Journal Issue

Mali: Selected Issues

International Monetary Fund
Published Date:
March 2006
  • ShareShare
Show Summary Details

I. How Fuel Prices Affect Household Spending: Evidence from Mali1

Using an input-output model and a household dataset for Mali, this paper assesses the distributional effects of a rise in petroleum product pump prices. The results show that rising gasoline and diesel prices affect mainly non-poor households while rising kerosene prices are most harmful to the poor. Overall, the impact of fuel prices on household budgets displays a U-shaped relationship with expenditure per capita. However, with the exception of kerosene, high-spending households would benefit disproportionately from not passing through increases of international prices to pump prices, i.e. implicit oil price subsidies. To mitigate the impact of rising fuel prices on poor households in an efficient manner transport subsidies, support for rural electrification, or kerosene coupons could be considered. A non-discretionary formula for passing through prices could make such increases less politically controversial.

1. Mali is highly dependent on imported oil products. Mali’s fuel import dependence, measured by the ratio of net oil imports to GDP, reached 5.4 percent in 2001, compared with an average of 3.3 percent for all oil importers and 4.0 percent for landlocked oil importers. Mali’s vulnerability results from a lack of domestic oil resources, and long and costly routes of importation (the nearest ports are nearly 1,000 kilometers from the capital).

2. The dramatic rise of international oil prices during 2004–05 poses a policy dilemma for policymakers in oil importing countries (Figure 1). A full pass-through of increases in international oil prices to administratively set pump prices lowers household purchasing power and may increase poverty. However, by not passing on higher prices, lowering fuel taxes, a revenue loss would result jeopardizing spending objectives. The size of the revenue loss is a function of the extent of pass-through and whether international price rises are permanent or temporary.

Figure 1:Mali: Pass through of world oil prices, 2002-05

(Index 2002=100)

A. International Oil Prices and Pass Through to Domestic Prices

3. Market fundamentals, together with expectations of continued narrow margins between global supply and demand, have been the primary influence on international crude oil prices during the past two years (IMF, 2005). The crude oil price increases of 2004 can be broadly explained by the unexpectedly rapid growth in consumption, which grew at the fastest rate in 20 years. Oil producers were unable to significantly increase production relative to original plans. Price increases in 2005, while still supported by market fundamentals, appear largely to reflect the uncertain environment and expectations of future tightness in the market. Strong demand pressures continue to put pressure on production capacity, in contrast to earlier periods when OPEC production increases were able to offset surges in demand. As a result, there has been a significant rise in longer-dated futures prices reflecting the perception of continued tightness in the oil market. CFA franc appreciation against the dollar has only moderately dampened the effect of the rising dollar-denominated price of international crude oil.

Indicative ceiling prices for fuel products in Mali are set administratively on a monthly basis. Through end-2002, changes in pump prices followed a more or less automatic formula, with an element of smoothing (in line with a 2001 UEMOA directive on fuel taxation). 2 From 2003 onwards, price changes have been more discretionary leading to larger divergences between domestic and international prices. The policy of a discretionary and gradual pass-through of increases in international oil prices to pump prices poses significant fiscal risks given that the recent international price increase is likely to be sustained.

Table 1.Mali: Structure of Fuel Pump Prices, [June] 2005

(In percent of pump prices) 1/

Supplier price60.134.349.3
of which: TIPP3.930.912.1
Transport costs13.29.111.7
Note:Source: Ministry of Economy and Finance.

B. The Impact of Rising Fuel Pump Prices on Households

Methodology and Data3

4. Rising pump prices have both a direct and an indirect effect on household spending. Pump price increases directly affect household spending through the higher prices for the fuel products they consume, and indirectly through price changes of other goods that use fuel as a production input. The direct effect on household expenditure of a rise in pump prices is calculated by multiplying the household budget share of each fuel product by its change in price. The indirect effect is calculated by multiplying the household budget share of each non-fuel product by its change in price derived from an input-output model. The input-output model requires less data than a computable general equilibrium (CGE) model. It is also.

5. While the input-output model is easy to implement and can indicate the magnitude and distributional effects of the impact of marginal price changes on real household spending it does rely on a fairly restrictive set of assumptions. Specifically, the input-output model is based on the following assumptions: homogeneity of output, no substitution between inputs, fixed proportions between inputs and outputs, absence of economies of scale, exogeneity of primary inputs and final demand components. Some of these restrictions are reasonable: spending patterns would likely not change significantly in the short term; given that there are few close substitutes for fuel products, the level of technology is also unlikely to change significantly in the short run. In the medium term, the first-order effects will nonetheless tend to overestimate the income loss that results from price increases. Finally, unlike a CGE model, the input-output approach does not take into account the labor market effects of producer price changes.

6. The expenditure data used in this paper are from the Mali 2000-01 household survey. The average expenditure per capita in urban areas is more than double that in rural areas, partially because of the smaller size of urban households. Expenditure on food is the primary spending item in both urban and rural areas. Oil product spending in urban areas is equivalent to 3.8 percent of spending (primarily for transport purposes), compared to 2.3 percent in rural areas (primarily for cooking/light).

7. Household expenditure levels and equipment ownership help to explain oil consumption patterns. The budget share (per capita) of gasoline and diesel rises with overall spending while the budget share for kerosene falls (Table 2). Holding constant the level of expenditure per capita, agricultural households that own a tractor or a cultivator tend to consume more oil products (Table 3). Similarly, households that possess a car or a moped tend to have higher petroleum consumption. A household’s access to electricity is also negatively associated with kerosene consumption.

Table 2.Household Budget Shares of Energy Spending by Product(percent of total spending)
Note: Shares are calculated using data from the 2000-01 household survey. Quintiles are based on the national distribution of consumption per adult equivalent.

8. The 1998 input-output table is used to define the technology of production.4 The petroleum sector is strongly linked to other sectors in the economy and has the highest sum of domestic input coefficients. The most oil-intensive sector consists of services intended for the agricultural sector, with a per unit requirement of 0.29, followed by the electricity and transport sectors, which have, respectively, 0.27 and 0.19 per unit requirements. 5 A sector’s dependence on oil products determines the increase in the cost of production following a rise in oil prices.

Table 3.Equipment Ownership and Oil Consumption Sign of coefficients from regressing fuel spending shares on the level of expenditure per capita (log) and a dummy variable for each piece of equipment.
Expenditure per capita+ ve-ve+ ve+ ve
Tractor+ ve+ ve+ ve
Cultivator+ ve+ ve+ ve
Mill+ ve+ ve+ ve
Generator+ ve- ve+ ve
Car+ ve- ve+ ve+ ve
Moped+ ve- ve+ ve+ ve
Pirogue (canoe)+- ve
Access to electricity+ ve- ve+ ve+ ve
Power-driven pump+ ve- ve- ve+ ve

9. Firms consume a major share of oil products. All the resources of the petroleum sector are imported. Approximately 76 percent of fuel products are used by firms as intermediate inputs, 20 percent consumed by household (balance is re-exported). Across sectors, transport and metallurgy consume the largest shares of oil products.

Simulation Results

10. Simulations are presented of a full-pass through of a 34 percent rise in domestic fuel prices (equivalent, for example, to the actual increase between September 2003 and September 2005). For convenience, we assume that a rise in expenditure resulting from higher product prices is equivalent to a decrease in real spending.

11. The direct expenditure effect as a share of spending of the rise in oil prices is modest, and its distribution is nonlinear. By expenditure quintile, the highest spending households are most affected, and the lowest spending the next most affected (Table 4). A 34 percent rise in the prices of all oil products leads to a 1.0 percent decrease in real spending for the top quintile, while the spending of households in the bottom quintile drops by 0.9 percent. The kerosene prices has a slight regressive effect on income, whereas the gasoline price is almost progressive. The impact of higher diesel prices is almost negligible for the three bottom quintiles. Nonparametric estimates support the above findings (not reported in this paper see: Kpodar, 2005).

Table 4.Direct Expenditure Effects of 34 Percent Fuel Price Increase
QuintileExpenditure Effect (percent of spending)Annual Expenditure Per Capita (CFAF ‘000)Annual Nominal Expenditure Effect (CFAF ‘000)Share of Expenditure Effect

(In percent)
All Oil ProductsKeroseneGasolineDiesel
Source: 2000-01 Household survey.Note: 1/Expenditure effects are obtained by multiplying the sum of oil product spending shares by the oil price increase. Quintiles are based on the national distribution of consumption per adult equivalent. Average expenditure per capita is based on annual per adult equivalent consumption.

12. The distribution of the direct expenditure effect indicates that higher expenditure groups would benefit most from not passing through increases of international prices to pump prices. Assuming the cost of not passing through is carried by the budget in the form of lower taxes, overall 40 percent of such tax expenditures would accrue to the top expenditure quintile, nearly four times the amount accruing to the lowest quintile. For gasoline and diesel the tax expenditures captured by the top quintile are 59 and 71 percent respectively. Urban households are, on average, more affected directly by fuel price increases than rural households in absolute spending terms particularly in the top quintile.

13. The indirect effects of fuel price changes depend on both the pattern of consumption on non-fuel products and the sensitivity of non-fuel product prices to fuel prices. Low expenditure households are likely to be disproportionately affected indirectly if food prices are highly sensitive to oil prices as their expenditure is skewed towards food. Using the input-output matrix to find the sensitivity of non-fuel products to fuel price changes shows that services to the agricultural sector are the most sensitive, followed by metallurgy products, transport and telecommunication. These simulations assume that cost increases caused by higher oil prices are fully passed through to output prices, except the prices for electricity and public services, that are assumed to be controlled by the government (Figure 3).

Figure 3.Simulated non-fuel product price changes, with 34 percent increase in fuel prices

Source: Based on 1998 input-output matrix.

14. The bulk of the indirect expenditure effect on households comes through increases in food expenditures, although food price increases are relatively small (Table 5). Expenditure effects for foods are the largest mainly because of their high budget shares. Likewise, the expenditure effects for transport and telecommunication are also large, albeit because of high price increases rather than high budget shares.

Table 5.Indirect Price and Expenditure Effects by Sector 1/(In percent)
SectorPrice changeBudget shareExpenditure effect
Flour and processed cereals1.813.40.24
Fatty substances2.13.80.08
Food products2.43.10.08
Transport and telecommunication services7.20.90.07
Textile and clothing1.24.50.05
Chemical products1.82.20.04
Cereal-based products3.11.10.04

15. The indirect expenditure effects of pump price increases are modest, although overall the bottom quintile and rural residents experience the biggest expenditure share increase (Table 6). This is the case because of the relatively high food budget shares of poor and rural households and the assumption that the electricity price is fixed (benefiting urban residents). Similar to the direct effects, the main beneficiaries of not passing through world prices to pump prices (by cutting taxes) would accrue to the top quintile, where the nominal effect of price increases is three times larger than that accruing to the bottom quintile.

Table 6.Indirect Expenditure Effects by Quintile, Rural and Urban(34 percent increase of pump prices)
Impact on total spending (percent)
QuintileAllUrbanRuralShare of


Note: Quintiles are based on the national distribution of consumption per adult equivalent. Data derived by aggregating the expenditure effects in Table 8 for each household.

16. Total expenditure effects yield a conclusion similar to the direct effects, which account for half of the total effect (Table 7). First, the simulated pump price increases have a limited impact on household expenditures, with urban households being more affected than rural households. Upper quintile households are most affected (in nominal spending terms) because they consume a larger share of oil products, and thus would benefit most from not passing through price increases.

Table 7.Total Expenditure Effects by Quintile(For 34 percent pump price increase)
Percent of total spendingShare of total expenditure effect
QuintileAllUrbanRural(in percent)

17. Upper quintile households also benefit from controlled electricity prices (Table 8). When we remove the assumption of a fixed electricity price, a 34 percent rise in oil prices leads to a 10.7 percent increase in the electricity price. The distributional effects show that the removal of electricity subsidies is slightly progressive. However, the urban poor are likely to be affected negatively by a higher electricity price and might therefore need to be protected.

Table 8.Indirect Expenditure Effects by Quintile, Rural and Urban: Distribution Effects of Controlled Electricity Price
Noncontrolled Electricity Price (1)Controlled Electricity Price (2)Difference (1)-(2)
Quintile 1/AllUrbanRuralAllUrbanRuralAllUrbanRural

C. Cross-Country Comparisons

18. The estimated average expenditure effect of oil price increases in Mali falls in the range of findings identified in other country studies.6 A 20 percent increase in oil prices leads to 1 percent rise in household expenditures in Mali,7 which is similar to that for Pakistan (0.85 percent) but significantly lower than that for Ghana (3.4 percent). The difference between expenditure effects in Mali and Ghana likely results from higher oil subsidies that lead to much higher oil product consumption in Ghana, e.g., the kerosene budget share is 3.5 percent in Ghana compared with 1.5 percent in Mali (2001).

19. The distributional effects across income groups vary across country studies. Fuel price rises have a progressive effect in South Africa and Indonesia, and regressive effect in Ghana, the Islamic Republic of Iran, and Pakistan. Mali and Mozambique show a different pattern, with the top quintile most affected and the bottom quintile second most affected. In part these differences can be explained by differing assumptions on the price increases by product. In Mali, the U-shaped impact results from a high level of kerosene consumption in low spending quintiles. However, the Pakistan case study is puzzling because the oil price increase is regressive even though the kerosene price is held constant. Further data investigation reveals that the poor are affected by increases in food prices (milled grains, vegetable oils, wheat, and so on) which are sensitive to transport costs (and hence diesel prices).

20. The relative importance of the direct effect varies across countries. In Mali, the direct effect of the oil price increase accounts for 50 percent of the total effect, compared with 20 percent in Ghana. This disparity is explained by the fact that households in Ghana directly consume 6.2 percent of petroleum products, less than a third of the level of household consumption in Mali. The indirect effects are higher in Ghana because households consume more energy-intensive products. In particular, transport costs represent 3.2 percent of household expenditures in Ghana and only 0.9 percent in Mali.

D. Policy Implications

21. oil price increases have a clearly adverse impact on household expenditures in Mali. A simulation of large increase (34 percent) in oil prices leads to a 1.67 percent average increase in household expenditures, with the impact on rural households (1.64 percent) being slightly smaller than that on urban households (1.76 percent). The indirect effect caused by the rise in the prices of other goods and services is calibrated through input-output linkages with the petroleum sector; it represents roughly 50 percent of the total effect.

22. Although the lowest and the highest expenditure quintiles are most adversely affected by oil price rises, the benefits of not passing through price increases accrue largely to the nonpoor. Mali therefore could realize significant gains by trying to target subsidies to achieve its poverty reduction objectives. Likewise, control of the electricity price to smooth the impact of oil price increases benefits high spending households rather than poor households.

23. However, these results represent the maximum short-run impact of the oil price increase. In the medium term, households and firms will adjust their demand for oil products, leading to smaller expenditure effects as consumers and producers reduce their demand for oil, either by switching to a different fuel or by switching to other goods or products.

24. In order to mitigate the economic and social impact of fuel price increases, a number of palliative measures can be considered:

(i) Limiting the impact of kerosene price increases is of more benefit to low spending quintiles than other products although still relatively inefficient (benefits are spread equally across quintiles);

(ii) kerosene coupons could more effectively target low spending consumers, though such subsidies would likely be temporary and transparent to discourage the diversion of subsidized kerosene to other purposes;

(iii) temporary support for public transport might provide a cushion for urban low income consumers in the event of sharp increases of fuel prices;

(iv) subsidies for rural electrification (as proposed by the World Bank in Mali) could also help shield the poor from kerosene price increases by substituting electricity for kerosene;

(v) below-cost electricity disproportionately benefits high spending urban households. To protect the urban poor, the existing subsidized block tariff rate should be maintained at a modest level of consumption. This would lessen poor households’ incentive to switch to kerosene.

25. Finally, the introduction of a formula to determine pump prices can lessen the onus of blame on the government for price increases as pass-through of international fuel prices to domestic prices is often politically difficult. A clear explanation to the public of the rationale for such a mechanism is important in this respect. The formula can employ price smoothing to lessen market disruption. In the medium run, it would be desirable to allow market mechanisms to determine domestic oil prices and to promote the efficient use of fuel.

Appendix I. Methodology for Calculating Impact on Household Expenditure of Fuel Price Increase

The Direct expenditure effect

The budget shares of oil products determine the direct effect on consumers of the increase in oil prices.

where ∂ log Ydir is the direct expenditure effect (expressed in percentage), bt is the budget shares of the oil product t, Pt is the change in the price of oil product t, and m is the number of oil products consumed by households.

The Indirect expenditure effect and the input-output approach

The formula of the indirect effect is similar to that of the direct effect:

where ∂ log Yind is the indirect expenditure effect (expressed in percentage), bi is the budget shares of final goods other than oil products, Pi is the change in the price of good i, n and m are respectively the number of final goods and the number of oil products.

To compute the change in commodity prices following a change in oil prices, we follow the input-output approach adopted by Coady and Newhouse (2005). We consider two categories of sectors: the noncontrolled sectors and the controlled sectors. For noncontrolled sectors, the relationship between user and producer prices is given by

where Pncu is the price paid by consumers, Pncp is the price received by producers,8 and t is the tax imposed by the government.9 Changes in the user prices are then given by

For controlled sectors, producer prices are determined by the government. Therefore, the change in consumer prices is equal to the change in producer prices, assuming that the tax remains constant.

where ΔPcu is the change in consumer prices, and ΔPcp is the change in producer prices.

The technology of production is captured by an input-output coefficient matrix A, withajjdenoting the cost of input i in producing one unit of output j. Also, ajj represents the change in the production cost of a unit of j resulting from to a unit change in the price of input i. This implies a Leontief production function, where the firm’s demand for inputs is relatively insensitive to the changes in input prices.

Using the input-output coefficient matrix and assuming that factor prices are fixed, the change in producer prices is derived as

Substituting in from (2) for ΔPncu and rearranging (4), we obtain

where I is an n × n identity matrix. Assuming that the tax on controlled-sector commodities is constant (Δt = 0), the following equation gives the effect of a change in controlled prices on the prices of noncontrolled sectors:

Appendix II. Distributional Effects of Oil Price Changes: Selected Country Studies
Study/AuthorCountryContextDataMain results for a 20 percent increase in average oil prices (unless otherwise specified)
I. Input-Output Approach
Coady and Newhouse (2005)GhanaThe application of a new pricing formula requires, on average, a 50 percent increase in pump prices.The 1999 Living Standard Survey and the 1993 Social Account Matrix (SAM).The average real income decreases by 3.4 percent, with the poor being the most affected (3.64 percent). However, the removal of subsidies is a progressive policy. The indirect effect accounts for 80 percent of the total effect.
Valadkhani and Mitchell (2002)AustraliaImpact of a rise in oil prices on price level and income distributionThe 1998-99 Household Survey and the 1996-97 Input-Output tableBased on budget share analysis, the authors conclude that a rise in oil prices is regressive. However, they did not provide a clear estimate of real income effects.
World Bank (2003)Iran, Islamic Rep. ofA 308 percent rise in the average energy price, intended to bring all energy prices to import parityThe 1994-95 Input-Output table and the corresponding household survey.Households experience, on average, a 1.98 percent decrease in real income. The effect is regressive because poorer households are hit harder than better-off households, especially in rural areas (3.1 percent for poor households compared with 1.92 percent for rich households).
ESMAP10Report (2001)PakistanAssessment of a 33 percent rise in gasoline and diesel prices; other oil products remained unchanged.The 1989-90 Input-Output table and the 1996-97 Household Survey.The cost of living of households increases by 0.85 percent on average. The impact is higher for urban households (0.90 percent) than for rural (0.79 percent). In both areas, the impact is regressive, with the poor experiencing a 1.15 percent increase in expenditures.
Nicholson and others. (2003)MozambiqueIncrease in oil tax to improve road maintenance, raise domestic revenue, and reduce aid dependency.Data come from the 1993-94 Social Account Matrix and the 1996-97 Household Survey.The increase in the average fuel price, assuming kerosene remains tax-exempt, leads to a 0.42 percent increase in the average expenditure of households. The lowest quintile feels the impact slightly more than all other quintiles except the rich. The impact is significantly higher on urban households than on rural households.
II. CGE Analysis
McDonald and van Schoor (2005)South AfricaSimulation of various oil price shocks.The model is calibrated on data from the 2000 SAM.The rise in oil prices is progressive. Poor households tend to be less adversely affected and rural households have a slightly smaller drop in income than urban households (0.76 percent versus 0.83 percent).
Clements, Hong-Sang, and Gupta. (2003)IndonesiaAssessment of real and distributives effects of petroleum price liberalization in Indonesia.The model is calibrated on data from the 1995 SAM.A 25 percent increase in oil prices would lead to a 2.5 percent decrease in average real consumption. The impact is slightly progressive because high-income groups are the most affected, especially in urban areas.

This paper was finalized by Chris Lane based on a working paper undertaken by Kangni Kpodar during a summer internship at the IMF (Kpodar, 2005; forthcoming). The assistance of the IMF Poverty and Social Impact Analysis (PSIA) unit in undertaking the quantitative analysis is also gratefully acknowledged.

Ceiling pump prices expressed in CFAF/liter are calculated as: Pump price = {(c.i.f. import price)(1+import duty rate)+excise tax}{1+VAT} + transport cost + margin.

Appendix I gives a full explanation of the methodology adopted.

By applying 1998 input-output data to the 2000-01 household expenditure survey it is assumed that the economy did not experience a major technological shock in the intervening period.

For example, cotton ginning, protection and treatment of crops, and tool and agricultural machine rental.

The impact estimated using the input-output approach is linearly proportional to the level of the price increase. In Mali, a 34 percent rise in oil prices leads to a 1.67 percent decrease in real income, so real income will decrease by 0.98 percent if oil prices rise by 20 percent.

Producer prices are a function of intermediate goods costs and factor prices.

Note that t is a trade tax when we consider traded goods, but a domestic tax when we consider nontraded goods. A tax on domestic production of traded goods does not affect user prices, but reduces producer prices.

The Energy Sector Management Assistance Programme.

Other Resources Citing This Publication