Even before the “Arab Awakening” that sparked the uprising in Yemen and initiated the political transition, the country faced huge socio-economic challenges. In 2006, 35 percent of all Yemenis lived below the national poverty line [1
], and estimates suggest that poverty and food insecurity increased substantially as a result of the global food crisis in 2008. Breisinger et al.
] estimate that poverty increased to 43 percent in 2009, and Ecker et al.
] and WFP [4
] consistently estimate that 32 percent of the population was food insecure in 2009. Growth did not trickle down well to the poor, mainly because the economy is dominated by the capital-intensive hydrocarbon sector (that is, oil and gas exploration) and non-tradable services. Labor-intensive sectors such as agriculture and manufacturing make up a relatively small share of the economy. Even though about 70 percent of the population lives in rural areas, only about 30 percent of the population earns their main livelihood from farming with little alternative opportunities for non-farm employment [3
While poverty and food insecurity have likely further increased during 2011, many of the structural and economic weaknesses and challenges remain. One of the major challenges has been the combination of declining oil revenues and rising fiscal costs to sustain Yemen’s subsidy scheme in combination with a large budget deficit in 2009—estimated at about 10 percent of the GDP [6
]. In fact, Yemen is among the high subsidizers in the Middle East and Northern Africa (MENA) region, and there are only a few other countries in the world with lower fuel prices than Yemen, such as Libya, Saudi Arabia, Bahrain, Qatar, and Kuwait, which all have significantly higher GDP and larger oil or gas reserves per capita [7
]. In Yemen, as in other countries such as Azerbaijan, Bolivia, Ecuador, Egypt, Indonesia and Jordan, fuel subsidies account for more than three percent of GDP and are comparable in size to public spending on health and education combined [8
]. Given that it is usually the better-off households that disproportionally benefit most from fuel subsidies, these subsidies are not only a blunt tool against poverty, but also increase inequality [9
]. Additional detriments include an often inefficient fuel-processing sector and, given the premiums involved in the shadow market, smuggling and fuel adulteration.
An increasing number of governments therefore question the usefulness of their energy subsidy schemes, and several governments have launched substantial reforms lately, including Chile, Ghana, India, Iran and Syria. However, there is often uncertainty about the economic and social impact of potential reforms [12
]. Many studies find that fuel subsidy reforms raise overall economic growth, mostly explained by economic efficiency gains over time [14
]. Hope and Singh [17
] studied the impacts of fuel subsidy reform in six countries (Columbia, Ghana, Indonesia, Malaysia, Turkey and Zimbabwe) and showed that in three of the countries studied (Columbia, Indonesia and Ghana) in the years during fuel subsidy reform, the GDP grew faster than before, and growth in the other three countries quickly accelerated in the years after the implementation of the reform [14
]. Household welfare effects of fuel subsidy reform have been less widely studied; Coady et al.
] found in their six-country study that real incomes of the poorest household groups declined between 1.8 percent in Mali and up to 9.1 percent in Ghana. Consistently, Hope and Singh [17
] found decreases in real household incomes of 1–3 percent due to subsidy reform in all of the six countries studied. These findings are further confirmed by the experience of the 2007–08 global food and fuel crisis, where rising prices for fuel products and food led to an increase in poverty [19
]. These studies further show that the magnitude of poverty effects significantly differs between countries.
Past experience with subsidy reform suggests that protecting the poor from the negative impact of reform is most important for success. The immediate loss in real household incomes, especially among the poor, may explain why fuel subsidy reform is often accompanied by social tensions or even riots. Nonetheless, social unrest may be mitigated, if targeted compensation is provided and the reform process is accompanied by broad publicity campaigns that raise awareness of the social inequality created by subsidies and their inefficiency in fostering sustainable growth [20
]. Several countries have successfully applied direct income transfers to protect the most vulnerable households from the negative impact. For example, Chile provided several rounds of cash transfers to the poorest 1.4 million households, China compensated the poor with monthly payments to offset rising fuel costs and Indonesia issued quarterly payments of US$30 over one year for 15.5 million poor households (or 28 percent of the population). Ghana used a more indirect approach and abolished fees for all public primary and secondary schools and established a program to improve public transportation [20
In this paper, we analyze the direct and indirect effects of reforming Yemen’s fuel subsidy policy on growth and poverty. In addition to and based on lessons from other countries, we hypothesize that fuel subsidy reform can have significant growth effects but may slow poverty reduction substantially, if no additional measures are taken. To test this hypothesis, we use a dynamic computable general equilibrium (DCGE) model combined with a microsimulation model to estimate growth and poverty effects under two alternative reform options relative to a baseline scenario. The first scenario represents an accelerated reform path where all subsidies are lifted within one year, and the second scenario depicts a more gradual reform scenario that phases out subsidies over a period of three years. Under both scenarios, we simulate three alternatives for spending the savings from reform. In the first set of simulations, the total amount is used for budget deficit reduction. In the second set, parts of the savings are used for direct income transfers to the poorest 30 percent of households and, in the third set, for public investment in infrastructure in addition.
The rest of the paper is organized as follows: Section 2
describes the role of fuel subsidies in the Yemeni economy and argues that, independent of the outcome of the current uprisings, reforming the subsidy scheme remains a political challenge but is necessary for sustainable economic development. Section 3
presents the DCGE model and the reform scenarios of the simulation. Section 4
discusses the simulation results, and Section 5
concludes with policy implications.
3. Modeling Growth and Poverty Effects of Fuel Subsidy Reform
3.1. Yemen Dynamic Computable General Equilibrium Model
As outlined in the previous section, fuel subsidies account for a large share of Yemen’s government expenditure and play an important direct and indirect role for household incomes. Reducing the subsidies is likely to significantly reduce household welfare and alter the production costs of economic activities in the short run while freeing up substantial resources, which may be used for alternative spending. Estimating both the direct and indirect effects of fuel subsidy reform requires an economy-wide model that captures the linkages between the reduction of the subsidies, production, consumption, and household incomes. Given that many of the effects arise from changes in relative prices, social accounting matrix (SAM)-based computable general equilibrium models (CGE) models are more suitable than SAM-based multiplier models that have previously been used in comparable studies.
In this paper, we apply a recursive-dynamic computable general equilibrium (DCGE) model as described in Diao and Thurlow [29
]. Recent applications of this model include, for example, Breisinger, Diao, and Thurlow [30
] and Wiebelt et al.
]. In the following model description we focus on the specific model features for Yemen, including a description of the social accounting matrix (SAM) as the main dataset, the choice of elasticities, macro- and labor-market closure rules and the description of exogenous shocks for the simulations.
We first update a 2007 SAM [32
] to represent Yemen’s economy in 2009 as the main database for the model. The main data sources for constructing the 2007 SAM include the latest supply-use table from the Central Statistics Organization of Yemen, balance of payments from the Bank of Yemen, government budget data from the Ministry of Finance, the 2008 Agricultural Yearbook from the Ministry of Agriculture and Irrigation, and the latest Household Budget Survey conducted in 2005–06. These data were complemented with data from the IMF and the World Bank. For updating the 2007 SAM to the economic conditions in 2009, we used national accounts data for 2009 provided by the Ministry of Planning and International Cooperation.
The Yemen DCGE model is very detailed at the production, commodity, factor, and household levels and includes 65 production activities, 65 commodities, 15 factors of production, and six household groups [33
]. Factors of production include labor for three skill levels (unskilled, semiskilled and skilled) and employment in the public and private sectors. In addition to the SAM as the main data source to calibrate to a set of parameters in both production and demand functions, a DCGE model also requires several elasticities. The main elasticities include the substitution elasticity between primary inputs in the value-added production function, which determine the ease with which, for example, users of fuel can substitute it for other inputs; the elasticity between domestically produced and consumed goods and exported or imported goods; and the income elasticity in the demand functions. The income elasticity with regard to fuel, for example, measures how consumers react to higher prices. We estimated the elasticity of expenditure for commodities and services with respect to household income from a semi-log inverse function suggested by King and Byerlee [34
] for rural and urban households separately, using data from HBS 2005–06. These elasticities range from, for example, 0.31 for cereal consumption to 2.2 for transport and 1.95 for fuel consumption, while most elasticities are lower for urban households than for rural [35
]. Instead of elasticities that could not be estimated econometrically due to lack of data, we used international standard estimates provided by the International Food Policy Research Institute. For the substitution between intermediate inputs and value added in the production function, we assume constant elasticities of transformation that are 1.2 for the factor substitution elasticity, 4.0 for the elasticity of transformation, and 6.0 for the Armington elasticities of all goods and services [36
In our simulations the dynamics of the DCGE model occur between 2010 and 2015 in each year. In the baseline scenario as well as in all subsidy reform scenarios, we assume that the nominal exchange rate is flexible. Exogenous variables in the model include government consumption, transfers to households, foreign inflows, population growth and, hence, growth of the workforce, which all grow exogenously according to their trends in recent years. Investments are savings driven, which means that an increase of either private or public savings increases the economy-wide investment rate. The government budget is flexible in the model, which means that the government can adjust to changes in revenues and spending by increasing or decreasing the budget deficit (or its savings). For example, if fuel subsidies are reduced, the government savings increase. This leads to an overall increase of savings in the economy, and thus to higher investment. It is important to note that real sector CGE models in general cannot capture the long-term benefits of low public debt or GDP levels and related lower interest rates for borrowing capital.
At the sector level, total factor productivity (TFP) increases exogenously to account for the differential growth patterns across sectors. Non-hydrocarbon capital is fully mobile across all sectors, and its inter-temporal allocation follows the highest profitability by sector and period. Capital employed in the hydrocarbon sector is sector-specific and cannot move to other sectors. Population growth, land and productivity growth are all exogenously determined. Baseline growth in the model is driven by population growth (three percent), supply of labor (three percent), annual TFP growth changes of three percent in all non-agricultural sectors from 2010 to 2015 and an increase in government spending consistent with annual growth rates (three percent). Changes in growth rates in the different reform scenarios are relative to the baseline scenario are mainly due to endogenous processes such as the change of relative prices for factors and commodities from subsidy removal. Changes in public spending from subsidy reform are accounted for by exogenous changes in government transfers to households and sector level changes in TFP [37
The six types of labor included in the DCGE model captures the distinct nature of the Yemeni labor market that is mainly characterized by public versus
private employment and different skill levels. Accordingly, there are different wage rates for labor employed by the government and by the private sector. Workers are fully mobile within these employments and wage rates differ among skilled, semi-skilled and unskilled labor. With this set-up, the model can capture some of the distributional effects of growth that has characterized the Yemeni economy over the past years such that growth has been oil-driven and did not trickle down to the poor and rural areas [1
]. In fact, the segmentation of the labor market, wherein only few highly skilled laborers in the oil sector and government employees benefited from oil production and related government revenues, has been deemed as a major obstacle to pro-poor growth and rural development [22
3.2. Microsimulation Model
The DCGE model links to a micro-simulation model that allows for the endogenous estimation of growth effects on poverty reduction. The micro-simulation model uses data available for 13,136 households from HBS 2005–06. Each household’s total consumption expenditure and expenditure shares on all consumed commodity groups are linked to one of the six representative household groups included in the DCGE model according to the household’s residential area (rural or urban), employment in agriculture (farm or non-farm in rural areas) and food security status (food secure or food insecure). Relative changes in the consumption expenditure levels of the household groups, which are endogenously estimated in the DCGE model, are applied on the reported consumption expenditure levels of all survey households to obtain their new expenditure levels. Each household’s new total consumption expenditure level is related to the official poverty line (available from HBS [27
]) to determine its poverty status, based on which new poverty rates at the national level and for rural and urban areas and farm and non-farm households in rural areas are estimated. Household expenditure levels and poverty rates are computed for each year in each scenario simulation.
3.3. Baseline and Fuel Subsidy Reform Scenarios
The DCGE and micro-simulation models are applied to estimate growth rates and changes in poverty rates under a baseline scenario, which represents a continuation of the fuel subsidy policy without reform, and six fuel subsidy reform scenarios, which represent two reform options each with three alternatives of spending the reform savings. To determine the net effects of the reform options, growth and poverty rate estimates of the reform scenarios are compared with the baseline estimates, and simulation results are reported as changes from the baseline.
Reform Option 1 represents an accelerated reform path, wherein all fuel subsidies are lifted within one year; Reform Option 2 represents a more gradual reform path, where subsidies are phased out over a period of three years (
). The base year of our simulation is 2009, and we assume exemplarily that the subsidy reform is implemented from year 2011 onwards. In the accelerated reform scenario, subsidies are eliminated from an estimated 391 billion YER in 2009 to zero through a one-time removal in 2011. This would, ceteris paribus
, imply a reduction of the fiscal space by one-half, from 6.9 percent of GDP in 2009 to 3.5 percent in 2011, which yield a surplus of 215 billion YER [38
]. In the gradual reform scenario, subsidies are phased out by equal amounts (130 billion YER) from 2011 through 2013. Ceteris paribus
, the overall savings from the gradual reform are smaller than from the accelerated reform due to continued fiscal costs for subsidies in 2011 and 2012, so that the full fiscal space is also reached later.
Fuel subsidy reform options.
Fuel subsidy reform options.
| ||Reform Option 1 (accelerated)||Reform Option 2 (gradual)|
|Percentage change (percent)||–100||0||0||–33||–50||–100|
|Absolute change (billion YER) ||–391||0||0||–130||–130||–130|
|Absolute change (million US$) ||–1777||0||0||–593||–592||–593|
|Remaining subsidies (billion YER) ||391||0||0||0||261||130||0|
|Remaining subsidies (million US$)||1777||0||0||0||1185||593||0|
|Fiscal deficit (percent of GDP)||6.9||3.5||3.5||3.5||5.8||4.6||3.5|
|Fiscal deficit (billion YER)||352||176||0||0||293||235||176|
|Fiscal deficit (million US$)||1600||800||0||0||1333||1067||800|
|Surplus from reform/spending (billion YER)||215||0||0||72||72||72|
For both the accelerated and the gradual reform options, we analyzed the growth and poverty effects assuming that savings from the fuel subsidy reform are used (A) only for budget deficit reduction, (B) for budget deficit reduction and direct income transfers to the poorest one-third of households, and (C) for budget deficit reduction, direct income transfers targeted to the poorest one-third of all households, and productivity-enhancing investments in infrastructure. The simulations of the Spending Alternatives B and C require assumptions on the allocation of the reform savings and the size of the changes in productivity induced by the increases in public investments in the case of Alternative C (Table 5
). Yet the empirical evidence on the effects of infrastructure spending is ambiguous and often country-specific. A recent analysis by Dorosh and Thurlow [39
] looking at road network investments in Ethiopia assumes TFP growth rates between 3.5 and 11 percent depending on the sector. We assume equal TFP growth rates for Yemen and consistently higher growth rates in construction sectors [40
]. Accordingly, for Reform Option 1C and 2C, we assume an overall investment-growth elasticity of 0.5; that is, a one percent increase in investment leads to 0.5 percent GDP growth.
Reform scenarios (model implementation).
Reform scenarios (model implementation).
| ||Government transfers||Subsidies||TFP|
|Reform Option 1 (accelerated)|| || || |
|1A: 100% increase in government savings ||as baseline||100% decrease of subsidies in 2011||as baseline|
|1B: Increase government savings by 50% and use remainder for direct transfers to poorest one third of households||increase transfers in 2011 between 40% and 380% depending on initial size of transfers and population shares||100% decrease of subsidies in 2011||as baseline|
|1C: Increase in government savings by 50%, compensate only the poorest of the poor, and use remainder for productivity-enhancing investments||increase transfers in 2011 between 22% and 155% depending on initial size of transfers and population shares||100% decrease of subsidies in 2011||22% in construction, electricity, water, trade transport in 2011; from 2013, 1 percent TFP growth in all sectors|
|Reform Option 2 (gradual)|| || || |
|2A: 100% increase in government savings ||as baseline||33%, 50%, 100% reduction from 2011 to 2013||as baseline|
|2B: Increase government savings by 50% and use remainder for direct transfers to the poorest one third of households||increase transfers from 30%-100% in 2011 to 15%-50% in 2013 ||33%, 50%, 100% reduction from 2011 to 2013||as baseline|
|2C: Increase government savings by 50%, compensate only the poorest of the poor, and use remainder for productivity-enhancing investments||increase transfers from 20%-74% in 2011 to 20% in 2013 ||33%, 50%, 100% reduction from 2011 to 2013||7% in construction, electricity, water, trade and transport 2011-2013; from 2014, 1% TFP growth in all sectors|
There is an urgent need for reforming Yemen’s fuel subsidy scheme. Yemen is among the countries with the lowest fuel-pump prices in the world. Fuel subsidies make up 85 percent of all public spending related to economic affairs and exceed total spending on health, education and social protection combined. Investments in infrastructure and welfare spending—critical for economic growth and poverty reduction—remain at especially low levels. At the same time, Yemen faces a severe budget deficit and the costs of fuel subsidies are expected to further rise because of declining oil production and export, growing consumption and import of petroleum products. There is consensus that phasing out fuel subsidies has a large potential for the consolidation of the budget, but the direction and magnitude of the effects on growth and poverty are debated controversially. Lessons from other countries suggest that, while efficiency gains are likely to lead to significant growth acceleration after reform, poverty often increases during reform. Expected income losses motivate people’s opposition to reform and may even spark civil unrest.
This paper contributes to the debate on the economic and social impact of fuel subsidy reform by analyzing the economic linkages between the existing fuel subsidy policy, government budget allocation and the role of the subsidies for production and consumption in Yemen. In addition, it provides a thorough analysis of the growth and poverty effects of alternative reform options. The results of the analysis may thus help Yemen’s policy makers in designing a fuel subsidy reform that accelerates both economic development and poverty reduction and gains broad public approval.
Consistent with findings from other countries, our analysis shows that the direct effects of fuel subsidy reform on household real income in Yemen is likely to be modest given the low share of petroleum products in private expenditure. To also capture the indirect effects, which we found to be more crucial for household welfare, we use an economy-wide DCGE model combined with a microsimulation model. Simulation results reveal that reducing fuel subsidies would increase poverty among both rural and urban households, if all savings from subsidy reform are channeled to increased investment. Comparison of an accelerated reform scenario (that is removing all fuel subsidies within one year) versus a gradual reform scenario (that is, phasing-out fuel subsidies over a period of three years) suggest that the timing and design of reform matter: rapid phasing-out leads to an initial drop in growth and a sharper spike in poverty, while gradual reductions smoothen the growth and poverty effects. Gradual cutback of fuel subsidies over several years is therefore preferable from a growth and poverty-reduction perspective. However, slow reform comes at a higher fiscal expense, since maintaining parts of the subsidies ties up urgently needed financial resources. Hence, the faster the phasing-out of subsidies, the more fiscal space exists for the government to compensate households and to invest.
Compensating the poorest of the poor for their income losses during reform will be important for success, though it may not be sufficient. Simulation results show that direct transfer payments to the poorest one-third of all households adds up to 19,700 YER per household and year under our accelerated reform scenario and 13,800 YER under our gradual reform scenario on average. Generally, using half of all reform savings for direct transfers (and the other half for budget deficit reduction) strongly smoothes the negative effects on household incomes, but growth impulses for sustainable development are limited. Yet the impact of transfers essentially depends on the targeting and the efficiency of service delivery, too [45
Phasing out the fuel subsidies and using the savings for a combination of fiscal consolidation, direct transfer payments to the poorest and productivity-enhancing investments in infrastructure is the most promising reform strategy. In the short term, direct transfers will mitigate the welfare impact from the reform among the poorest, and infrastructural investments will enhance income-earning opportunities especially in construction. Investments in water, electricity and transport, trade and construction sectors will also lower transaction costs and allow for the integration of economic spaces across Yemen. In the medium term, reforming fuel subsidies thus offers the creation of an impetus for restructuring of productive, agricultural, industrial and service value chains, which could be exploited by enabling domestic and foreign private investment. The short and medium-term effects from a need-oriented spending of the reform savings could not only avoid setbacks in poverty reduction but also facilitate pro-poor growth in Yemen.
The Government of Yemen has made first attempts to reform the fuel subsidies by reducing fuel prices in 2010. This paper has shown that continuing this reform process offers a great opportunity for development, if the transition to higher fuel prices is designed properly and the overall petroleum subsidy reform is integrated into Yemen’s overall development strategy.