines), such as depth, severity, or
the number of new poor. Where energy
shares of household expenditures are
high and aordability is a critical issue,
measures of affordability, such as the
energy share, the rates of energy poverty,
or poverty itself are useful indicators
(See annex C for more information on
measuring energy poverty). Estimates of
103. A QUANTITATIVE ASSESSMENT OF DISTRIBUTIONAL IMPACTS: KEY QUESTIONS AND HOW TO APPROACH THEM
the numbers of those who would qualify
for some important means tested program
might also be obtained, if relevant to the
policy dialogue. As mentioned above, this
is often the part of the analysis that attracts
most policy attention, although ultimately
it is just a way of summarizing information
produced in the previous stage.
How Much Would It Cost to Compensate
Vulnerable Groups?
Based on the estimates of how much households
have been aected described above, one
can simulate how existing or potential social
assistance programs could help mitigating
the adverse effects of reforms on poor
households or specific vulnerable categories
of the population. Good Practice Note 5 oers
a comprehensive view of how compensations
options through social assistance should
be assessed, and Good Practice Note 7, by
providing a macro-perspective, highlights how
a plurality of compensation options might be
considered beyond compensation through
a social protection program. Yet, depending
on the country dialogue, there might be
scope for producing highly stylized estimates
of compensation costs to give a sense of
the overall magnitudes being involved. For
example. it might be useful to compare the
amount of resources absorbed by subsidies
with the amount that would be absorbed if it
were possible to compensate the poorest x%
of the population for the price increase, quite
irrespective of whether it is possible to put in
place a program that perfectly compensates
the bottom x% of the population.
With this understanding in mind, and as an
input to the broader (iterative) discussion
on compensation options mentioned in
section 2, dierent types of simulations can
be conducted, such as the following:
• Simulations conducted assuming perfect
targeting capture the mitigation budget
required if there were no administrative
costs and perfect information. While both
assumptions are unrealistic, this highly
stylized measure can be useful in the policy
dialogue to compare the budgetary costs of
dierent mitigation strategies (for example,
compensating all the poor or compensating
only the poor who would become eligible for
some targeted program) or dierent price
increase scenarios (for example, a big bang
scenario versus a scenario of more gradual
increases).
• Simulations trying to replicate the eligibility
criteria for dierent programs in contrast oer
the advantage of showing how well dierent
program designs would protect eectively
dierent groups of interest, although such
simulations might be limited if the household
survey does not cover all the indicators used
to determine program eligibility.
An important caveat to keep in mind is that
for the comparison between dierent program
designs to be meaningful, if at least one of
the programs is already in place, eligibility in
the baseline should also be simulated. This is
to make sure that the comparison between
scenarios is not biased by such issues as
low take up or administrative error (which
would be captured by the real data on transfer
receipt), as well as by the limited replicability
of real life eligibility criteria with the household
survey information.18
11 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
4. DIFFERENT METHODS TO ESTIMATE HOUSEHOLD-
LEVEL WELFARE IMPACTS OF ENERGY SUBSIDY REFORM
PARTIAL EQUILIBRIUM ANALYSIS
Most analyses rely, at least as an intermediate
step, on the estimate of partial equilibrium
eects. Such direct impacts are estimated
following one of two approaches. The first
focuses on the purchasing power loss the
household experiences following the price
increase. Therefore, it measures the impact
in terms of real incomes and expenditures.
Intuitively this is proportional to the share
spent on the item in the household expenditure
(figure 1). For example, if a household spends
2% of its expenditure on a given energy source,
whose price is going to double (increase by
100%), and it makes no adjustment in the
amounts it consumes, its purchasing power
loss will be 2%. This can be thought of as the
real income or expenditure of the household
contracting by 2%.19 Note that this approach
lends itself very naturally to estimating (partial
equilibrium) poverty impacts incorporating the
purchasing power loss that every household
experiences into the welfare indicator of
interest (for example, household per capita
income or consumption).
A second approach is to measure the impact
on households in terms of their welfare.
Following Hicks (1942), the welfare impact
of price change on households in the case of
a purchased good is commonly understood
as the additional expenditures required to
reach the same level of utility following a
price increase than with the initial price, a
measure called the Compensating Variation.20
Such a measure requires prior knowledge of
the preferences and various methodologies
have been presented in the theoretical and
empirical literature to approximate this
FIGURE 1: First Order Impact Based on the Share of Energy Expenditures
Without mitigation
Price
Price increase
Indicators at the
Household level
Share of energy
expenditures
Increased share of
energy expenditures
Loss of
purchasing power
Access to Energy consumption
Energy expenditures
Increased energy expenditures
124. DIFFERENT METHODS TO ESTIMATE HOUSEHOLD-LEVEL WELFARE IMPACTS OF ENERGY SUBSIDY REFORM
welfare measure with limited assumptions.
The simplest approximation of the welfare
measure is the Laspeyres Variation, the change
in income required to purchase the original
quantities of the good after the price has
changed. This measure would be an upper
bound of the actual welfare change in the
absence of energy shortages before the price
increase. If energy rationing was preventing
households from purchasing more energy,
should shortages be significantly reduced
following the price increase, some or many
households may purchase greater quantities
despite the price increase.
Another widely used measure is the loss
of Consumer Surplus, which requires
information on the demand function, but
not on preferences. For non-negligible price
elasticity,21 as Consumer Surplus does not
assume a constant utility, this measure
underestimates the true welfare eect and
is smaller than the Compensating Variation
measure. The Laspeyres Variation and
Consumer Surplus loss can be used as the
lower and upper bound of the true value
of the real welfare change for an energy
price increase. The lower the price elasticity
are and the closer these measures and the
Laspeyres Variation are is a good measure
of the Compensating Variation when price
elasticity is close to zero (see Cory and others
[1981], as well as Araar and Verme [2016] for
an updated review of these measures).
GENERAL EQUILIBRIUM EFFECTS
Estimates of the indirect welfare effects
of higher energy prices would ideally be
captured through the use of an input-output
matrix incorporating the energy intensity
(disaggregated by source) of each sector.
Making some simple assumptions on how
an increase in energy will increase the prices
in each sector, and knowing how much each
energy source under consideration contributes
to the value of production of the goods and
services produced by each sector one obtains
an estimate of the first-round eects of the
increase in one energy price on the prices of
all other goods and services produced in the
economy. Of course, increases in the prices
of the goods produced in each sector will
feed through to other sectors, so that the
process of identifying how much the overall
price increase will be has to be performed
iteratively. The indirect real-income eect is
then calculated by multiplying the expenditure
shares of the various goods and services by
the estimated percentage price increases
in these sectors. Note that the precision of
this simple method depends heavily on the
possibility that households have to switch their
consumption away from fuels and goods and
services with relatively high price increases
toward those with relatively low increases.22
Depending on the energy source under
consideration (increases in the price of
petroleum products typically have indirect
effects in many sectors if not all through
transport costs, while increases in prices that
aect only households, such as district heating,
which is typically provided to apartment
buildings, can be expected to have negligible
indirect eects) and the availability of an
I/O matrix, one can use a simpler approach
to incorporate only some indirect effects
which, a priori, one can consider most relevant.
This approach, which can be defined as
one of “limited general equilibrium,” would
involve identifying and quantifying based
on expert judgment of which channels are
likely to matter the most in a given context.
For example in an agricultural economy
there might be important repercussions of
changing natural gas prices through their
indirect effects on the cost of fertilizer if
13 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
fertilizer manufacturers are benefiting from
subsidized natural gas, which is used as a
feedstock for fertilizer production, while the
impacts on other sectors outside of agriculture
may be less. Depending on the coverage of
variables related to farm production in the
household survey, this important element
could be modeled and included to the analysis
of the direct distributional impacts of the gas
price increase.
5. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
This section discusses first the logic, then some
of the common issues faced when having to
apply the approaches described above in
practice. Those mostly refer to shortfalls in
the data and pragmatic ways in which they
can be addressed.
SETTING UP THE ANALYSIS AND
COMPLEMENTARY DATA NEEDS
To assess distributional impacts in practice
one might rely either on pre-packaged options
or on one’s own programming. Either way
it is helpful to understand the logic of what
is being done, both to understand how the
results can be interpreted and to choose the
tool that might be most appropriate.
ASSESSING THE DISTRIBUTIONAL
IMPACT OF THE DIRECT EFFECTS
OF ENERGY SUBSIDY REFORM
Figure 2 provides a summary of the two main
steps involved assessing the distributional
impact of the direct eects of energy subsidy
reform (ESR), namely constructing a simulated
baseline and simulating the impact. One
typically needs to simulate the baseline if
the household survey is not recent (as a rule
of thumb if the survey is more than two years
old) or if growth and inflation have been
significant between the time the survey was
run and the time when the energy reform has
taken place or is being considered.23 Once
the counterfactual scenario is constructed
(usually under the simplifying assumption
that economic growth has resulted in a
distributionally neutral shift, that is, that all
incomes have been growing at the same rate),
FIGURE 2: Overview of the Simulation Steps
Increase in the share of
expenditures, by quintiles
Change in real income
Poverty headcount using
the (adjusted) poverty line
ENERGY SHARES
for
Survey Year
ADJUSTED
ENERGY SHARES
for
Reference Year
INCREASED
ENERGY SHARES
Latest Household
Budget Survey
Constructed Baseline Simulation
ENERGY PRICE INCREASE
145. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
the simulated energy expenditures shares
and the real income after the energy price
increase (actual increase or expected increase)
can be compared with the forecasted energy
expenditures shares under the constant price
scenario of the baseline. Note that if poverty
impacts are to be calculated, the poverty line
also needs to be updated to make sure that it
retains the same constant purchasing power
as in the base year.
As a first step, the total nominal expenditures
(prior to any adjustment) is required in order
to calculate the share of energy expenditures
as a share of the total nominal expenditure. It
is usually dierent from the welfare indicator,
which is already adjusted for the dierent
cost of living across the sample. In computing
energy expenditure shares, the following issues
need to be considered:
• What is included in total expenditures varies
from country to country: For poverty analysis,
most countries exclude the purchase of large
durables that can cause a one-o inflation
of the total expenditures of the aected
households. Some countries also exclude
large expenditures on rare events, such as
weddings. It is important to understand
what is excluded especially when making
cross-country comparisons.
• Not all expenditures are in cash: Energy
price increases are relevant for energy
products for which households have to
pay. Total expenditures, by contrast, often
include items for which households did
not pay. For the poor, the largest of such
“expenditures” is food that was given to
or grown by household members. Among
energy items, biomass collected by
household members (such as wood and
agricultural residues) is a common example,
particularly in rural areas and in low- and
lower-middle-income countries. These items
may or may not assigned imputed values,
typically market values in the vicinity of the
household in question. For the purpose of
poverty analysis, total expenditures inclusive
of such imputed values are important. To
understand the impact of higher energy
prices on the poor, however, it would be
useful to examine energy expenditures
shares of cash-only expenditures (excluding
imputed values of all items for which the
household did not pay), as well as total
expenditures used in poverty analysis.
Depending on the type of analysis conducted,
and whether an actual reform is mimicked, or
potential scenarios are simulated, dierent
sets of variables are required, such as the
following:
• Distribution of the subsidies: To estimate
the distribution of energy price subsidies
before and after the reform, unsubsidized
prices need to be estimated. Good Practice
Note 1 reviews the data requirements and
methodologies for dierent types of energy.
However, it is worth repeating the points
made earlier that household expenditures
on energy may not be correlated with their
actual consumption—because households
are not individually and accurately metered,
or payment arrears are common, or prices
paid dier from the ocial prices (utility
sta extracting unocial payments, fuel
shortages pushing up prices on the black
markets). In the case of network energy, it
should be possible in theory to compare
the aggregated imputed consumption from
the household data with utility data on total
consumption for such users to assess if the
extrapolations of quantities from expenditure
data are broadly reasonable. A general sense
of whether such extrapolations on average
results in an over- or underestimation of
15 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
one type of energy consumption, while
helping qualify the results, however, does
not necessarily help in the distributional
analysis.
• Distributional impact of a uniform price
increase: Simulating the direct impact of
a uniform price increase requires limited
data at the household level, as only the
share of expenditures for energy, the
total expenditures and the price increase
are required. These simulations can be
conducted by software. such as SUBSIM,24
or with International Monetary Fund (IMF)
Stata do-files25 described below, if quantities
can be reasonably accurately estimated
from the price data available. Yet even in this
relatively straightforward setup, challenges
might arise.
• If prices are regulated, but black
market activity means that households
are paying higher (and possibly very
dierent in terms of location) prices,
simulations based on the ocial price
are going to represent an upper bound
of the welfare loss.
• If rationing is present, it is impossible
to predict future consumption based
on pre-reform consumption levels.26
In the case of electricity, for example,
which is rationed in most countries in
Sub-Saharan Africa, if price increases
translate into improved service,
consumption could increase. In this case,
the increase in spending would capture
an increase in welfare as opposed
to a negative impact. A sense of
distributional impact, however, could be
obtained by simulating how the value of
a basic amount of, say, electricity, would
change and analyzing whether such
basic amount would be aordable by
dierent groups (for example, answering
the question “Could the bottom quintile
aord to cover its minimum consumption
needs considering the basic electrical
appliances they need?”).
• Distributional impact of non-uniform price
increases: A more common reform approach
is to dierentiate price increases by energy
type. For fuels, this could mean increasing
the prices of higher-grade gasoline (higher
octane) and diesel (lower sulfur) much
more than lower-quality fuels, or LPG sold
in large cylinders (LPG sold in 15-kilogram
cylinders more than that sold in 5-kg
cylinders). In the case of network energy,
uniform price increases are virtually unheard
of. With a few exceptions, such as Liberia,
virtually all countries have tari schedules
unique to residential consumers, and within
residential consumers, many have two or
more schedules. Where there are two or
more schedules, without knowledge of
which schedule each household subscribes
to, only an approximation of the increase
using a uniform (average) increase can be
conducted if the household survey does
not include the information needed to
dierentiate dierent types of customers.
Where the tari schedule applicable to each
household is known, knowledge of the actual
tari structure allows simulation of the price
increase applicable to each household.
The analysis requires the variables to
identify the categories targeted by each
specific tariff (for example, in Belarus
and Ukraine, the tariff varies according
to the type of heating appliances used by
the household), in addition to the tari
structure. In the case of dierentiated tari
increases in multiple-block tari structure—
for example, the tari for the first block is
unchanged while the upper block taris are
increased—the information must be even
more accurate, since the consumption level
165. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
must be estimated prior to the simulation
of the price change.27 When tari diers by
season, the survey should cover the dierent
quarters. Information on the taxes and other
fees applied is essential, but gathering that
information is not necessarily easy because
taris are often reported by the utilities
excluding taxes and fees not retained by
them, while household expenditures include
all charges.
ASSESSING THE DISTRIBUTIONAL
IMPACT OF THE INDIRECT EFFECTS
OF ENERGY SUBSIDY REFORM
As already discussed, several studies have
underlined the magnitude of raising the prices
of some petroleum products, such as through
indirect eects caused by higher costs of
transport for distributing goods and services
to households, and higher costs incurred in
agriculture from higher irrigation and fertilizer
costs. This evidence suggests that although
direct eects display a strong distributive
pattern in average, indirect eects of fuel
price increase appear distributionally neutral.
Atamanov, Jellema, and Serajuddin (2015) show
dierent eects in Jordan depending on the
fuel considered. The removal of diesel subsidies
would mainly impact households through large
indirect eects (households hardly use any
diesel directly), while the removal of gasoline
subsidies would mainly aect households
directly, with a greater eect on the wealthiest.
The same paper, when analyzing electricity
subsidy reforms, shows a greater impact on
wealthier households, through direct as well
as indirect eects because richer households
consume more non-food goods and services,
whose production is electricity-intensive. In
magnitude, electricity reform indirect eects
may amount to 40% of total eect, depending
on the reform scenario considered.
Analysis of the indirect effects of a price
increase, such as in Coady, Flamini, and Sears
(2015) study, are based on a “price shifting”
model, which describes and quantifies the
magnitude of sectoral changes in producer
and retail prices resulting from an exogenous
price shock. It uses information on the
current structure of an economy, at current
levels of production, reflected by an Input/
output Matrix, thus is a static model (see
also Good Practice Note 7 for limitations of
such fixed coecients models compared to
more comprehensive Computable General
Equilibrium models able to estimate longer
term eects).
The model assumes that exogenously
generated price changes are either “pushed
forward” to output prices or “pushed
backwards” onto factor payments when
output prices are fixed (determined by world
prices or controlled by the government).
The model also assumes constant returns
to scale in production, perfect competition,
and reproducible fixed factors of production
economywide. These assumptions allow the
analyst to use the input-output matrix—which
describes the input shares (of all sectors) in
the output of all sectors at a point in time
and given prevailing prices—to generate
producer price changes assuming production
technologies and production input shares
remain fixed. Results generated are considered
as an upper-bound estimate of the impact
of any change in government-administered
price policy on household welfare.
Under the specific assumption that all sectors
are either Cost-Push or Controlled,28 the
change in Cost-Push retail prices separates
the direct eect of the shock from the indirect
eects arising from changes in producer prices
in the Cost-Push and Controlled sectors. In
order to solve the price-shifting model using
17 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
one of the software alternatives and to use
results to trace the impact of price policy on
household welfare, the following steps should
be completed.29
1 | Prepare the Input-Output (I/O) matrix. The
analyst should choose an I/O year closest
to the year of the primary household
survey. Both the OECD (http://www.oecd.
org/trade/input-outputtables.htm) and
the World Input-Output Database (www.
wiod.org) maintain I/O databases that are
regularly updated. I/O matrices are usually
stated in flows: each row will describe the
value of that sector’s output by destination
(that is, did the sector’s output go to other
sectors for use as production inputs or
to households for consumption?), and
each column will contain a complete list
of the value of production inputs (from
each sector). To figure out the weight
of each input in each output, one must
calculate the technical coecients. This
is done from the flows in the I/O matrix
by dividing each cell in column j with the
row sum (that is, total output) from the
final row (where i=j). Technical coecients
express the value of inputs (in a sector) as
a share of the value of total output from
that same sector.
2 | Map household consumption expenditures
to I/O table sectors. The analyst will need
to use his or her judgment in mapping
each household questionnaire item to
the relevant I/O sector. In cases where an
item consumed by the household could
plausibly come from more than one sector,
it is reasonable to split each household’s
total consumption of that item between
all plausible sectors according to sectoral
share in total output (according to the
I/O table).
3 | Calculate the subsidy as a percentage of
the market price/reference price and map
the subsidy schedule to I/O table sectors
4 | Determine which (if any) I/O sectors
would continue to have regulated/non-
market prices if the price policy under
consideration were revised. For example,
in the case of fuel subsidies, the relevant
counterfactual may more likely be one
where the government still controls the
price of fuel even after eliminating the
current subsidy. In such a counterfactual,
fuel would be sold at a higher price, but the
price at which it sold would not necessarily
be freely determined by market supply
and demand.
5 | Read in the I/O matrix with the software.
6 | Enter exogenous price shocks and
designate sectors with fixed prices.
7 | Solve the model.
USING PREPACKAGED
SIMULATION MODELS
Various prepackaged simulation models
exist, with SUBSIM and the so-called “IMF
files”30 being popular. However, they are useful
primarily if quantities can be back-calculated
with reasonable accuracy based on available
price information. Unfortunately, as explained
in detail elsewhere, this is not the case in many
countries. Where the requisite information is
not available to estimate quantities consumed,
more tailored albeit often less ambitious
approaches based on whatever information
is available are called for.
SUBSIM is a World Bank tool designed to
facilitate and standardize rapid distributional
analyses of subsidies and simulations of
subsidies reforms (Araar and Verme 2012),
185. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
especially when indirect eects are expected to
be significant. The model estimates the impact
of subsidies reforms on household welfare,
poverty and inequality, and the government
budget with or without compensatory cash
transfers. It either estimates an upper bound
of the welfare eect on household of the price
increase, assuming that the loss in welfare
is approximated by the total expenditures
required to maintain the same level of
consumption than before the price increase
(thus the Laspeyres Variation) or a lower
bound (the Equivalent Variation) using them as
a reference for the welfare level after the price
increase, with an assumption on preferences
based on a Cobb-Douglas utility function, as
well as a price-elasticity. There are two versions
of the model to estimate direct and indirect
eects using household expenditure survey
data and input-output matrixes. Note that
the analysis with an I/O table, by requiring
that expenditure items from the household
survey are matched with those covered in
the I/O table, requires aggregating items and
considering their average prices. Depending
on the detailed nature of the I/O table, it might
require aggregating all petroleum products in
one category, or almost certainly considering
only one type of “gasoline” without reference
to the quality (such as the octane level).
For the estimation of the indirect eects,
the SUBSIM software creates automatically
a technical coefficient matrix from the
input-output matrix. The price shifting
model presented above is simulated under
a “permanent price shock” option with a
long-term price adjustment (in the short-
term option, the increased prices in the non-
shocked sectors do not become higher input
prices for all sectors).
Although the model has been developed
for the Middle East and North Africa region,
with a focus on oil prices such as in Jordan
(Atamanov, Jellema, and Serajuddin 2015)
and Libya (Araar, Choueiri, and Verme 2015),
it can also be applied to energy, food, or
water subsidies and accommodates linear
and nonlinear pricing for the direct eects,
assuming some simplifying assumptions.
The IMF has also developed a set of publicly
accessible Stata do-files that estimate the
direct and indirect effects of indirect tax
or subsidies, using the price shifting model
described previously.31 Both tools provide a
rapid analysis, including the indirect eect of
price increases on household welfare.
While these ready-made tools are already
available, the needs of the country engagement
might be such as to require and justify the
expense of time and eort to conduct a finer
modeling of the impacts through custom-
made programs. This is especially the case
for networked utilities with multiple tari
schedules for residential customers. In these
circumstances, ready-made products, which
lend themselves to rapid analysis, might
disregard some of the specificities of the taris
that apply to the most vulnerable households
and provide a poor approximation of the
impacts that would matter the most for the
analyst. Regarding the modeling of behavioral
responses, a strong distributional pattern of
consumption may exist as evidenced by Zhang
(2011) in Turkey where price elasticities for
electricity consumption are much lower for
lower-income households than for wealthier
ones. As in Zhang (2011), most of the tailored
analyses rely on the estimation of the loss
of consumer surplus, or provide the upper
bound (Laspeyres Variation) and lower bound
(Compensated Variation) of the welfare
impact, with constant consumption for the
upper bound and post-increase consumption
19 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
with a price-elasticity assumption for the lower
bond (see a selection of studies in Annex B).
Note also that, depending on the social
assistance programs available in the country
and the dialogue on whether and how they
can be adapted to cope with the impacts
of reforms, the existing social assistance
schemes might require tailored modeling
(as they might require tailored modeling
for possible modifications of the existing
schemes). In coordination with the Social
Protection specialist working on those issues,
it might be agreed that creating customized
files for both the poverty impacts and the
social protection response might be the most
ecient way of organizing the work.
MEASUREMENT CHALLENGES
The application of the methods discussed
in this note is crucially dependent on the
quality, reliability, and comprehensiveness of
the information contained in the household
surveys. This section summarizes some of the
most common diculties when deriving the
key variables needed for the analysis of ESR,
and points briefly to additional data sources
that might help in addressing some of these
challenges.
LIMITATIONS IN THE ENERGY
SPENDING VARIABLES
The distributional analysis discussed in this
note typically relies on household expenditure
surveys or other surveys that collect
disaggregated expenditure information (such
as a Living Standard Measurement Survey
with an expenditure module). This implies
that information on household-level spending
on energy will be available,32 although not
necessarily with the required level of detail. It
might not be possible, for example, to obtain
disaggregated information on a given energy
source, such as when fuels expenditure are
presented as an aggregate, or residential
energy spending is aggregated with other
housing expenditures.33
Similarly, the lack of information or ambiguity
regarding the time reference period for
dierent energy expenditures might make it
dicult to extrapolate the expenditure to an
annual amount (or averaging to a monthly
one) to make it consistent with the total
expenditure variables used. It is not uncommon
for questionnaires to adopt vague reference
periods such as “winter and summer” or
“heating period” (the questions would then
ask for typical expenditures over such periods).
Finally, the seasonality of expenditures might
be critical where heating and cooling needs
are significant.34 Even surveys that are run
over the entire year might not be released
with details of inter-year variation, so that
even if average values adequately capture
yearly averages, it might be hard to estimate
correctly some key variables that depend on
monthly consumption. This is most evident in
the case of utilities when block taris are in
place and there is strong seasonality. In such
a case, average monthly consumption might
fall in a tari block that diers from either
the tariff block of the peak consumption
season (for example, gas for space heating
in winter, electricity for cooling purposes for
the summer) and of the low consumption
season, or both.
These challenges are compounded when
seeking to compare across countries if survey
questionnaires dier.
Addressing such challenges requires
some creativity. It might require relying on
complementary sources to triangulate and
205. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
possibly impute variables (for example, when
ratios of consumptions in dierent periods
of the year are derived from administrative
data and used as a basis for imputation) or
reducing the ambition of the analysis. For
example, for the purposes of assessing energy
aordability when the required expenditure
data is not available, but some information
(even if aggregate by quintile) on overall
expenditure exists, one can at least calculate
the expenditure share of a “minimum energy
bundle” defined as covering essential needs
(for example, in electricity sector reform
focusing on the costs of using several light
bulbs, a fan, and a fridge) to have a sense of
aordability. In the case of liquid fuels, for
example, the estimate of the basic bundle can
be informed by assumptions or information on
the amount of kerosene needed for lighting
or LPG used in cooking (where use of LPG
is widespread even among the poor). It is
also important to look at total expenditures
on energy, and not just expenditures on the
subsidized forms of energy, to assess whether
meeting basic energy needs is aordable.
In addition to helping in approximating an
assessment of aordability, the costs of a
minimum energy bundle can be an input
in approximating a rough distribution of
the household-level direct benefits of the
subsidies. Finally, depending on resources, the
time available and the characteristics of the
survey (so that data quality is not aected),
one could consider collecting more data by
adding ad hoc energy modules to existing
household surveys.
CHALLENGES IN EXTRAPOLATING
ENERGY QUANTITIES CONSUMED
FROM ENERGY SPENDING DATA
As discussed above, since household surveys
typically do not include information on the
quantities of energy consumed, those need
to be extrapolated from energy expenditure
data. Special care should be taken in ensuring
that the extrapolation is based on nominal
expenditures, rather than on expenditures
that have been adjusted for dierences in
cost of living across regions.
The reliance on energy expenditures implies
that all possible concerns for the way energy
expenditures are reported, as described above,
will aect such an extrapolation. In addition,
additional challenges might arise, such as
the following:
• Ocial prices may be lower or much lower
than the actual prices paid due to fuel
shortages. Absent a price survey at the
same time as household expenditure data
are being collected, it is not possible to
back-calculate quantities from expenditures.
If fuel shortages are driving higher prices,
the actual prices paid can be location- and
time-specific, with large variations from
purchase to purchase, making it virtually
impossible to track quantities based on
expenditure data.
• Prices may be location-specific, and for
certain energy types also dependent on
the volume purchased. For example, LPG
purchased in large quantities (such as
large-cylinder refills) typically has lower unit
prices, but the lumpiness of each purchase
makes it unaordable for the poor.
• Even if the prices paid are precisely known,
if fuels with dierent prices are lumped
together in the household expenditure
survey—for example, dierent grades of
gasoline lumped into a single category
called gasoline, gasoline and diesel lumped
into a single category called automotive
fuels, and natural gas and LPG lumped into
21 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
a single category called gas—again it is not
possible to calculate quantities consumed.
• If an energy source is paid for less frequently
than the recall period in the household
expenditure survey, a large number of
households would record zero expenditures.
A classic example is a recall period of a
month when LPG sold in large cylinders is
refilled less frequently than every month. This
has resulted in a large number of households
citing LPG as their primary cooking fuel in
Mexico reporting zero expenditures on LPG
(Kojima, Bacon, and Zhou 2011).
• Quantities consumed may be curtailed by
energy shortages, which in turn may be
caused partially or even solely by energy
subsidies (see Good Practice Note 1). If
subsidy removal can largely eliminate energy
shortages—which is possible with liquid
fuels—actual consumption may rise, even
among the poor, to meet their basic needs
(say for lighting with kerosene) and especially
among the better-o. Depending on which
energy shortages are being addressed,
consumption of other forms of energy may
fall. For example, if electricity was previously
rationed because tari levels were too low to
enable the utility to purchase adequate fuels
for generation, and higher taris after the
subsidy reform enable the utility to purchase
more fuels and eliminate rationing, those
households previously purchasing diesel
for backup generators may switch back
to grid electricity, substantially increasing
electricity consumption and correspondingly
decreasing diesel consumption.
• Network energy presents additional
challenges:
• If there are multiple schedules of
regulated tariffs for households, it
would not be possible to back-calculate
consumption, unless the household
expenditure survey asks which schedule
the household subscribes to, which
household surveys almost never do.
For example, Mali has a total of 27
schedules for residential consumers
depending on the level of service
(Kojima and others 2016), whereas the
household expenditure survey provides
no information on the household’s tari
schedule, making it impossible to back-
calculate quantities consumed. As an
example of dierentiated utility taris,
Ukrainian gas taris are dierentiated by
categories defined on the basis of the
appliances used by the household and
by consumption levels; electricity taris
are dierentiated between rural and
urban areas as well as consumption; and
district heating taris dier by location,
being a function of the local provider.
• Unlike liquid fuels, for which payments
are made at the time of purchase,
utilities issue bills at fixed time intervals,
and late payments are allowed or
tolerated to varying degrees. As a result,
payments made at one point in time
may be a poor reflection of the quantity
consumed: expenditures may be low
on account of underpayment, or high
on account of making up for payment
arrears. The 2012 National Survey of
Household Income and Expenditures
in Mexico asked a series of questions
to understand arrears, including when
the last payment for electricity had
been made, while the 2005 Integrated
Sample Household Budget and Labor
Survey in the Kyrgyz Republic asked for
the quantity of electricity consumed,
the amount billed, and the amount paid
for three successive months as well as
the amount of subsidy received. Most
225. DOING DISTRIBUTIONAL ANALYSIS IN PRACTICE
household surveys, however, simply
ask how much the household paid
during the recall period. Asking just
one question about how much the
household paid over a fixed period of
time could under- or overestimate (the
latter if past debts are being repaid to
utilities) monthly expenditures. Cross-
checking household survey data against
data from the utilities could indicate the
magnitude of these problems and ways
of adjusting data for a more accurate
picture. Lampietti and Junge (2006)
combined billing and payment records
from the utility and merged them with
household survey data to address recall
errors, under- and over-reporting, and
the presence of arrears, which enabled
more accurate estimation of current
and historical electricity consumption
as a function of household income and
other characteristics. However, such an
approach would entail considerably
more work, as well as data gathering
challenges.
• The converse of the above is prepaid
metering, which is increasingly popular
in Sub-Saharan Africa. With prepaid
metering, households pay in advance
for energy they plan to consume in the
future. Depending on how little or how
much they pay, the expenditures may
bear little resemblance to the quantity
consumed during the recall period.
Further, few household surveys, if any,
ask if the household has a prepayment
plan, making it impossible to tell what
the expenditures represent.
• Each household may not be individually
metered and in a timely manner. If
several households are connected to
a single meter, or if households are
billed according to estimation (as in
a number of countries in Sub-Saharan
Africa), expenditures are not correlated
well with the charges calculated from
tariff schedules, and in some cases
expenditures and theoretical bills based
on tari schedules are virtually delinked.
• Some or all of the utility bills may be
covered in rent or by employment
benefits. In such cases, information on
neither expenditures or consumption
may be available.
• If every household is accurately and
individually metered, utilities can provide
the requisite data. However, utility data
have no information on total household
expenditures, and matching utility data
with household expenditure data by
household identity is virtually impossible.
Moreover, if there are multiple dierent
utilities covering dierent regions of
the country (as in Namibia and South
Africa), consolidating the information
from different utilities alone would
require considerable work, even if every
utility is willing to provide the data.
In the absence of requisite data, vastly
simplifying assumptions have to be made to
estimate the amount of price subsidies each
household receives.
METHODOLOGICAL CHOICES
IN CONSTRUCTING KEY
VARIABLES OTHER THAN ENERGY
CONSUMPTION
While the findings are extremely sensitive to
the way energy consumption is measured,
the construction of other variables needs also
careful thought.
Selection of a Welfare Indicator. Identifying
an appropriate welfare indicator is essential
23 GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES
AND THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
for the distributional analysis, independently
from the type of reform to be simulated. The
welfare indicator is used to rank households
according to their living standards. When
using a nationally representative household
expenditure survey, the indicator already
exists and is usually the one used for poverty
analysis—this might be total expenditures
for consumption per capita or per adult
equivalent, regionally and temporally adjusted
or total income, similarly adjusted.
Creation of the Energy Shares Indicator. As
mentioned above, when examining energy
expenditures, it is important to ensure that
those are taken in nominal terms, rather than
being adjusted for spatial dierences in costs
of living. Similar concerns aect the calculation
of energy shares for aordability purposes.
If an unadjusted expenditure aggregate
is not already available in the survey, the
analyst might need to reconstruct it using
all expenditures.
In addition, where it is possible to separate
out imputed values, computing shares out of
cash-only household expenditures, as well as
out of an all-inclusive household expenditures,
would be useful to provide a full picture of
the welfare impacts of higher prices. In the
case of low-income pensioners who fully own
their homes, for example, an aordability
assessment might reach dierent conclusions
if considering expenditure shares based on
cash expenditures only or on a full expenditure
measure.
Other Data Needs. Finally, it is worth
mentioning some of the other data sources
that can help fill gaps in the household survey
information available; triangulate some
indicators, especially for access, service quality,
and payments pattern; and identify seasonal
effects or for further simulating indirect
eects. Table 2 contains a list of various data
sources that teams can use depending on their
reliability, accuracy, and relevance.
TABLE 2: List of Additional Data Sources
Utility data Detailed information on payment and expenditures for formal consumers, data on
tari structure, costs, quality, payment recovery, and consumption. As such, utility data
is extremely relevant to assess the impacts of an increase in the cost of fuel or energy
services.
However, utility data do not give any information on the household socioeconomic
context and are available only for formal consumers. Utility data is therefore irrelevant
for households drawing electricity from ocially connected households or for
households with illegal connections.
National
statistics
Administrative sources, such as utility data from the Ministry of Energy or data on
social assistance recipients, poverty, and socioeconomic development from the
ministry in charge of social policy, can help provide context and crosscheck findings
from the survey.
International
statistics
Some specialized databases, such as the tari database of the Energy Regulator
Regional Association for Europe (https://erranet.org/knowledge-base/tari-database/)
can oer some simple cross-country comparisons (based on average taris in this
case).
Qualitative
assessment
Complements quantitative data by illustrating impacts beyond monetary ones. Can
informing the design of new survey modules. For more info, see Good Practice Note 4.
Census Data on energy access and housing characteristics can help provide general context
for the analysis.
24
ANNEX A: SAMPLE OUTLINE FOR A REPORT ON THE DISTRIBUTIONAL ASSESSMENT OF ENERGY PRICE INCREASES
ANNEX A: SAMPLE OUTLINE FOR A REPORT ON
THE DISTRIBUTIONAL ASSESSMENT OF ENERGY
PRICE INCREASES
BACKGROUND
1 | General background on poverty and severe poverty
1.1.1 | Table: Incidence of, by region, urban or rural, etc.
1.1.2 | Table: Profile of special groups of interest (if discussed in policy dialogue,
for example, unemployed or pensioners)
Info needed:
• National poverty lines and details on local consumption aggregate
• Special group of interest—particularly if emerging in social assistance
dialogue
2 | Background on the energy sector—note that the information on prices and
structure is needed for subsequent steps of the analysis, but might have been
collected under other “modules” of the work; other items listed that might provide
useful complementary information to the analysis of the distributional impact
on residential customers clearly do not apply to all energy sources and will be
omitted depending on the scope of the work.
2.1 | Background on reforms in the sector and how it aected prices over time—
details on energy price-setting mechanisms
2.2 | Current prices for major sources of energy
2.3 | Details on the type of subsidies available, eligibility criteria, amounts permitted
per eligible beneficiary, and delivery mechanism on paper
2.4 | Subsidy delivery mechanism in practice and associated consequences
2.5 | Cross-subsidies between consumers (industrial/residential) and among energy
sources; information on non-collection of network energy bills; taxes and
other fees where they are not captured in the schedules of regulated taris
2.6 | Costs of connection and metering (needed if access is not universal or if
each household is not individually metered)
25
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
CORE ELEMENTS
3 | Patterns of energy use
3.1 | Use of dierent sources (by poverty status, quintile, for selected groups)
3.2 | Access
3.3 | Details on use of dierent types of energy-consuming assets or appliances
by dierent groups.35
3.4 | Evidence on seasonality if available from either the household survey or
administrative sources
Info needed:
• Evidence from household surveys
• Complementary evidence from qualitative analysis, particularly on
seasonality if not covered by quantitative data
4 | Energy expenditures (possibly evolution over time if multiple data points)
4.1 | Table: Share of energy expenditures in total HH expenditures (by poverty
status, quintile, for selected groups)
4.2 | Table: Share of energy expenditures in monetary HH expenditures (by
poverty status, quintile, for selected groups)
4.3 | Table: Composition of energy expenditures by source (by poverty status,
quintile, for selected groups)
5 | Quantities consumed
5.1 | Average quantities consumed (by poverty status, quintile, for selected groups)
5.2 | Quantities consumed in dierent seasons if relevant (say for networked
energy if there are significant cooling and heating needs, depending on
the season).
6 | Performance of current policy measures.
6.1 | Distribution of the price subsidy
6.2 | Additional information to what is listed above under patterns of energy use
and which can help understand such findings. In the case of utilities, for
example, details on the block taris: distribution of households by block,
share of households that consume within blocks that are priced below
average (by region, quintile, poverty status, rural/urban—if cross-subsidization
between residential consumers). Where subsidized fuels are rationed,
comparison of the ocial expenditure on the rationed fuel—for example,
26
ANNEX A: SAMPLE OUTLINE FOR A REPORT ON THE DISTRIBUTIONAL ASSESSMENT OF ENERGY PRICE INCREASES
if a household is entitled to 5 liters a month of subsidized kerosene at a
discounted price—with actual expenditures reported by eligible households
would be another example.
6.3 | Breakdown of subsidy received by energy source, by quintile and group
of interest
7 | Distributional impact of tari increases
7.1 | Table: Changes in energy expenditure share (by quintile and for groups
of interest)—this can be by energy source if looking at multiple increases,
or overall. If calculating both direct and indirect eects, they might be
presented in dierent tables.
7.1.2 | As background for calculating the indirect eects: Share of spending
on energy intensive items (for example, food, which, is very sensitive to
transport costs; passenger transport) by quintile and group of interest
7.3 | Table: Poverty impact (overall, and for groups of interest)
7.4 | Incidence of the poverty increase
7.5 | Energy poverty impact (where relevant or where there is information to
compute the eect)
7.6 | Evidence from focus groups on coping mechanisms and behaviors, if available
Info needed:
• Evidence from qualitative analysis
• Elasticity estimates if available—evidence from focus groups can
help triangulate
POSSIBLE EXTENSIONS, DEPENDING ON NEEDS OF THE
DIALOGUE AND DIVISION OF LABOR WITHIN THE TEAM
8 | Energy aordability—aordability is one of the criteria adopted by the Multi Tier
Framework (MTF) to evaluate “usable” access to energy.36 While in the MTF, it
does not lend itself to a simple binary classification, it is worth noting that in the
framework, aordability is defined in the case of electricity as the basic service
not costing more than 5% of income, while in the case of cooking, it requires that
the levelized cost of the cooking solution absorb less than 5% of income. Such
estimates can form the basis of a summary aordability indicator (analogous to
the concept of energy poverty adopted by countries in the European Union). As
an alternative energy shares can be analyzed without resorting to the creation
of a binary variable capturing whether energy is not aordable.
27
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
8.1 | Aordability—share of energy expenditures in total HH expenditures: incidence
of “unaordable energy” or energy poverty across groups, its correlation
with consumption poverty, sensitivity of estimates to the adoption of a
dierent threshold to defined energy “not aordable”
8.2 | Profile of those with high energy shares or without access to aordable
energy as identified before, by user, defined by “main type of energy source”
8.3 | Complementary info from qualitative analysis
Info needed:
• Evidence from household surveys
• Local definition of energy poverty if one or several have been adopted
or are under consideration
• Complementary evidence from qualitative analysis on other concerns
that households have: debt arrears, lack of metering and unfairness
of bills, quality of service
9 | Analysis of current programs and descriptives to inform the overall discussion
on compensation options. As discussed in the text these additional tabulations,
data permitting, can be an input in a full assessment of existing social protection
measures and in the discussion of which compensation options might be pursued.
9.1 | Table: Distribution of direct subsidies or energy-related transfers and social
programs if they exist by quintile, poverty status, and energy poverty—
coverage, distribution of beneficiaries, distribution of benefits, generosity.
9.2 | Table: Distribution of other major social assistance programs by quintile
poverty status, energy poverty—coverage, distribution of beneficiaries,
distribution of benefits, generosity
9.3 | Table: Share of household expenditure spent on other items under discussion
as part of the design of broader compensation options—for example, food
or other basic items for which a value added tax (VAT) reduction might be
considered, out-of-pocket expenditures in health or education if (targeted)
measures to decrease those are being considered.
Info needed:
• Information on safety nets, earmarked energy subsidies (possibly to
be collected by a local consultant if not available)
• Information on eligibility criteria for social assistance (if transfers are
known to exist, but are not captured by the surveys)
• Information of cost recovery tari
28
ANNEX A: SAMPLE OUTLINE FOR A REPORT ON THE DISTRIBUTIONAL ASSESSMENT OF ENERGY PRICE INCREASES
9 | 10. Stylized policy simulations on compensation costs for various tari increase
scenarios and/or highlighting impacts on groups of interest (for example, the
poor, the bottom 40 percent of the population and those defined as “vulnerable”
in the relevant legislation)—as discussed above these stylized simulations can
be a useful input in the overall dialogue, for example, helping provide an order of
magnitude for the fiscal costs of dierent price increase scenarios and of course
highlighting their distributional implications.
N.B. All simulations to be evaluated for their impact on poverty, incidence,
concentration of benefits, energy share spent by dierent quintiles.
Base case: Distributional impact of a given set of increases.
Simulation 1—reallocating resources through social assistance (least cost solution):
Consolidating energy-related transfers into one budget (or another discretional
or appropriate budget) and distribute it uniformly per household to
a. the bottom x% of the distribution
b. those who are receiving major social assistance programs (independently
of the quintile, poverty etc.)
Further simulations could focus on ballpark estimates of mitigating impacts through
energy eciency measures or other measures discussed in the policy dialogue.
29
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
ANNEX B: SELECTED STUDIES FROM THE WORLD
BANK AND THEIR MAIN METHODOLOGICAL ISSUES
A number of resources documents the welfare impact of energy sector reforms in countries
around the world, including Coady, Flamini and Sears (2015), and the Asian Development
Bank and Global Subsidies Initiative assessments of subsidy reform in Asia (https://www.
iisd.org/gsi/fossil-fuel-subsidies/modelling-impacts-fossil-fuel-subsidy-reform-asia).
This annex does not seek to provide a comprehensive guide to the evidence, but
rather to examine a few examples of country studies, which might provide a picture
of the type of analytical challenges encountered and how they have been addressed.
ENERGY SUBSIDY REFORM IN ARMENIA. SEE KOCHNAKYAN AND
OTHERS (2013).
In order to bring the prices to cost recovery, electricity and gas prices have been
increased in steps from 2009 to 2013—electricity price by 17% and 28% and gas price
by 37% and 18% (see figure B1). In parallel, in order to mitigate the adverse impact on
vulnerable household, a discounted gas price has been applied since 2011 to Family
Benefit recipients for a total gas consumption up to 300 m3/year (increased to
450 m3/year in 2013). As new investments are planned for the supply of electricity,
FIGURE B1: Evolution of the Average Prices for Gas and Electricity for Residential
Customers Armenia
0
20
40
60
80
100
120
140
160
180
200
0
5
10
15
20
25
30
35
40
2000/1
2000/3
2001/1
2001/3
2002/1
2002/3
2003/1
2003/3
2004/1
2004/3
2005/1
2005/3
2006/ 1
2006/3
20 07/1
20 07/3
2008/1
2008/3
200 9/1
200 9/3
20010/1
2010/3
2011/1
2011/3
2012/1
2012/3
2013/1
2013/3
Gas prices in AMD/m3
Electricity price in AMD/kWh
Electricity in real term s (2000) Electricity in current terms G as in real terms (20 00)
Gas in current term s Index CPI 2000
Source: ERRA database and author calculations.
30
ANNEX B: SELECTED STUDIES FROM THE WORLD BANK AND THEIR MAIN METHODOLOGICAL ISSUES
significant electricity tari increase will occur in the coming years (from 40% to 240%
depending on the generation and the financing scenario).
Heating Source and Energy Expenditures
Gas is the main heating source for half of Armenian households; 44% of the poorest
quintile and 57% of the wealthiest relied on gas for heating in 2012 (80% of the
population has access to gas and 72% of the Family Benefit recipients). Wood is also
an important source of heating (30% of all households and 36% of the poorest) along
with electricity (15% in average). Energy expenditure as a share of total expenditures
reaches 9% in average (gas represents 4.6% and electricity 4%), with a slight distributive
pattern (8.7% for the wealthiest quintile and 9.9% for the poorest).
Distribution of the Subsidies
Prior to the gas price increase, the gas subsidy (when the tari was assumed to be
30% below the cost recovery) was highly regressive as 39.2% of the subsidy was
supplied to the wealthiest quintile and only 9% to the poorest. The new gas subsidy
(the discounted tari is 100 AMD/m3 instead of 156 for Family Benefit beneficiaries)
is quite progressive as an estimated 43% of the transfer is supplied to the poorest
quintile. However only 28% of the poorest quintile and 18% of the poor are covered
by such scheme. For reference, the current Family Benefit program, while limited
in coverage (only 36% of the poorest quintile is covered by the program in 2012), is
more progressive as 52% of the transfer goes to the poorest quintile.
Impact of the Reform
Using the household survey for 2012 the combined increase in gas and electricity
taris in 2013 is estimated to increase the share of the population living in poverty by
about 2.8%, and would reach 3.5% without a gas discounted price oered to Family
Benefit families.
Data and Methodological Issues Arising from the Armenian Context
Because of the gas price discount oered to a specific category of the population (Family
Benefit recipients) for whom the tari applied depends of the annual consumption,
the actual tari applied to these users is not precisely known. Using the regular tari
structure underestimates the subsidies provided to the Family Benefit recipients.
Wood cost is not correctly reported in the household survey, since only 1% of
those using wood to heat their homes report wood expenditures. This prevents the
highlighting of any substitution behavior in favor of wood use following the gas and
electricity price increases.
31
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
The simulation of direct energy subsidy supply relies on current coverage, which was
still very limited in 2012, and cannot model an extension of the program (only perfect
coverage of the poorest is modelized, thus without inclusion and exclusion errors).
ENERGY SUBSIDY REFORM IN MOLDOVA. BERHOLET (2015).
Since the 2012 energy price adjustment, costs have risen continuously and energy
taris have fallen short of cost recovery (especially for the gas bought in U.S. dollars).
The implementation of the 2015 tari adjustment decided by ANRE (Agenţia Naţională
pentru Reglementare în Energetică), the regulatory agency for gas and electricity, has
been suspended. Based on parameters influencing the end-user taris, the cumulative
electricity tari increase required is estimated to range between 42–61% from 2014
to 2016 and 73–113% from 2014 to 2020. The range of cumulative heat tari increase
is estimated to be 21–80% by 2016 and 30–78% by 2020. The consumer gas tari
increase is assumed to be 25% by 2016, based on the 2015 tari adjustment decided,
and the cumulative increase should reach 50% by 2020.
Energy Expenditures and Energy Mix
The Household Survey Data (2013) indicates that 80% of Moldova’s population spend
more than 10% of their expenditures on energy bills. On average, energy expenditure
is 17% of the total, which is high compared to other countries in the region (and 21%
among the poorest quintile and even 22% among isolated women). The spending
pattern for energy or “energy mix” is very heterogeneous, with urban households
spending 15% on utilities (central heating, gas, and electricity) while rural households
spend more on solid fuel (wood and coal). Central heating is only available in the
Capital City Chisinau and one secondary city (Balti) and is the main heating source
in both cities. Where central heating is not available, gas use for heating increases
with wealth. Energy consumption is highly seasonal with central heating and gas
expenditures twice the annual average during the first quarter. However, the share
of household resources spent on electricity remains rather constant across the year,
since few households rely on electricity for heating. Wood is often purchased ahead
of the heating season, during the third quarter.
Impact of the Reform
The estimated range of energy tari increases would increase the average share of
energy costs in total expenditures to 18–20% in 2016 and with projected economic
growth, the share would decrease to 17–18% in 2020. Within the estimated range of
energy tari increases, the poverty rate is expected to increase by 1.1–1.9 percentage
points in 2016 compared to a baseline and in 2020 by 1–1.5 percentage points. Wood
users are vulnerable to electricity tari increase, since they are poorer and thus more
vulnerable to poverty due to a price increase.
32
ANNEX B: SELECTED STUDIES FROM THE WORLD BANK AND THEIR MAIN METHODOLOGICAL ISSUES
Subsidies and Social Assistance
A heating allowance program complements the social assistance cash transfer
(remaining modest with 7% of the population covered in 2014—136,000 beneficiary
households—and only 12% of the poorest quintile.). It complements the Ajutor Social
program to compensate the poor for increased cost of living during 5 months of
heating season (flat monthly benefit of MDL 250 oered to all recipients of Ajutor
Social and to those households whose income is below 1.6 times the Guaranteed
Minimum Income).
Data and Methodological Issues Arising from the Moldovan Context
Seasonality is well managed in the household budget survey as for all energy
items, and current month’s expenditure is reported, but also the 12-month estimate
(except for electricity where only current expenditure is reported). The consumption
expenditure aggregate as per the Statistical Oce procedure includes annual average
for energy items except for the electricity (current month). Inconsistencies or missing
are corrected using the current expenditures and the median reference period.
Regarding the mitigation through the Social Program, a perfect take-up could cover
446,000 households. This is not easy to simulate using the household budget survey
because of all the eligibility criteria. Thus estimation is based on current coverage
or perfect coverage of the poorest. Also Municipal Heating benefits in Chisinau and
Balti are not reported in the survey (in Chisinau 33,000 households covered from
189,000 households targeted): The average monthly benefit paid during five months
in Chisinau to gas, wood, and coal users is MDL 450 and for central heating users
MDL 285, thus higher than the national Heating allowance. Average monthly benefit
in Balti during five months is MDL 200.
ENERGY SUBSIDY REFORM IN UKRAINE. BASED ON
UNPUBLISHED NOTES. SEE ALSO WORLD BANK (2013).
The Government of Ukraine (GoU) implemented substantial natural gas and heating
tari increases under the IMF’s Extended Fund Facility (EFF) arrangements, resulting
in a substantial decrease in implicit subsidies for residential users. The implicit
subsidies in the form of low end-user taris disproportionately benefited richer
households, hindered investments in the energy sector and, ultimately, proved to
be extremely costly for the national budget. As part of a three-year transition to
cost recovery, gas and district heating prices increased in 2014 (by 56% and 40%,
respectively), and in 2015 a major gas price increase has been applied (+450% with
a lifeline for households heating with gas, and consequently an increase in a district
heating tari of 70%). The devaluation of the hryvnia has eroded the impact of the
tari increase on gas, restoring a de facto universal subsidy on the 50% of yearly gas
consumption that does not fall into the lifeline. In May 2016, the transitional lifeline
33
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
tari for consumers heating with gas has been abolished and the tari switch to a
uniform tari. In parallel, electricity tari (increasing block) have increased twice a
year since 2015. As a consequence, the average increase between 2014 and 2016
reached 470% for gas, 155% for electricity, and 190% for district heating, while the
inflation reached about 70% during the same period.
Energy Expenditures
In 2013, poorer households spend a higher share of their resources on energy (almost
8% for the 30% poorest versus less than 6% for the upper 70%). 27% of the poorest
quintile in large cities, and 22% of the poorest quintile in small cities and rural areas
spend more than 10 % on energy (the national average is 14%). Expenditure patterns
reflect the sources to which households in dierent locations have access: households
in large cities mostly rely on district heating and gas, while in smaller cities and rural
areas—mostly on gas.
Subsidies and Impact of the 2015 Reform
Existing subsidies (subsidies from the taris) prior to the price increase are regressive:
in 2013, only 15% of the direct and quasi-fiscal subsidies are provided to the poorest
quintile while 24% of the wealthiest are covered. Post energy increase, had the
devaluation not occurred, the distribution of the gas subsidy would have been more
progressive (17.4% to the poorest and 17.5% to the wealthiest, instead of 16.8% and 18.4%
respectively with the devaluation). The lifeline per se cannot guarantee a progressive
distribution of the benefits, since there are many large users of gas also among the
poor. Those are poor households living in individual houses in rural areas, with very
inecient heating systems (limiting the possibility to adapt their consumption level).
As a consequence, the gas implicit subsidy in 2015 is quasi-neutral. In addition, the
most regressive subsidies are the ones supplied through the district heating tari;
they are still in place in 2015 (less than 11% of the district heating implicit subsidy
goes to the poorest quintile while 35% goes to the wealthiest). As a consequence,
the combined gas and district heating subsidies remain regressive even after the
2015 reform.
In contrast, social assistance programs, which are the main tool to address the
distributional impact of the tari increase, have become already much more progressive
than they were (only 12% of the Housing and Utilities Subsidies (HUS) are supplied
to the poorest quintile in 2014—where the coverage was very limited- and it would
reach 36% in 2015 already if the HUS program could be extended to all eligible
households). The formula for HUS calculation is based on the ratio of the cost of a
normative consumption bundle to income (the total cost for a normative consumption
should not exceed 15% of disposable income for a household earning the equivalent
of 2 Minimum Salary per capita; it decreases for lower income levels according to a
sliding scale). When energy price increases, more households become eligible for
34
ANNEX B: SELECTED STUDIES FROM THE WORLD BANK AND THEIR MAIN METHODOLOGICAL ISSUES
the HUS as their consumption (normative as well than actual) increases and exceeds
the threshold. Thanks to the sliding scale, about 92% of the bottom decile is eligible
against 14% in the top decile. In parallel with the extension of the coverage, the
generosity also increases. The updated administrative data in 2016 shows that 44%
of the households were actually covered (from 4% in 2014). In order to accompany
the program extension, the norms have been reduced in 2016 and 2017. Recent
simulations conducted in 2017 allow the estimation of the fiscal needs for dierent
scenarios of changes in the eligibility criteria, the distributional impact of dierent
alternatives and the poverty impact of reducing transfers under those scenarios.
Data and Methodological Issues Arising from the Ukrainian Context
The Housing and Utility Subsidies (HUS) are based on the normative consumption,
depending on the surface area of the housing and the number of persons. For each
household eligible, a proportion of this normative consumption is covered, not the
actual consumption. Heating being the main source of energy consumption, norms
vary for heating and nonheating purposes, thus across users and across the year.
Although available data are annual averages, subsidies at the household level must
be calculated separately for heating and nonheating seasons then aggregated.
Also because the energy price increases are significant and because they are partly
compensated by HUS, no elasticity assumption can be made. All simulations are
conducted holding constant consumption levels and considering total expenditures
before HUS. Finally, the analysis does not include the possible eects of nonpayment,
which might increase with further restrictions on HUS eligibility.
ENERGY SUBSIDY REFORM IN BELARUS. ZHANG AND
HANKINSON 2015.
Despite nominal increases, residential gas, electricity, and district heating taris
have not kept pace with rising production costs. District heating production costs
in particular have risen sharply since 2005, driven by the cost of imported natural
gas and depreciation of the Belarussian ruble against the U.S. dollar. Residential
taris are currently at 11–25% of cost-recovery levels, depending on the provider
and technologies used. Belarus still benefits from preferential tari for gas imported
from Russia (the US$263/tcm applied in 2011 was even reduced to US$163/tcm in
2012), but in case of an import gas price hike, financial losses in the district heating
sector would more than double. The fiscal costs borne by the government reached
2.5% of GDP in 2012 and benefit disproportionately to wealthier households and the
cross-subsidies imposed on nonresidential customers increase the cost of energy in
all other sectors and indirectly to households.
35
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
Energy Expenditures
In 2012, since most households are connected to district heating (almost all of them in
urban areas and more than half of them in rural areas), district heating and electricity
each account for 40–50% of the total expenditures on energy. Energy as a share of
total expenditures accounts for less than 5% of the total expenditures in 2012 and
decreases with wealth (from 4.9% for the poorest quintile to 1.6% for the wealthiest).
Impact of Higher Taris
Study conducted using Household Survey from 2012 shows that higher taris will
have the most impact on the poorest households—and households in rural areas.
Urban customers in the poorest quintile can be expected to spend, on average, 21% of
their incomes on district heating services if taris are set at full cost-recovery under
a uniform cost-recovery price scenario. If the tari is dierentiated based on each
provider cost-recovery, rural customers in the poorest quintile would spend 23% of
their income on district heating services for such scenario.
Social Protection
The fiscal savings could be used to fund social protection mechanisms with better
coverage and targeting. Options include expanding or topping-up the existing Public
Targeted Social Assistance Program (GASP) and/or refining the Housing and Utilities
(H&U) subsidies program discontinued in 2009.
Data and Methodological Issues Arising from the Belarussian Context
Since the tari is calculated according to information on metered and unmetered
customers, as well as ownership of equipment, such as water heaters and gas stoves,
these variables are required to identify the tari applied. The dierentiated tari per
category of customers and per season would aect the distribution of the implicit
subsidy and the distributional impact of the price increase to cost-recovery.
Remark: In Belarus household surveys, district heating expenditure was included in
the “sum of payment for the use of living quarters, the maintenance and the public
services,” which regroup district heating, water, waste collection, and other building
items. This item is further disaggregated using the ratios for typical households with
the normative consumptions.
ENERGY SUBSIDY REFORM IN JORDAN. ATAMANOV, JELLEMA,
AND SERAJUDDIN 2015.
The simulation of petroleum and electricity removal uses the SUBSIM tool and
assesses the direct and indirect impacts of the reforms conducted in 2013, adjusting
36
ANNEX B: SELECTED STUDIES FROM THE WORLD BANK AND THEIR MAIN METHODOLOGICAL ISSUES
for economic growth (GDP per capita nominal), inflation (CPI), and for population
size changes, using 2010 Household Expenditure and Income Survey.
Impact of the Subsidy Removal
The study simulates the full removal of petroleum and LPG subsidies and assesses
the direct impact of two scenario on the per capita consumption of households
per quintile: full subsidy removal with no government mitigation measure, and full
removal combined with a cash transfer program that accompanied the petroleum
price increase. Both scenarios are conducted using the historical data on butane and
propane from the utility Saudi Aramco contract price to calculate the ecient LPG
price. Poverty would be expected to increase by 1.6 percentage points. Indirect impacts
of petroleum product price increase are estimated using a Jordanian Input/Output
table (I/O - 2010) with HIES data and used the disaggregated production figures of
the state-owned refinery as a proxy of the industry-wide petroleum—production mix
since the Jordanian I/O does not have disaggregated-by-type petroleum products
statistics.
The study also simulates the removal of electricity subsidy along three scenarios (tari
increases as planned for 2015, full removal of electricity price subsidies, progressive
removal of the subsidy for all consumer categories but for the first two quintiles).
Full elimination of electricity subsidies is expected to increase the poverty rate by
2.4 percentage points.
ENERGY SUBSIDY REFORM IN MOROCCO. VERME AND EL
MASSNAOUI. 2015. SEE ALSO KOJIMA 2016.
Food and Fuel Subsidy Reform
Morocco’s subsidy system had been in place since the 1930s, with the aim of
stabilizing prices for consumers, in part to protect vulnerable population groups,
and to promote domestic industries. By 2007 or 2008 the rising fiscal pressure was
out of the government’s control. Food and fuel subsidies reached 6.6% of GDP by
2012, with the bulk (70%) going to energy products. The subsidy system was no
longer achieving its objectives, as much as 75% of energy subsidies were going to
the richest 20% of the population.
Launched in June 2012, the reform process was introduced in phases over three years
until full price liberalization. In 2013, the government introduced a partial fuel price
indexation mechanism for diesel, gasoline, and industrial fuel oil. Indexation was based
on a moving average of the previous two months and entailed automatic adjustment
of domestic prices when the dierence between the market reference price and
the domestic retail price exceeded 2.5%. In 2014, the Government of Morocco first
removed subsidies for gasoline and industrial fuel oil, then the fuel subsidy for power
37
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
generation. It introduced quarterly price increases to reduce the unit subsidy for diesel
and entirely eliminated diesel subsidies by the end of the year. In December 2014, the
Government of Morocco signed an agreement with petroleum product suppliers. This
paved the way for a transitional period between January and November 2015, during
which prices would be announced by the government twice a month. LPG remains
excluded from the price deregulation though, and annual subsidies have averaged
DH 13–14 billion (US$1.5–1.7 billion). In addition, an agreement signed between the
government and the national oce for electricity and water provides for gradual retail
price increases of electricity over three years to match production prices to the sale
prices, which entails operational cost savings in addition to price rises of about 3.5%
annually. Only the price of the first consumption bracket is maintained unchanged
for low-consumption households (less than 100 kWh per month).
In parallel, the government expanded existing targeted social protection programs
(support to school-age children and medical assistance for the poor), introduced
new social protection programs in support of low-income widows and the physically
disabled, and provided support for the public transport to compensate for the cost
of higher fuel prices and limit fare increases (see figure C1).
The government implemented the policy as agreed, and deregulated prices at the
end of 2015.
The compensation fund budget decreased from 72% between 2012 and 2016.
FIGURE B2: Evolution of the Total Compensation Budget
0.00
1.00
2.00
3.00
4.00
5.00
6.00
4.72 5.14
3.84
3.05
1.45
2011 2012 2013 2014 2015 2016
1.41
Source: Hassan Bousselmame, Réforme du système de subvention au Maroc. Communication June 2017.
38
ANNEX B: SELECTED STUDIES FROM THE WORLD BANK AND THEIR MAIN METHODOLOGICAL ISSUES
Simulation of the Welfare Eect of the Reforms Implemented between
January and October 2014
Direct eects (using household expenditure data)
The elimination of subsidies reduces welfare by about 1% on average with the impact
being larger for the poorest quintile (-0.61%) as compared to the richest quintile
(-1.07%). The transformation of the upper three blocks in volume-dierentiated taris
(VDTs) has been much more significant for water than for electricity, despite the fact
that the water sector had no subsidies and that taris per block have not changed.
Hence, the change in taris structure toward a VDT system can have an even greater
impact on welfare than the simple increase in prices.
TABLE B1: Welfare Eects of the 2014 Reform, Direct Eects, million Moroccan
dirhams
Electricity Water Gasoline Diesel Total Total
(% of expend.)
Quintile 1 -118.0 -94.5 -0.3 -1.1 -213.9 -0.61
Quintile 2 -241.4 -263.7 -1.4 -4.5 -511.1 -0.87
Quintile 3 -366.5 -462.5 -6.3 -20.6 -855.9 -1.04
Quintile 4 -490.8 -677.8 -17.6 -57.1 -1,243.2 -1.05
Quintile 5 -1,182.0 -1,221.2 -154.1 -502.7 -3,060.0 -1.07
Total -2,398.7 -2,719.8 -179.7 -586.0 -5,884.1 -1.01
Source: Verme and El Massnaoui (2015).
Indirect Eects Using I/O Data Combined with Household Data
With I/O tables, it is not possible to simulate price increases by product or by tari
block, given that the I/O tables are aggregated by sector. Therefore, averages prices
are used across products belonging to the same sector or across taris blocks.
The shock to the petroleum sector is a price increase of 11.15%, which is an average
of the price shocks applied to diesel and gasoline. The assumption here is that
gasoline and diesel have a similar weight in the I/O oil refining sector and that they
represent the almost totality of the sector. The price shocks applied to electricity is
2.1%, which is an average price increase across taris blocks weighted by the number
of households in each block.
Results of the simulations show that the relation between direct and indirect
eects varies significantly across products and across quintiles. When simulated
independently, indirect eects are 88% of the total for petroleum products and 37%
for electricity (see table B2). The relative weight of indirect eects is also dierent
39
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
across quintiles. Indirect eects on petroleum products are the quasi-totality of eects
for the first (poorest) quintile while they become 81.33% for the upper quintile. This
is understandable because the poor consume very little gasoline and diesel. Instead,
for electricity, indirect eects represent 30.1% of total eects for the first quintile,
and this share increases to 42.17% for the fifth quintile. That is because the coverage
of electricity is very large in Morocco, and many if not most of the poor consume
electricity.
TABLE B2: Indirect Eects of 2014 Reform (% of total eects)
Petroleum Electricity
Quintile 1 99.55 30.10
Quintile 2 98.87 30.31
Quintile 3 96.48 30.52
Quintile 4 93.43 33.89
Quintile 5 81.33 42.17
Total 8 7.79 36.55
Source: Verme and El Massnaoui (2015).
40
ANNEX C: THE MEASUREMENT OF ENERGY POVERTY
ANNEX C: THE MEASUREMENT OF ENERGY
POVERTY
Energy poverty, a measure of deprivation that seeks to capture aordability problems
as they relate to energy, is an indicator mostly used in Europe and Central Asia, where
energy bills can represent very high shares of total household income due to heating
expenses. The concept is enshrined in Third Energy Package of legislative proposals
for common rules for the internal electricity and gas markets of the European Union
(EU), adopted and entered into force in 2009, though not explicitly defined. While
only one-third of EU Member States (EU MS) have defined it, the Third Energy
Package also requires EU MS to formulate national energy action plans to provide
benefits through the social assistance system or to improve energy eciency for
the energy poor.
The concept is generally approximated through a dichotomous variable that captures
those who spend more on energy than a given share of their household budget,
typically 10%. The 10% threshold was originally defined in the United Kingdom to
measure “fuel poverty,” with reference to twice the median consumption of low-income
households, but has since been used rather mechanically as opposed to in reference
to a specific context. This measure of energy poverty is often contested as a basis
for targeting support to ensure energy aordability, since it includes an element of
preference. In this respect, a major contribution has been made by the Hills report in
the United Kingdom (Hills 2012), which suggests identifying households that have
both low income and high energy needs as those aected by energy poverty. The
report also emphasizes how energy (fuel) poverty per se is not just a facet of income
poverty, but a specific challenge requiring a dedicated policy strategy.
Another approach to measuring energy aordability is to focus on the share of
domestic energy expenditure in final consumption expenditure for the lowest quintile—
the share of energy-related expenditure in
total household expenditure for the poorest
20% of the population. Such a share can
be examined either by how it is aected
by tari increases, or in reference to the
share for another type of basic consumption
item (such as food, health, or education) to
give a sense of the competing priorities on
limited budgets such low-income households
might face.
The revised definition of fuel poverty
in the United Kingdom states that a
household is said to be in fuel poverty
if they have required fuel costs that are
above average (the national median
level) and, were they to spend that
amount, they would be left with a
residual income below the official
poverty line.
41
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
ENDNOTES
1 The note draws extensively on an earlier methodological piece focused on the analysis
of ESRs in the Eastern Europe and Central Asia region of the World Bank (Olivier 2016).
2 Note that some of these eects might be implicitly part of the reform—for example, if
ESR is accompanied by measures to increase access or improve accountability of local
utilities—or might be part of its “side eects,” such as by making local utilities solvent
and therefore capable of improving quality, or if they result in the elimination of energy
shortages.
3 Proponents of energy subsidies typically argue that energy subsidies help make
energy aordable for lower-income groups; facilitate access to improved household
energy sources (as opposed to traditional biomass), particularly in rural areas where
unsubsidized prices can become prohibitively high; facilitate the shift from more-
polluting to less-polluting fuels, such as from solid fuels or kerosene burned in wick
stoves to LPG (liquefied petroleum gas) or natural gas; or help shield the economy as
a whole from volatile energy prices. All of these points are hotly contested, since critics
focus on energy subsidies’ ability to achieve these objectives, as well as on their side
eects.
4 For example, in Indonesia, it has been estimated that the population that would enter
poverty following a reduction in fuel subsidies would increase by 0.4 percentage points
based on ocial price increases (Dartanto 2013). Impacts of about 5 percentage points
have been estimated for Egypt (World Bank 2005; Soheir, El-Laithy, and Kheir-El-Din
2009), but arguably these represent the very top of the distribution of eects, since fuel
subsidies there were both very high and pervasive. While estimates vary significantly
with the methodology used, the studies quoted appear to cover the spectrum of
poverty impacts typically found in empirical analyses.
5 For example, using a computable general equilibrium model, Burniaux and Chateau
(2011) show that a coordinated subsidy removal could reduce the competitiveness
of energy-intensive industries in certain economies (especially in Indonesia, Nigeria,
Venezuela, and the Middle East), which in turn would reduce employment in that sector
to the extent that labor is not a good substitute for energy inputs. Note, however, that
this drop in competitiveness would be accompanied by higher incomes resulting from
increased fossil fuel exports. In addition, EU experience suggests that reliable energy
supply and a productive labor force, wider access to markets, and so on, are significant
drivers of industrial competitiveness—so that an emphasis on fuel prices might only
prove reductive.
6 Some transitional measures (cash transfers) can exacerbate the inflationary impact
of ESR (see, for example, Clements and others 2013). Note that the finding on the
magnitude of indirect eects is likely to be significantly overstated due to the use of
input-output analysis, which is based on fixed coecients (that is, there is no scope
for substitution)—as recognized by Coady, Flamini, and Sears (2015). Estimates should
therefore be considered as short-term eects or upper bounds of the long-term eects
(see also Good Practice Note 7 for more details on these assumptions).
42
ENDNOTES
7 Recent evidence on the shift from fuels to biomass has been provided, for example, by
the recent policy pilots in India, involving switching from in-kind to cash benefits for
LPG and kerosene. Because of the specific design of those measures, household facing
poor banking facilities and other barriers to accessing the benefit dramatically reduced
their consumption of fuels (Lang and Wooders 2012).
8 Available estimates of the price elasticity of demand vary substantially by type of
fuel and with the level of income per capita in a country. A recent review (Dahl 2012)
reports estimates of between -0.11 and -0.33 for gasoline, and of between -0.13 and
-0.38 for diesel. The price elasticity for gasoline appears to be higher in richer countries.
The income elasticity of demand for fuels is much larger in magnitude than the price
elasticity. Vagliasindi (2013) reports that long-run elasticities are significantly higher
than short-run elasticities. Zhang (2011) estimates the price elasticity of demand
for electricity by dierent groups and finds that demand from poorer households is
significantly more inelastic than that from richer households.
9 In addition, the table does not capture complexities, such as policies that may lower the
price of LPG relative to kerosene for the very poor.
10 Consumption patterns in Eastern Europe and the former Soviet Union depart markedly
from those in most other regions, because of the essential need for heating in winter,
limiting the extent to which consumption of energy for space heating can be curtailed.
Another distinguishing feature of Eastern Europe and the former Soviet Union is that
natural gas and district heating networks may be extended to rural areas in some
countries, whereas natural gas pipeline networks in rural areas are virtually unheard of in
other regions of the world.
11 Similarly, in the case of networked utilities, subsidies often result in insucient
investment in the sector by the utility, which over time loses the ability to provide
services of appropriate quality. Improvements in service quality (for example, reductions
in blackouts) following removal of subsidies are not captured in the table.
12 Note that the two questions of “who benefits from the subsidies” and “who is going to
be mostly aected by their removal” are very dierent—it is commonly found, indeed,
that those that benefit the most are the better o, since they consume more, while
poorer groups are the most aected, since their consumption of subsidized goods
represents a larger share of their budgets.
13 In addition, the analysis of household survey data can also help quantify the resources
absorbed by the subsidies, that is, the size of the overall subsidy. Such an assessment
is partial, however, since it would cover only the household sector. A full quantification
of the size of subsidies in the energy sector is discussed in Good Practice Note 1. In
addition, while household survey data should provide an accurate estimate of the
resources distributed to households through energy subsidy programs, such estimates
should at best be considered an approximation, since at the very least they do not
include the cost of administration of the program.
14 The ADePT Social Protection module, for example, oers an easy way of calculating key
performance indicators of social protection measures and can be downloaded at http://
go.worldbank.org/1HHHLLELG0.
15 In the incidence literature, generosity is typically measured as a share of pretax and
pretransfer income (that is, at market income). It can also be done at post-transfer
income or post-consumption.
43
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
16 In certain countries, policy makers are interested in specific categories that are
predetermined as “vulnerable,” such as pensioners, the unemployed, and single parents.
17 Access should ideally be defined in terms of “the ability to avail energy that is adequate,
available when needed, reliable, of good quality, convenient, aordable, legal, healthy
and safe for all required energy services” (https://www.esmap.org/node/55526). This
definition, which is at the basis of the Multi Tier Monitoring Framework for the SGD 7
involves focusing on the energy services available to the household, independently of
the technology that provides it and recognizing that energy access refers to a spectrum
of service levels (http://www.se4all.org/sites/default/files/Beyond-Connections-
Introducing-Multi-Tier-Framework-for-Tracking-Energy-Access.pdf). While new tools
(such as new household surveys and modules that can be integrated into existing
surveys) are being deployed, however, existing surveys might allow the measurement
only of more simplistic binary measures of access.
18 An example might clarify. When trying to compare whether an existing targeted
program (Program A, targeted based on 3 indicators combined as a proxy for low
income) should be used to provide compensation, or whether a new program with
better targeting (Program B, at the moment only discussed in policy circles, targeted
based on a proxy constructed with 5 indicators) should be used, using data on real life
recipients of Program A versus data on those who would theoretically be recipients
of Program B biases the outcome in favor of Program B, as it implicitly assumes that
Program B would have perfect take up.
19 This loss of purchasing power, assuming a constant consumption, is expressed as a share
of total income and corresponds in nominal terms to the Laspeyres Variation, the amount
of money required for the household to maintain their initial level of consumption.
20 We do not retain here the notions of Equivalent Variation or Paasche Variation, since
these variations use as a reference the welfare or consumption level after the price
increase, and thus do not take into account the loss of welfare incurred by the reduction
of the consumption. These two measures are commonly used in the fiscal literature
where the reference period is the one after the change in price.
21 Note that estimates of the relevant elasticities are hard to come by. Zhang (2011) used
the consumer surplus change to approximate the welfare impact of an electricity
price increase in Turkey in 2008 after estimating the demand function, allowing the
elasticity to vary with income. She estimates an electricity own price elasticity of -0.08
for the poorest quintile and -0.5 for the wealthiest. These results confirm the finding
highlighted by the qualitative literature on Eastern Europe that poorer households have
a lower price elasticity, since they are closer to a subsistence level. Assumptions of
quasi-inelastic energy consumption for the estimation of the impact on poverty is thus
reasonable for electricity in richer countries, though it seems likely that other energy
sources (such as cooking fuels) in other contexts (such as Sub-Saharan Africa or South
Asia) would require dierent assumptions. In the upper part of the distribution, as well
as for most petroleum products, price elasticity is likely to be significant. Finally, note
that in cases of acute shortages, the price elasticity may even be positive.
22 Longer-term eects taking into account these adjustments would require the use of a
more sophisticated Computable General Equilibrium Model, also relying on an I/O table,
yet allow the behavior of firms and consumers to be fully flexible, as detailed in Good
Practice Note 7.
44
ENDNOTES
23 If energy prices have been frozen at a time of high inflation, as is often the case with
utilities, this step is all the more necessary.
24 SUBSIM, as described below, allows also the tari to be nonlinear, with increasing block
or volume-dierentiated design.
25 Those can be downloaded from http://www.imf.org/external/np/fad/subsidies/index.
htm, under “tools.”
26 Local knowledge, either through exploring the local press (including through such
resources as Factiva) or by relying on local experts, can provide evidence of rationing.
Such evidence might be particularly hard to identify if shortages are very localized.
27 The considerations made above on the diculties of accounting for rationing or spatial
disparities in prices would therefore be relevant also in this context.
28 When analyzing a price change resulting from government policies, the proportion
of Traded/Non Cost-Push sectors in the sectoral outputs can be ignored and the
framework can be reduced to a Cost Push and Controlled sectors (price change either
fully pushed onto output /retail prices or controlled by the government).
29 See Inchauste and Jellema (2016) for more detailed steps.
30 See endnote 26.
31 These do-files can be downloaded from http://www.imf.org/external/np/fad/subsidies/
index.htm, under “tools.”
32 In the analysis of residential utilities, the survey might collect information on dierent
variables, including actual cash expenditure for utility consumption, cash transfers
for energy consumption, and arrears and fines. If the focus of the analysis is on the
financial burden on households, the actual expenditure for consumption should be
used excluding cash transfers targeting energy consumption (whose impact should
be analyzed separately) and potential arrears or fines. Similar adjustments should be
conducted on the aggregate expenditure variables for consistency.
33 For example, in Eastern European surveys often district heating and hot water are
grouped under a single expenditure item.
34 Cooling needs might not be very relevant to the analysis of low-income countries if the
use of air conditioning is limited to few better-o households.
35 See endnote 17.
36 See, for example, http://www.se4all.org/sites/default/files/Beyond-Connections-
Introducing-Multi-Tier-Framework-for-Tracking-Energy-Access.pdf.
45
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
REFERENCES
N.B. Several World Bank sources have been quoted freely and adapted with the
permission of the authors. They are marked below with an asterisk (*).
Araar, Abdelkrim, and Paolo Verme. 2012. Reforming Subsidies: A Tool-kit for Policy
Simulations. Policy Research Working Paper, no. WPS6148. Washington, D.C.: World
Bank Group. http://documents.worldbank.org/curated/en/892071468331042804/
Reforming-subsidies-a-tool-kit-for-policy-simulations
Araar, Abdelkrim, and Paolo Verme. 2016. Prices and Welfare. Policy Research Working
Paper, no. WPS7566. Washington, D.C.: World Bank Group.
Araar, Abdelkrim, Nada Choueiri, and Paolo Verme. 2015. The Quest for Subsidy Reform
in Libya. Policy Research Working Paper, no. WPS7225. Washington, D.C.: World
Bank Group. http://documents.worldbank.org/curated/en/344571467980552949/
The-quest-for-subsidy-reforms-in-Libya
Atamanov, Aziz, Jon Robbert Jellema, and Umar Serajuddin. 2015. Energy
Subsidies Reform in Jordan. Welfare Implications of Dierent Scenarios.
Policy Research Working Paper, no. WPS7313. Washington, D.C.: World Bank
Group. http://documents.worldbank.org/curated/en/595761467998814982/
Energy-subsidies-reform-in-Jordan-welfare-implications-of-dierent-scenarios
Bacon, Robert. 1995. Measurement of Welfare Changes Caused by Large Price Shifts: An
Issue in the Power Sector. World Bank Discussion Papers, no. WDP 273. Washington,
D.C.: World Bank Group.
Bertholet, Fabrice Karl. 2015. Moldova—District Heating and Electricity
Tari and Aordability Analysis. Washington, D.C.: World Bank Group.
http://documents.worldbank.org/curated/en/101211468060253519/
Moldova-District-heating-and-electricity-tari-and-aordability-analysis
Burniaux, Jean-Marc, and Jean Chateau. 2011. Mitigation Potential of Removing Fossil Fuel
Subsidies: A General Equilibrium Assessment. OECD Economics Department Working
Papers. No. 853. Organisation for Economic Co-operation and Development, Paris.
Clements, Benedict, David Coady, Stefania Fabrizio, Sanjeev Gupta, Trevor
Alleyne, and Carlo Sdralevich (eds.) 2013. Energy Subsidy Reform: Lessons
and Implications. Washington, D.C.: IMF. http://citeseerx.ist.psu.edu/viewdoc/
download?doi=10.1.1.800.3440&rep=rep1&type=pdf
Coady, David, Moataz El-Said, Robert Gillingham, Kangni Kpodar, Paulo Medas, and
David Newhouse. 2006. The Magnitude and Distribution of Fuel Subsidies: Evidence
from Bolivia, Ghana, Jordan, Mali, and Sri Lanka. IMF Working Paper No. WP/06/247.
Washington, D.C.: IMF.
Coady, David, Valentina Flamini, and Louis Sears. 2015. The Unequal Benefits of Fuel
Subsidies Revisited: Evidence for Developing Countries. IMF Working Paper No.
WP/15/250. Washington, D.C.: IMF. http://www.imf.org/external/pubs/ft/wp/2015/
wp15250.pdf
46
REFERENCES
Cory, D., R. Gum, W. Martin, and R. Brokken. 1981. “Simplified Measurement of Consumer
Welfare Change.” American Journal of Agricultural Economics 63:715–17.
Dahl, C. A. 2012. “Measuring Global Gasoline and Diesel Price and Income Elasticities.”
Energy Policy 41:2–13.
Dartanto, T. 2013. “Reducing Fuel Subsidies and the Implication on Fiscal Balance and
Poverty in Indonesia: A Simulation Analysis.” Energy Policy 58:117−34.
Deaton, A. 1989. “Rice Prices and Income Distribution in Thailand.” Economic Journal
99:1–37.
Freund, C., and C. Wallich. 1997. “Public-Sector Price Reforms in Transition Economies:
Who Gains? Who Loses? The Case of Household Energy Prices in Poland.” Economic
Development and Cultural Change 46(1):35–59.
Grainger, Corbett Alden, Fan Zhang, Andrew William Schreiber. 2015. Distributional Impacts
of Energy Cross-Subsidization in Transition Economies: Evidence from Belarus. Policy
Research Working Paper, no. WPS 7385. Washington, D.C.: World Bank Group.
Hicks, J. 1942. “Consumers Surplus and Index-Numbers.” Review of Economic Studies
9:126–37.
Hills, John. 2012. Getting the Measure of Fuel Poverty: Final Report of the Fuel Poverty
Review. CASE Report 72. London: Department for Energy and Climate Change and
Centre for Analysis of Social Exclusion. https://assets.publishing.service.gov.uk/
government/uploads/system/uploads/attachment_data/file/48297/4662-getting-
measure-fuel-pov-final-hills-rpt.pdf
IEA, OPEC, OECD, and World Bank. 2010. “Analysis of the Scope of Energy Subsidies and
Suggestions for the G-20 Initiative.” Joint report prepared for submission to the G-20
Leaders’ Summit.
*Inchauste, Gabriela, and Jon Jellema. 2016. “Constructing Consumable Income: Including
the Direct Eects and Indirect Eects of Indirect Taxes and Subsidies.” Chapter 6 in
Nora Lustig (ed.), Commitment to Equity Handbook—A Guide to Estimating the Impact
of Fiscal Policy on Inequality and Poverty. Commitment to Equity (CEQ) Working Paper
Series. New Orleans, LA: Tulane University.
Ivanic, Maros, and Will Martin. 2008. Implications of Higher Global Food Prices for Poverty
in Low-Income Countries. Policy Research Working Paper, no. WPS4594. Washington,
DC: World Bank. http://documents.worldbank.org/curated/en/253001468150585777/
Implications-of-higher-global-food-prices-for-poverty-in-low-income-countries
Jacoby, H. 2016. “Food Prices, Wages, and Welfare in Rural India.” Economic Inquiry
54(1):159–76.
Kochnakyan, Artur, Ani Balabanyan, Pedro Antmann, Caterina Ruggeri Laderchi, Anne
Oliver, Lauren Pierce, and Denzel Hankinson. 2013. Armenia—Power Sector Tari Study.
Washington, D.C.: World Bank.
Kojima, Masami. 2016. Fossil Fuel Subsidy and Pricing Policies: Recent Developing Country
Experience. Policy Research Working Paper no. WPS 7531. Washington, DC: World Bank
Group.
47
GOOD PRACTICE NOTE 3: ANALYZING THE INCIDENCE OF CONSUMER PRICE SUBSIDIES AND
THE IMPACT OF REFORM ON HOUSEHOLDS — QUANTITATIVE ANALYSIS
Kojima, Masami, Robert Bacon, and Xin Zhou. 2011. Who Uses Bottled Gas? Evidence from
Households in Developing Countries. Policy Research Working Paper 5731. Washington,
D.C.: World Bank. http://go.worldbank.org/IF3RXBPQK0
Kojima, Masami, Xin Zhou, Jace Jeesun Han, Joeri de Wit, Robert Bacon, and Chris Trimble.
2016. Who Uses Electricity in Sub-Saharan Africa? Findings from Household Surveys.
Policy Research Working Paper, no. 7789. Washington, D.C.: World Bank. https://
openknowledge.worldbank.org/handle/10986/25029
Lampietti, Julian, and Nils Junge. 2006. “Europe and Central Asia: Power Sector Reform.”
In Aline Coudouel, Anis Dani, and Stefano Paternostro (ed.), Poverty and Social Impact
Analysis of Reforms: Lessons and Examples from Implementation. Washington, DC:
World Bank. http://elibrary.worldbank.org/doi/abs/10.1596/978-0-8213-6486-4
Lang Kerryn and Peter Wooders (2012) India’s Fuel Subsidies: Policy recommendations for
reform, Global Subsidies Initiative Policy Brief
*Olivier A. 2016. “Knowledge Note on How to Conduct Distributional Analysis with
Household Survey Data.” Knowledge Note for Design and Delivery of Energy Subsidy
Program. Washington, D.C.: World Bank.
Ruggeri Laderchi, Caterina, A. Olivier, and C. Trimble. 2013. “Balancing Act: Cutting Energy
Subsidies While Protecting Aordability.” Washington, D.C.: World Bank.
*Ruggeri Laderchi, Caterina. 2015. Transitional Policies to Assist the Poor While Phasing
Out Inecient Fossil Fuel Subsidies that Encourage Wasteful Consumption. World Bank
Technical Paper to the G20.
Soheir, Abouleinein, Heba El-Laithy, and Hanaa Kheir-El-Din. 2009. The Impact of Phasing
Out Subsidies of Petroleum Energy Products in Egypt. ECE Working Paper 145. Cairo:
Egyptian Center for Economic Studies.
Son H and N Kakwani. 2006. “Measuring the Impact of Price Changes on Poverty.” Journal
of Economic Inequality.
Swinkels, R. 2014. “Assessment of Household Energy Deprivation in Tajikistan: Policy
Options for Socially Responsible Reform in the Energy Sector.” Washington, D.C.; World
Bank Group.
Vagliasindi, Maria. 2013. Implementing Energy Subsidy Reforms: Evidence from Developing
Countries. Directions in Development; Energy and Mining. Washington, D.C.: World Bank.
Verme, Paolo, and Khalid El-Massnaoui. 2015. An Evaluation of the 2014 Subsidy Reforms
in Morocco and Simulation of Further Reforms. World Bank Policy Research Working
Paper, no. 7224.
World Bank. 2003. India: Access of the Poor to Clean Household Fuels. Energy Sector
Management Assistance Programme (ESMAP); no. ESM 263/03. Washington, D.C.:
World Bank. http://documents.worldbank.org/curated/en/612631468771668053/
India-access-of-the-poor-to-clean-household-fuels
World Bank. 2005. “Egypt—Toward a More Eective Social Policy: Subsidies and Social
Safety Nets.” Washington, D.C.: World Bank.
World Bank. 2013. Special Focus: Residential Gas and District Heating Taris in Ukraine.
Washington DC: World Bank Group.
48
REFERENCES
World Bank. 2015. “Beyond Connections: Energy Access Redefined.” Washington, D.C.
https://openknowledge.worldbank.org/handle/10986/24368
Zhang, Fan. 2011. Distributional Impact Analysis of the Energy Price Reform in Turkey. Policy
Research Working Paper, no. WPS 5831. Washington, D.C.: World Bank.
Zhang, Fan, and Denzel Hankinson. 2015. Belarus Heat Tari Reform and Social Impact
Mitigation. World Bank Studies. Washington, D.C.: World Bank.
Energy Subsidy Reform
Assessment Framework
LIST OF GOOD PRACTICE NOTES
NOTE 1 Identifying and Quantifying Energy Subsidies
NOTE 2 Assessing the Fiscal Cost of Subsidies and Fiscal Impact of Reform
NOTE 3 Analyzing the Incidence of Consumer Price Subsidies and the
Impact of Reform on Households — Quantitative Analysis
NOTE 4 Incidence of Price Subsidies on Households, and Distributional
Impact of Reform — Qualitative Methods
NOTE 5 Assessing the readiness of Social Safety Nets to Mitigate the
Impact of Reform
NOTE 6 Identifying the Impacts of Higher Energy Prices on Firms and
Industrial Competitiveness
NOTE 7 Modeling Macroeconomic Impacts and Global externalities
NOTE 8 Local Environmental Externalities due to Energy Price Subsidies:
A Focus on Air Pollution and Health
NOTE 9 Assessing the Political Economy of Energy Subsidies to Support
Policy Reform Operations
NOTE 10 Designing Communications Campaigns for Energy Subsidy Reform
