David Coady
PSIA Group
Fiscal Affairs Department
International Monetary Fund

Structure of Presentation
 Background to PSIA on fuel subsidies
 Objective of the PSIA studies
 Methodology, data, impacts (five steps)
 Mitigating measures plus pro-poor and progrowth expenditures
 Policy messages from PSIA

Background I: Market Structure
 Most developing countries control the domestic pricing and distribution of petroleum products
 Recent FAD survey found that from 48 countries
 15 had fully liberalized systems
 8 had functioning automatic pricing formulae (+8 suspended recently)
 21 had ad hoc pricing

Background II: Prices and Subsidies
(World prices have increased substantially since 2002)
Text Table A. Change in International Fuel Prices, 2003-06
1
Crude oil prices
Gasoline
Kerosene
Diesel
US$ per liter
0.4
0.6
0.6
0.6
Percent change
128.0
140.7
126.7
142.1
1/Increase during end-2003 to June 2006. The crude oil price is the average spot prices for Dated Brent,
WTI, and the Dubai Fateh. The prices for the other fuels are the average fob prices for Rotterdam, New York, Gulf
Coast, Los Angeles and Singapore.

Major Events and Real Price of U. S. Oil Imports, 1970 –2006
$80
Iran-Iraq War begins; oil prices peak $70
$60
$50
$40
Saudi Arabia abandons "swing producer" role; oil prices collapse
Prices spike on Iraq war, rapid demand increases, const rained
OPEC capacit y, et c.
Prices rise sharply on OPEC cut backs, increased demand
Gulf War
$30
$20
Iranian revolut ion;
Shah deposed
$10
1973 Oil Embargo
Iraq invades
Kuwait
Asian economic crisis; oil oversupply; prices fall sharply
Prices fall sharply on 9/11 at t acks
$-
1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006

Text Table C. Gasoline Pricing Mechansims, Prices
and Price Pass-Through
Pricing mechanism
Number of countries
Average price
(US$ per liter)
2003 2006
Price pass though
Ad hoc
Automatic
Liberalized
21
8
15
0.61
0.56
0.70
0.98
0.84
1.03
0.83
1.00
1.13

Background II: Prices and Subsidies



Controlled prices have resulted in rising budget subsidies in many countries (% 2005 GDP, estimated)
–
–
Yemen, 9.2; Jordan, 5.8; Indonesia, 4.2; Bolivia, 0.8
Subsidy rates typically higher for kerosene and diesel as well as in exporting countries
Countries often respond by decreasing taxation, socalled tax expenditures (especially kerosene and diesel)
– e.g. Bangladesh, India, Sri Lanka, Kenya, Zambia
Implicit subsidies also often substantial and take form of quasi-fiscal deficit financed by debt (%GDP2005, estimated)
– Azerbaijan, 13.9 (2.8ex); Egypt, 4.1; Ecuador, 3.6; Bolivia, 5.2

(a) Explicit subsidies
Argentina
Azerbaijan
Bolivia
Cameroon
Congo, Republic of
Dominican Republic
Ghana
Honduras
Indonesia
Jordan
Lebanon
Nigeria
Pakistan
Senegal
Sri Lanka
Yemen, Republic of
2003
0.2
...
1.5
0.0
...
0.0
0.1
...
...
5.0
0.0
5.1
0.6
0.0
0.8
...
Est.
2005
0.9
...
4.2
5.8
0.1
0.0
0.2
0.6
0.8
9.2
0.2
2.8
0.8
0.2
1.0
0.5
Proj.
2006
0.7
0.6
2.5
1.2
0.1
1.0
...
0.8
...
8.5
0.2
1.9
1.3
0.3
1.0
0.4

2003
Est.
2005
Proj.
2006
(b) Implicit subsidies
Armenia
Azerbaijan
Bangladesh
Bolivia
Cameroon
Colombia
Congo, Republic of
Dominican Republic
Ecuador
Egypt
Ethiopia
Gabon
Indonesia
Nigeria
Sri Lanka
3.9
...
0.4
...
1.6
...
0
10.0
...
1.7
0.1
1.2
...
...
1.4
4.1
0.7
1.6
...
2.2
1
0
13.9
1.0
5.2
0.0
1.6
...
0.2
3.6
1.0
10.4
...
6.6
…
0.3
...
6.2
2.8
0.3
...
...

Text Table B. The Average Price Pass Through, 2003-2006
1
Net oil importers
Net oil exporters
Gasoline
1.09
0.46
Kerosene
0.91
0.43
Diesel
1.15
0.70
AFR
APD
EUR
MCD
WHD
G-7 countries of which : USA
1.06
1.05
1.25
0.56
1.00
1.11
0.89
1.07
0.37
...
0.78
0.92
Average
2
Countries in sample
2
0.96
44
0.83
29
1/ Post-tax retail prices; latest observation for the fisrt half of 2006.
A number lower than one indicates less than full pass-through.
2/ excluding G7 countries
1.11
0.83
1.54
0.78
1.30
1.07
39

Pricing Regime (Selected Countries)
Country
Price mechanism
Argentina
Bolivia
Brazil
Colombia
Liberalized
Ad hoc
Liberalized
Automatic
Dominica Automatic
DominicanRep Liberalized
Ecuador
Honduras
Peru
Uruguay
Ad hoc
Ad hoc
Liberalized
Ad hoc
Retail fuel price (US$ per liter) Price pass-through
Gasoline Kerosene Diesel Gasoline Kerosene Diesel
Tax % Gas
Retail
Prices
(2006)
...
1.03
...
3.33
1.25
1.45
0.65
0.46
1.27
0.64
...
0.83
...
2.27
0.91
0.89
0.47
0.34
...
...
...
0.79
...
...
0.85
0.95
0.64
...
0.92
0.47
...
1.78
...
...
1.64
1.40
0.09
0.21
1.14
0.74
...
1.49
...
...
1.28
0.84
0.08
...
...
...
...
1.29
...
...
0.99
1.14
0.83
...
2.92
0.65
...
33.4
...
...
42.0
43.9
46.4
36.4
...
38.4

Background III: Reform Agenda


Fuel subsidies seen as undesirable because
–
–
High fiscal cost with consequences elsewhere in budget
(Indonesia/Yemen: subsidies exceeded combined health and education budgets)
Inefficient: leads to over-consumption
Governments still reluctant to increase domestic prices in line with world prices
–
–
Concerns about impact on poor and politically unpopular
PSIA can inform choice of appropriate policy response (so far:
Angola, Bangladesh, Bolivia, Ethiopia, Gabon, Ghana,
Honduras, Jordan, Madagascar, Mali, Moldova, Sri Lanka,
Sudan)

Objective of PSIA
 To identify the magnitude and financing of consumer subsidies
 To evaluate the aggregate and distributional incidence of their withdrawal on household real incomes
 To identify appropriate mitigation measures to offset adverse impact on poorest households
 To identify higher priority public expenditures
(more pro-poor and pro-growth)

Methodology and Data
 Higher domestic prices affect consumers through two channels
– Direct effect from increase in price of fuels consumed by households
– Indirect effect from increase in prices of goods and services that use fuel as inputs
 Indirect effect often substantial since over 50 percent of total consumption of fuel is as intermediate product

Step I: Identify magnitude and financing
 This requires a reference price for each product and required price increases
–
–
For most countries, border (cif,fob) price
(plus,minus) domestic trade and transport margins
Often existing or desired tax levels included in reference price to allow for “tax expenditures”
 Average price increase ranged from 34-68 percent (mostly including taxes)

P m
P
P s
P c
P p
Magnitude and Financing of Subsidies
A
D
Demand
Q c
B
E
C
Q s Q
• Domestic refinery that imports product
• Import at P(m), produce at P(c)
• Subsidized domestic price is P(s)
• Produces Q(c), imports Q(s)-Q(c)
• Total consumer subsidy =
(A+B+C)=Q(s)[P(m)-P(s)]
• Where shows up depends on price to producer. If taxes, P(p), P(s)
– Explicit import subsidy=(B+C)
– Loss in profits=(A+D)+E
– Tax revenue=(D+E)
– Net fiscal position
• On budget: (D+E)-(B+C)
• Off budget: -(A+D+E)

Cameroon: More Transparent Formula
P e tro l
3 0 0
2 0 0
10 0
0
8 0 0
7 0 0
6 0 0
5 0 0
4 0 0
3 3
1 2 0
2 1
6 4
3 0
1 0 7
3 0 0
2 0 0
7 7
9 8
3 0 0
1 2 0
7 7
9 8
3 0 0
1 5 7
7 7
9 8
3 0 0
5 9 5
E x i s i t i ng R e f o rme d E f f i c i e nc y E q ui t y A c t ua l
Im por t P r ic e Ma r gins Cust om s VAT Input VAT Exc ise AE Ac t ua l P r ic e

Sri Lanka: Eliminating subsidies required: gas (12%), diesel (20%), kerosene (58%), average (23%)
104
93
84
79
61
43
Formula Actual
Gasoline
Formula Actual
Kerosene
Landed Cost
Distribution Margins
Excise Taxes
Formula Actual
Diesel
Value Added Tax
Consumer Price

Step II: Calculate direct effect




Need household survey with information on different fuel expenditures
For each household, calculate budget shares as expenditure on fuel divided by total household consumption
Multiply required price increases by budget share to get approx. real income impact
Look at distribution of percentage real income effect across income groups (regressive vs. progressive)

Example of fuel consumption patterns in Sri Lanka
5.3
3.0
2.7
2.5
2.6
3.7
Bottom
Decile
Second
Decile
Second
Quintile
Kerosene
LPG
Third
Quintile
Fourth
Quintile
Diesel and Petrol
Electricity
Top
Quintile

Magnitude of direct effect



Fuel budget shares varied from 2-4.3 percent (3.1-
6.6 percent including electricity)
– Therefore, a 50 percent increase in average price implies a
1-2.1 percent (1.6-3.3 percent) decrease in real incomes
Fuel budget shares for lowest welfare quintile varied from 2-6 percent (2.7-7.1 percent)
– Therefore, a 50 percent increase in average price implies a
1-3 percent (1.4-3.6 percent) decrease in real incomes
Direct effect found to be either neutral of regressive
– Reflects importance of kerosene, which is typically relatively heavily subsidized

Step III: Calculate indirect effect





An input-output table and a simple model can be used to calculate the increase in prices for other goods and services from higher fuel costs
Aggregate household consumption data to get budget shares for input-output sectors
Multiply budget shares by percentage price increases to get percentage real income effect
Aggregate to get total indirect effect and look at distribution across different income groups
Add to direct effect to get total impact of fuel price increase on household real incomes and distribution

Example from Ghana
Sector
Agriculture
Utilities and mining
Manufacturing
Construction
Trade
Transport
Business
Community
Electricity
Budget Share
(BS)
0.452
0.021
0.253
0.000
0.070
0.032
0.025
0.097
0.008
Price Effect
(dP)
0.066
0.116
0.052
0.107
0.107
0.267
0.025
0.048
0.000
Impact=BS*dP
0.030
0.002
0.013
0.000
0.007
0.008
0.001
0.005
0.000

Magnitude of indirect effect
 Indirect effect at least as large as direct effect and approximately neutral incidence
 A 50 percent average increase associated with a 3 percent decrease in real incomes
 Most of indirect effect comes through higher food and transport costs

Magnitude of total effect
 Total effect ranged from 2-8.5 percent
 A 50 percent increase associated on average with a 4.6 percent decrease in real incomes
 Distribution typically regressive reflecting role of higher kerosene price increases

Step IV: Evaluate targeting efficiency
 Calculate the share of the total subsidy (or, equivalently, the burden of subsidy removal) accruing to each income group
 Can do this separately for each product as well as the direct, indirect and total effects
 Individual product shares useful later when comparing alternative approaches to protecting the real incomes of low-income households

Fuel subsidies are badly targete d


A relatively high share of total fuel subsidies go to higher income groups
–
–
Share of bottom two quintiles varied from 15-25 percent (so
75-85 percent of subsidy benefit accrues to top three quintiles)
So costs 4-6.7 units of income for every 1 unit transferred to bottom two quintiles
Even direct (mainly kerosene) subsidy is badly targeted
– Between 70-80 percent leaks to top three quintiles so costs
3.3-5 units of income for every unit transferred to bottom two quintiles

Step V: Identify mitigating measures


Although badly targeted, withdrawal of fuel subsidies can have substantial adverse effect on poor (c2-9%)
Can consider a number of alternatives and simulate using household-level data (budgetary cost minimized by better targeted transfers/expenditures)
–
–
–
–
Gradual withdrawal of specific fuel subsidies (kerosene, LPG) to minimize revenue-poverty trade-off
Using some of budgetary savings to finance targeted public expenditures (education, health, roads, transport, electricity)
Restructure electricity tariff schedules to reduces cost for poor
Use savings to finance existing/reformed/new social safety net for poorest households

Example from Ghana
Benefit Shares
Education
Untargeted
Targeted
Health
Untargeted
Targeted
Rural electrification
Urban transport
Proxy-means targeting
Kerosene subsidy
Bottom
0.215
0.204
0.149
0.148
0.329
0.299
0.373
0.178
2 nd
Quint
0.225
0.279
0.193
0.229
0.251
0.128
0.277
0.211
3 rd
Quint
0.219
0.249
0.208
0.208
0.212
0.185
0.205
0.227
4 th
Quint
0.187
0.170
0.207
0.226
0.135
0.280
0.111
0.209
Top
0.154
0.098
0.244
0.189
0.074
0.108
0.035
0.174

Example from Sri Lanka
 Kerosene subsidies
 Use of electricity lifeline rates
– Potential benefits from restructuring tariff schedule
 Use of existing Samurdhi transfer program
–
–
Highlight performance level of existing program
Emphasize gains from reforming design and implementation

Even kerosene subsidies involves substantial leakage to the non-poor
18.9%
15.2%
13.0%
10.7%
22.0%
20.2%
Bottom Decile
Second Quintile
Fourth Quintile
Second Decile
Third Quintile
Top Quintile

Share of Gasoline Burden (Cameroon)
Share of Burden from Direct Effect--Gasoline
10.3%
83.0%
Bottom Decile
Second Quintile
Fourth Quintile
Second Decile
Third Quintile
Top Quintile

Share of LPG Burden (Cameroon)
Share of Burden from Direct Effect--LPG
0.7%
6.0%
11.0%
58.1%
22.4%
Bottom Decile
Second Quintile
Fourth Quintile
Second Decile
Third Quintile
Top Quintile

Alternatively could subsidize electricity.......
0 50 100 150 200
Monthly Electricity Consumption (Kw/H)
Existing Tariffs
Restructured Tariffs
Scaled Tariffs
Cumulative Density
250

......but these appear badly structured.....
20th percentile
40th percentile
6 8
Log Per Capita Consumption
Existing Tariffs
Restructured Tariffs
10
Scaled Tariffs

.....and involve very substantial leakage to non-poor
49.6%
Scaled Tariffs
2.0%
3.9%
9.5%
12.5%
22.5%
Bottom Decile
Second Quintile
Fourth Quintile
Restructured tariffs
4.8%
7.9%
26.8%
17.7%
21.9%
21.0%
Second Decile
Third Quintile
Top Quintile

The Samurdhi program reduces leakage substantially.........
Samurdhi Food Stamps
4.6%
17.4%
14.7%
14.8%
22.4%
26.1%
Bottom Decile
Second Quintile
Fourth Quintile
Reformed Samurdhi
1.0%
8.0%
24.9%
18.1%
29.1%
Second Decile
Third Quintile
Top Quintile
18.9%

....and potentially provides a more costeffective approach to social protection
16.9
7.9
7.3
Scaled Electricity
Fuel
Existing Samurdhi
4.2
3.1
2.3
Restructured Electricity
Kerosene
Restructured Samurdhi

Policy messages from PSIA


Fuel subsidies are often substantial fiscal drain, crowdout priority expenditures and badly targeted
So should be able to identify alternative uses that are more pro-poor and pro-growth:
–
–
–
Alternative approaches to social protection can provide same or better protection at substantially lower fiscal cost
Higher priority public expenditures (nutrition, health, education, infrastructure) – e.g. based on PRSP
Access to effective system for targeting expenditures can be a crucial component for promoting efficiency-enhancing structural reforms

Policy messages from PSIA
 Important to announce reforms as part of a package: budgetary savings to finance better targeted, higher priority expenditures that benefit low- and middle-income households
 Gradual reduction of better targeted fuel subsidies should be seen only as short term measure are developed since revenue-poverty trade off is large and efficiency cost from interfuel substitution large