Energy subsidy reform is necessary in the Western Balkans to contain the fiscal cost of supporting the status quo in the energy sector, to incentivise the transition to renewable energy sources, and to ensure markets provide appropriate signals to prompt energy savings and energy efficiency. This chapter probes the need to balance energy sector reform with anticipated fiscal, economic and social impacts. It presents simulation results focusing on three key themes: 1) reforming retail energy pricing; 2) deregulating retail electricity markets; and 3) phasing out direct subsidies to energy.
Energy Prices and Subsidies in the Western Balkans
Reforms for a Fair and Green Future
Abstract
The energy sector in the Western Balkans is characterised by three main features: heavy reliance on coal for electricity production; the predominance of large, often state-owned incumbent firms; and price regulation that keeps retail prices low. In practice, significant variations exist across the six economies. Albania relies mainly on hydroelectric power, which is also significant in Montenegro and Bosnia and Herzegovina. North Macedonia and Kosovo have advanced in unbundling vertically integrated sectors. North Macedonia has also progressed with deregulation. While price regulation remains the norm, some economies and/or entities have recently adjusted electricity prices, including Serbia, Republika Srpska (RS) within Bosnia and Herzegovina, and North Macedonia. In the latter two, introduction of new tariff schedules opens the door for retail tariffs to more clearly reflect market price signals.
Energy subsidies and support measures play a vital role in sustaining the equilibrium in the region’s energy markets. In particular, they compensate key energy sector actors for the public service obligations and regulated prices that underpin supply of electricity at relatively low prices, particularly to households. The subsidies and support measures also sustain activity in less profitable mining operations, helping to maintain livelihoods and contributing to social peace. In turn, universally low electricity prices contribute to reducing inequality and sustaining real incomes, especially among the poor.
This equilibrium, however, comes at significant fiscal, economic and environmental cost. Lower than market prices do not fully reflect production or opportunity costs, supporting energy-intensive production at the expense of other activities. In addition, sustaining the status quo necessitates significant direct subsidies with the corresponding fiscal cost (Chapter 3), while price distortions reduce incentives for energy efficiency investments and reduce the relative profitability of renewable energy production.
Radical reform of the energy systems of the Western Balkans is necessary. It will, however, require balancing associated fiscal, economic and social impacts. This chapter presents simulation tools that support analysis and debate regarding actionable policy reform scenarios. It then probes key mechanisms and parameters at play in energy reform and presents tools and approaches that can guide policy dialogue on future reforms. The chapter first presents the key considerations motivating energy reform, it then turns to a presentation of the methods used in this report to discuss energy sector reforms. The following three sections feature selected results from the three archetypical scenarios as contained in the economy-specific chapters (Chapters 7 to 12): 1) reform of retail energy pricing; 2) expanding the scope of deregulated markets; and 3) phasing out direct subsidies. The final section discusses policy implications of the results.
The status quo of energy systems in the Western Balkans sustains domestic production while keeping prices low for households and small consumers – both at significant cost to a variety of sector actors. In fact, high levels of fiscal support to the energy sector, worth EUR 3.8 billion over the period 2018-23, has put public finances in some of the economies under stress. This need for public intervention to support the energy sector in the economies of the region contrasts with others in which support for households was partly financed through measures such as windfall taxes and solidarity contributions (OECD, 2023[1]). The specific mechanisms through which support is channelled carry the risk of distorting market competition and, in some cases, are a disincentive for incumbent state-owned energy firms to participate in the necessary green transition. Finally, subsidies sustain a polluting energy sector: electricity production generates 66% of carbon dioxide (CO2) emissions in the region (Chapter 2). Coal-fired power stations also generate significant amounts of sulphur dioxide (SO2), nitrogen oxides (NOX) and dust pollution, which deteriorates air quality locally and has negative impacts on human health (OECD, 2022[2]).
The existing toolbox of energy subsidies and support measures is not sufficiently conducive to the transition to renewable energy sources. Fixed-price feed-in tariffs for renewable energy production have been the main policy tool to encourage investment in the past. In recent years, the economies of the Western Balkans have made progress in adjusting the legal framework to support entry of renewable energy sources, with recent legislative reforms in Albania, Kosovo, North Macedonia, Republika Srpska in Bosnia and Herzegovina, and Serbia. Support mechanisms are evolving from fixed feed-in tariffs towards more market-friendly schemes (in particular auctions and contracts for difference) (Chapter 2). However, substantial financial resources are still flowing towards fossil fuels, which received up to 63% of financial support in 2018-23 (Chapter 3). While renewable energy sources received, on average, EUR 26 million per year in 2018-20, direct financial support and credit to fossil fuels averaged EUR 191 million annually over the same period.
Energy intensity of GDP, kilogramme of oil equivalent (KGOE) per thousand euro in purchasing power standards (PPS)
Note: Data on GDP in EUR PPS not available for Kosovo. The use of GDP in purchasing power standards ensures comparability by correcting the data for differences in prices across economies.
Source: Eurostat (2025[3]), Energy intensity (database), https://ec.europa.eu/eurostat/databrowser/view/nrg_ind_ei/.
Low energy prices blunt incentives for energy savings or reducing energy intensity. Energy intensity remains high across in most of the Western Balkan economies, and the economies of the region made less progress in reducing it during the 2010s than the average EU country (Figure 4.1). Energy intensity is the amount of energy used to satisfy all needs in an economy (Gross available energy) divided by gross domestic product (GDP). It measures the degree of efficiency at which an economy uses energy to generate value. In some cases, the fall in energy intensity is directly attributable to changes in the activity of highly energy-intensive sectors. For example, the notable fall in Montenegro can be linked to the end of operations of the KAP aluminium plant, which reduced production at the end of 2021 and ceased operations in June 2023. Despite the progress achieved, energy intensity in the Western Balkans fell by 29%, which compares to a 36% fall in the European Union. Energy intensity is particularly high in Bosnia and Herzegovina, Kosovo and Serbia, all of which rely extensively on coal to generate electricity. Although fully comparable GDP data on PPS is not available for Kosovo, Kosovo’s energy intensity is high: if measured at market rates (299 kgoe/thousand EUR), it is second only to Bosnia and Herzegovina’s (313 kgoe/thousand EUR) and above the regional average (230 kgoe/thousand EUR). An additional consequence of this situation is that that the degree of productivity in relation to CO2 emissions is particularly low. Indeed, Bosnia and Herzegovina generates USD 2.11 of GDP per kilogramme (kg) of energy-related CO2 and Serbia 2.81 USD/kg, compared with the OECD average of 5.46 USD/kg, according to the OECD Green Growth Indicators (OECD, n.d.[4]).
Reforms in the energy sector will need to balance impacts on the energy sector, on public finances and on the economy. Indeed, current low energy prices, for electricity and gas in particular, constitute a significant transfer of public and/or sector resources to households. It can be expected that prices for categories of consumers currently shielded by price regulation would increase under a more market-oriented energy system, at a minimum to guarantee full cost recovery. Under current practices, this transfer is quite broadly distributed and, to a degree, contributes to reducing inequality relative to market incomes (see Chapter 5). It follows, then, that increasing retail prices is likely to increase the risk of poverty among certain categories of households. Compensating such groups is necessary to ensure that the green transition is just. In turn, changes in price setting and the scope of regulation are likely to affect productive firms and sectors differently, depending on how they procure energy prior to reform.
The ability of each economy to appropriately target and redistribute income to energy-vulnerable households is key to gauge the impacts of potential reforms. This report uses the framework of the Commitment to Equity (CEQ) approach (Lustig, 2022[5]) to gauge the capacity of Western Balkans economies to redistribute income. The CEQ approach can also assess the distributional incidence of social assistance programmes, of programmes to support energy-vulnerable customers, and of potential compensatory measures that could be implemented in the context of energy reform (see Chapter 5).
Changes in energy markets will also impact the economy widely. At the macro-economic level, changes in market regulation generate a transfer of value from certain categories of households to the energy sector and to the state. This occurs through three channels: increased tax receipts on energy consumption; increased tax receipts linked to increased profitability; and dividend payments linked to improved performance of state-owned enterprises (SOEs). In addition, sectors and firms will be differentially impacted depending on their current status. At present, large consumers already procure electricity in the open market. In many Western Balkan economies, micro-, small- and medium-sized enterprises (MSMEs) have access to regulated prices that cross-subsidise (to varying degrees) lower household tariffs. Different sectors and firms will also be differentially impacted depending on the energy intensity of their activity and their ability to shift energy consumption between sources of energy. To shed light on how the various effects combine, this report applies macro-economic computable general equilibrium (CGE) modelling as a flexible tool to analyse impacts of reform in individual economies in the region.
This report relies on a Computable General Equilibrium (CGE) model to analyse the macroeconomic impacts of potential energy sector reforms. CGE models describe the equilibrium in a stylised version of an economy, after all sectors and actors have adjusted to changes in technology or external parameters. They are particularly appropriate when the changes examined are likely to have sizeable impacts on other sectors in the economy and when those impacts are likely to differ across sectors. They are also superior to alternatives (such as an Input-Output approach) in this case because they explicitly incorporate the behavioural reaction of households and productive sectors to changes in prices (Box 4.1).
The POWER-CGE model developed for this report is a single open economy model. A different instance of the model is run in the analysis presented in each of the economy chapters (Chapters 7 to 12) in this report. They are named by prefacing the model’s name with the 2-letter code for each economy. All instances share the same specification and design features (see Annex 4.A). Each instance relies on a separate social accounting matrix (SAM) as the base data for the exercise, and each instance is separately calibrated on its corresponding SAM.
The approach allows the analysis to overcome the limitations of available data sources. The usual starting point to build the data needed for CGE analysis are input-output tables for the modelled economies. In the Western Balkans, official input-output tables are available for Albania, North Macedonia, and Serbia. They are not available for Bosnia and Herzegovina, Kosovo, and Montenegro. In addition, Albania and Serbia are included as individual geographies in the latest version of the Global Trade Analysis Project (GTAP 11) (Aguiar et al., 2022[6]). The GTAP database is a multi-regional input-output database widely used for CGE analysis, in particular of trade policy, and is also one of the primary data sources for the OECD ENV-Linkages model (Château, Dellink and Lanzi, 2014[7]). The analysis relies on GTAP data for Albania and Serbia. In other cases, social accounting matrices (SAMs) for each economy are built by collecting available data from official sources, which are then used to extract data for individual economies to match available national accounts aggregates (see Annex 4.A).
The POWER-CGE model introduces a more detailed power sector to the standard CGE framework. The model isolates electricity production sectors (grouped by technology). In addition, electricity suppliers are modelled separately as grids that purchase electricity from producers and imports and sell it to their customers. The model allows for multiple grids serving different market segments. Typically, instances of the model are built using three grids respectively for households in the regulated segment, other customers in regulated segments, and all customers in the unregulated segment. Within each grid, electricity sources have the same degree of substitutability. Therefore, changes in final demand for electricity in one segment impact the power sector through derived changes in demand for individual electricity inputs. This design makes it possible to analyse how diverse energy policy reform affects both firms and households, and places emphasis on the effect of energy prices for end-users in different parts of the economy.
Computable general equilibrium (CGE) models are a popular framework to study the implications of policy reforms on the broader macro-economy (Burfisher, 2020[8]). At the highest level, a CGE model tracks the joint behaviour of all market participants in a national economy. To accomplish this, four basic assumptions are required:
1. The observed data are assumed to represent an economic equilibrium.
2. Market actors are assumed to be rational utility maximisers with complete information that respond to incentives.
3. Market actors’ behaviour is reasonably described by mathematical equations, the parameters of which are chosen, in part, by the modeler.
4. After a policy change, markets will eventually find a new equilibrium.
While unrealistic, these assumptions unlock a powerful tool for accounting not only for the direct effects of a policy change, but also higher order (“general equilibrium”) effects. For example, reducing energy subsidies for households directly affects households by raising the price they pay for electricity. It also affects firms that face reduced demand, which causes reduced demand for labour services, which further effects households, and so on. The GCE model results offer a representation of the situation after the economy has reached a new equilibrium. In short, the tool allows policy makers to “test” different policy changes and assess their potential implications in a logically consistent fashion.
Features of the POWER-CGE model
The POWER-CGE model designed for this report draws heavily from ENV-Linkages (Château, Dellink and Lanzi, 2014[7]) and the Global Trade Analysis Project (GTAP) (Corong et al., 2017[9]). Within it, the economy is comprised of six agents: households, goods firms, energy firms, electricity grids, government, and the rest of the world. As in all CGE models, households earn income by selling their labour and capital to firms and use this income to purchase consumption goods and services, and pay taxes.
The model is a single country open economy model. The model considers the economy under analysis as a small open economy which is a price taker in international markets. As such, different versions of the model are used for each of the economies of the Western Balkans (and each of Chapters 7 to 12 in this report). Such versions of the model are prefixed by the 2-letted economy code in each case (AL-POWER-CGE for Albania, BA-POWER-CGE for Bosnia and Herzegovina, XK-POWER-CGE for Kosovo, ME-POWER-CGE for Montenegro, MK-POWER-CGE for North Macedonia, and RS-POWER-CGE for Serbia).
The POWER-CGE model extends this framework by a layer to the power sector. This is done by adding energy producing firms – one representative firm per technology including both fossil fuel and renewable sources – that sell their energy to national electrical grids. Grids can be “regulated”, in which case they meet the posted demand of their clients at the regulated prices. In most versions of the model used in this report, there are two regulated grids, serving respectively household and non-household customers eligible for regulated supply. The unregulated grid sells energy for the remaining non-household sectors (typically businesses). This allows scenarios to shock the economy by independently altering household and non-household electricity prices. The “unregulated” grids, in contrast, behave as suppliers in a competitive market.
The source data does not allow the analysis of the heat energy sector, a key limitation of the approach. Data on energy flows used in this report is drawn from the GTAP database. The GTAP energy dataset is a comprehensive complement to the GTAP dataset, built primarily on the basis of energy balances and consistent with the global input-output database. The GTAP database aggregates heat energy and electricity (under the Electricity heading). Heat energy constitutes 14.8% of the aggregate in the Western Balkans as of 2023 and, as described in Chapter 3, receives several subsidies and support measures. In several cases, the price of heat is aligned to the price of electricity – in particular when heat is generated by combined heat and power plants. However, at the microeconomic level, the treatment is often different. The model used in this report can be expanded to include heat energy explicitly but it would require substantial statistical work to develop domestic heat balances and align them with social accounting matrices.
The level of aggregation of sectors is a second key limitation of the approach. The various instances of the model are built on the same 26-sector aggregation of production sectors. Production sectors correspond to NACE Rev. 2 (Nomenclature statistique des activités économiques [NACE]) (Eurostat, 2008[10]) sections, augmented by a subdivision of the Mining sector into the extraction of coal, oil, gas, and other products, and a residual sector B capturing supporting services. This level of aggregation allows the combination of the data with business statistics to calibrate the size of the regulated segments in several economies. It also allows for greater confidence in the results of the procedure to build SAMs in economies for which no input-output data are available. However, it imposes a series of limitations. The most relevant limitation for this report is that it is not possible for the standard model to examine the impact of subsidies and price regulations on hydrocarbons, which are included in the data as outputs of the manufacturing sector.
Finally, the data used in this report do not allow for a regional model to be ran. A 6-economy model for the region would require a similarly shaped regional input-output table and SAM. However, such data is not readily available. For this reason, the models are utilised independently and focus on domestic interactions. The use of a foreign economy in the model allows the integration of key mechanics linked to import dependence in the case of energy.
Evidence on the social impact of reform is based on microsimulation analysis. Where relevant, the chapter presents evidence on the social impact of selected reform scenarios. These are based on the microsimulation analysis based on the Commitment to Equity (CEQ) framework presented in Chapter 5. The POWER-CGE and the microsimulation model for each economy are not linked, but rather present results for similar parameter values. The CEQ framework allows the comparison of the impact of implicit and explicit energy subsidies across the distribution of income and can compare them with the effectiveness of social programmes to reach those in need and therefore compensate vulnerable populations. However, the CEQ framework does not include a demand model and would therefore tend to overestimate the impact of changes in markets on household wellbeing measures, such as the degree of income or energy poverty.
The following sections present results from policy scenarios for the economies of the Western Balkans. The scenarios presented in this report were selected and parametrised on the basis of specific considerations for each of the economies in the region, including the results of the Inventory of Energy Subsidies and Support Measures in the Western Balkans. Therefore, identical scenarios are not run in each of the economies and the results correspond to independent scenarios analysed in each of the economy-specific chapters of this report (Chapters 7 to 12). However, for a selected number of issues, similarly parametrised scenarios were run for several or all economies. The remainder of this chapter focuses on three key issues with the help of results from policy scenarios:
Reform in retail energy prices,
Progressive deregulation of electricity markets, and
The reduction of direct subsidies to the energy sector.
Across the Western Balkans, a large share of retail electricity sales take place in regulated markets, with below-market prices. The overwhelming majority of households, as well as most small and medium (SME) firms and in some economies, even larger firms, are served under regulated prices set through price regulation or set with political considerations via state ownership of incumbent electricity producers or suppliers. According to the Inventory presented in Chapter 3, if electricity is valued at international market prices, existing price interventions effectively transferred, on average, EUR 2.9 billion per year to final consumers1 over the period 2018-23. As discussed in Chapter 3, the level of the estimated induced support is heavily influenced by the period of high international energy prices in 2021-23. However, even before the energy crisis, the value transferred to final consumers through implicit subsidies was as high as EUR 655 million per year across the region.
The first set of scenarios examines the impact of changes in electricity market regulation that would drive retail prices towards market prices. Since prices do not currently reflect market signals, the increase in tariffs reflects a situation where price interventions would be modified to converge towards market prices. This can be the result of a number of different reforms. A change in regulation could drive tariffs up, for example through ensuring that the full cost of energy procurement is included. But increases in prices can also result from reforms that introduce market elements into the determination of retail prices in other ways. For example, reducing the public service obligation of incumbent producers to supply the needs of universal service suppliers (USS) at low prices would force the latter to purchase a larger fraction of their needs in the open market, thus increasing their costs.
Market-consistent prices provide an alternative benchmark for electricity prices
This report relies on the set of market-consistent prices to reflect an alternative situation where prices would be determined by the market. Market-consistent prices as constructed as part of the Inventory of Energy Subsidies and Support Measures in the Western Balkans. Regulated prices in the region are chiefly determined through a process that identifies total costs of provision on the part of the supplier, determines maximum allowed income from regulated activities, and apportions that income between market segments to determine the tariff structure. Allowable costs include the cost of energy, network charges, operational or supply costs, as well as a profit margin for the supplier. For every economy, market-consistent prices correspond to the situation where regulated suppliers would procure electricity at international prices. Network charges and operational margins are added to international prices to construct the series. Where data on network charges is available for different market segments market-consistent prices are built for different segments – for example, for households and non-household customers and for customers connected at different voltage levels. Across the analysis, the annual average of day-ahead market (DAM) prices from the Hungarian Power Exchange (HUPX) is taken as the international reference price. As a simplifying assumption, network charges are applied as per the recorded data – that is no adjustment is made for the fact that higher wholesale prices would lead to higher network charges as network operators purchase electricity in the market to compensate for network losses.
The market-consistent prices calculated in this report can be considered as an upper bound of prices in a liberalised market, especially during the energy crisis. Indeed, they correspond to a situation where energy suppliers would purchase all their electricity in the open market at spot prices. The report refers to these as market-consistent because they correspond to a hypothetical open market situation but are not observed in the market. In practice, most wholesale exchanges in the region take place through bilateral contracts, which can take place at significantly lower prices and include stabilisation elements. In addition, a fraction of the purchases made by retail suppliers include electricity produced by renewable energy producers covered by incentive schemes. Traditionally, these are long-term take-or-pay power purchase agreements at favourable prices to the producer. Since they are multi-year contracts, they provide a certain degree of insulation against the volatility even of annual price movements, such as those observed in 2021-23.
Average prices for households and non-household customers served under regulated prices increased notably in 2022 and 2023 after remaining stable in the period 2018-21. On average, non-household regulated prices increased by 24% between 2021 and 2023, while household prices increased by 17%. In contrast, between 2018 and 2020, prices increased by 2% for households and by 4% for non-household customers. This evolution hides notable regional disparity. Price increases in the 2021-23 period were particularly pronounced in Serbia (34% for households, 31% for non-household customers) as price adjustments took place to shore up the finances of state-owned EPS, and in North Macedonia (29% for households, 46% for non-households) as exposure to international markets drove procurement costs up. In contrast, prices in Montenegro stayed stable (falling by 2% and 3% for household and non-household customers respectively).
Price regulation and intervention in electricity markets has contributed to keeping prices down in the Western Balkans in the past 6 years. Even though prices increased in the region, they did not reflect the magnitude of changes in international markets. The average market-consistent prices largely follow the evolution of international prices, proxied by the HUPX DAM average, by construction. Average HUPX prices increased rapidly during the energy crisis and were at 272 EUR/MWh in 2022, over five times their historical values in the pre-COVID-19 period. Despite the increase in prices in 2021-22, as of 2023, household prices remained below international wholesale prices, and therefore well below market-consistent prices (Figure 4.2).
Regional averages of regulated prices and market-consistent prices
Note: The series presented are simple average of economy-specific series.
Source: Authors' calculations based on regulator reports and the Inventory of Energy Subsidies and Support Measures in the Western Balkans.
Gaps between regulated and market-consistent prices increased notably during the energy crisis. Market-consistent prices were substantially higher in 2023 (161 EUR/MWh) than before the energy crisis (93 EUR/MWh on average in 2018-19). The before-crisis data is taken as indicative of the long-term value of these quantities; the reference period excludes 2020 because the fall in economic activity due to the COVID-19 pandemic. In contrast, average data for the year 2023 includes the tail end of the energy crisis – prices moderated at the end of the 2022/23 winter. However, average prices have stayed high in 2024, indicating that there may not be a return to the pre-COVID-19 normal, and that higher prices may indeed be the new long-term equilibrium. These are therefore considered as valid benchmark prices.
Note: Non-household price for North Macedonia refers to small customers only in 2023.
Source: Authors’ calculations based on annual reports of energy regulators and on the Inventory of Energy Subsidies and Support Measures in the Western Balkans.
The price adjustment scenarios use the 2018-19 and 2023 market-consistent prices as bounds for analysing the impact of price changes induced by reform. This approach is consistent with the methodology implemented in the Inventory presented in this report.2 Figure 4.3 provides a visual representation of the gaps between average prices for household and non-household customers and the average market-consistent prices. As prices are higher for firms in regulated segments than households, gaps between recorded and market-consistent prices are systematically larger for households. Where available, specific gaps for each segment are used. By construction, they differ only in the network charges applied to each market segment so that differences are small. These are not comparable across economies because the set of non-household customers considered differs (small customers in Bosnia and Herzegovina, Serbia and North Macedonia, all USS customers in Albania, all firms in Kosovo where price data is available for broad categories only, and in Montenegro, where prices are not regulated but all customers are served by the state-owned incumbent).
The gaps applied in scenarios therefore vary by economy and by type of customer. Lower bound gaps are particularly large in Albania (51% for households), Bosnia and Herzegovina (37% for households) and Kosovo (39% for households). In several economies, non-household customers appear to have been served at prices consistent with the market, including in North Macedonia (where average non-household prices were 22% above the benchmark in 2018-19), in Kosovo, and in Montenegro. Upper bounds are significantly larger across the board, averaging 93% for households, and 45% for non-household customers).
Economic impacts of price reform
The main channel by which electricity price deregulation would impact Western Balkan economies is through the transfer of funds from households and other consumers to the energy sector. On the one hand, households and regulated consumers would no longer benefit from low prices for which the sector covers the cost and face higher prices. This leads to a fall in household consumption expenditure that drives aggregate demand down. In addition, productive sectors that use electricity as an input see increases in their production costs, which drive prices to increase further, also affecting aggregate demand. On the other hand, since demand for electricity is relatively inelastic, the increase in prices that result from reform lead to an increase in value-added and profitability in the energy sector.
The impact of price regulation on public finances can compensate the contractionary impact of price increases. Price liberalisation impacts public finances through three separate channels. First, it changes the composition of indirect tax revenue: taxes levied on electricity increase and taxes levied on other goods fall as demand falls. Second, the increased profitability of energy sector firms increases direct tax proceeds. Third, the incumbent energy producers and suppliers who benefit from the reform are state-owned enterprises. Therefore, the budget can receive transfers in addition to tax revenues in the form of dividends. In the cases where the firms are not SOEs, these funds would need to be collected through other means, such as special taxes. The first two channels are embedded in the POWER-CGE model; however the third one is not, and will be discussed separately. In the POWER-CGE model, any additional income earned by the government is spent in proportion to the government demand as recorded in the SAM. Therefore, the fiscal multiplier is a key determinant of whether government consumption can compensate for the falls in private demand.
Macroeconomic modelling helps determine how these multiple effects combine. To provide a synthetic view of the results, the key results from across the economy chapters of this report (Chapters 7 to 12) are summarised in this section. The section pays particular attention to the impact on aggregate output – which indicates whether the fiscal channel compensates for the fall in private demand – and to the impact on household consumption – as a summary measure of the impact on households which can quantify the dimension of the social impact of the reform.
Removing cross-subsidies has positive impacts on aggregate output
Cross-subsidies remain prevalent in retail electricity markets in the Western Balkans. Firms in the region generally pay higher prices for electricity, contrary to the situation in most EU countries. It is generally costlier to supply electricity per unit to residential customers than to larger industrial customers, who buy higher volumes and can be connected at higher voltage rates, lowering losses. On the other hand, residential customers are more likely to benefit from lower prices at non-peak hours. According to the Inventory of Energy Subsidies and Support Measures in the Western Balkans, cross-subsidies ranged in value from EUR 5 million per year in Montenegro to EUR 33 million per year in Albania.
To analyse the impact of cross-subsidies, two analyses are carried out in Chapters 7 to 12. First, a scenario is simulated that brings the price of electricity per unit for each of the regulated segments to the same average. This implies an increase in unitary prices for households, and a corresponding fall in the prices offered to non-household customers in the regulated segment. In Albania, where the baseline AL-POWER-CGE model allows for customers supplied by the supplier of last resort to belong to the regulated segment (to account for subsidies received through price regulation in the period 2022-23, this simulation omits medium-voltage customers from the regulated segment. Second, scenarios are run that shift the retail price for all regulated customers by the same amount, and then that shift the average price by the same amount but effect some degree of convergence in prices by making them tend toward the segment-specific (household and non-household) market-consistent price. The second set of simulations was carried out for both the lower bound average gap (corresponding to the pre-crisis period) and the upper-bound (corresponding to 2023).
Removing cross-subsidies generates aggregate gains for the economy. The simulated removal of cross-subsidies generates aggregate output gains in all the economies of the Western Balkans. Across the region, households make up the largest group in regulated segments, both in number of customers but also in volume. As a consequence, when shifting prices to equate household and non-household customers, the falls in prices for non-household customers are larger than the increases for households. The increase in prices lowers households’ real disposable income and generates a substantial fall in electricity demand. However, the fall in prices for productive sectors and the increase in government spending – fuelled by increased tax revenues – drive an aggregate output increase. In turn, in the new equilibrium, household consumption is in fact higher than in the baseline (Table 4.1).
Changes relative to baseline (%)
Note: For Albania, the table reports results from the scenario assuming medium-voltage customers are not affected by the change in price.
Source: Authors’ calculations.
Larger initial cross-subsidies and more regulated markets drive larger simulated impacts. The impact of the simulated removal of cross-subsidies depends critically on the size of the initial subsidy. Since a key factor driving the increase in output is the fall in prices for productive sectors, the effects are larger where prices were initially higher. This is the case in North Macedonia, where energy prices (excluding network charges) were three times higher for non-household than household customers as of 2023 (ERC, 2024[11]). Impacts are also larger in cases, such as Kosovo, where a large share of non-household customers are considered in the analysis due to the manner of determination of prices.
In practice, cross-subsidies should be phased out in the process of market reform. The isolated analysis of cross-subsidies is a useful analytical device, but policy should integrate the phasing out of cross-subsidies with broader reform. The scenarios depicted in Table 4.1 represent relatively small changes in household prices – of the order of magnitude of annual price changes effected in the region in since 2021. However, since average prices across the region do not reflect market conditions and in some cases do not reflect costs, it would not be efficient to lower prices for non-household customers in most cases as they would go down to inefficient levels – with the exception of North Macedonia where a more active and contestable supply market may help bring prices down. In addition, the removal of cross-subsidies is an explicit target of electricity sector management (for example in Bosnia and Herzegovina, where the State regulatory commission and the entity regulatory commission work towards the gradual elimination of cross-subsidies (SERC, 2024[12]).
Rebalancing prices towards market-consistent levels requires complementary measures
The main scenarios to analyse the impact of price shifts effect an increase in average prices and a convergence between the prices of household and non-household regulated customers. In light of the results on cross-subsidies, the lower-bound and upper-bound scenarios for price rebalancing are those that incorporate a convergence in prices between segments. Chapters 7 to 12 examine the differential impact of shifting prices uniformly and doing so along with convergence. The main impact of a price increase is to drive a contraction in aggregate demand and in total input. In all cases the smaller relative increase in non-household prices when rebalancing is introduced counteracts this effect, albeit to varying degrees. Whenever rebalancing scenarios would require a significant fall in prices for the non-household segment, this is censored as it corresponds to cases where those segments are served at market-consistent prices that may not be affected by shifts in regulation.
Shifts to output and household consumption relative to the baseline in lower bound and upper bound scenarios (%)
Source: Authors’ calculations.
The adjustment of prices towards market-consistent prices has positive macroeconomic impacts when it is moderate. Across the economies in the region, the smaller-sized adjustment drives increases in output ranging from 0.10% in Bosnia and Herzegovina to 0.26% in Montenegro. In contrast, the larger-sized adjustment produces mixed results. The growth impact remains in North Macedonia and Serbia, but the scenario is essentially output neutral in Bosnia and Herzegovina, and it leads to a contraction in Albania, Kosovo and Montenegro. The economies where the negative impact is larger are also those where the share of the productive sectors that is concerned by the shift is larger as calibrated. Indeed, in Montenegro, price shifts are assumed to cover all firms, while in economies where they correspond to regulated segments, the share of regulated electricity is approximated by the sector-wise share of SMEs in total value added.
Price adjustments also affect the sectoral composition of output. The most affected sectors are those that are most energy intensive, but only when they are concerned by price regulation. In Kosovo and Montenegro, where they are supplied by the same incumbent firm, they are included in the affected sectors. In Albania, medium-voltage sectors are still included among regulated segments in the data. In most other economies, regulated prices are only offered to small and medium-sized firms. This means that the simulated shift in prices does not affect them. The increase in government revenue also affects the composition of output, as sectors commanding a larger share of government expenditure (construction, education, public administration, health) see an increase in output.
Price reform can heighten vulnerability in the population
The adjustment to market prices has large impacts on households. At the aggregate level, the simulated price increases – consistent with full market opening and marginal cost pricing – have significant negative impacts on household consumption expenditure. In the lower bound scenarios, they range from a 0.65% reduction in North Macedonia to a 2.11% reduction in Serbia. In practice, as shown in individual economy chapters, introducing price rebalancing leads to smaller declines in household consumption, as it preserves economic activity. For the upper bound shifts, the falls in consumption expenditure are even larger, ranging from 0.71% in Montenegro to -3.92% in Serbia.
Beyond their aggregate impact on households, price regulation reforms would have significant impacts on poverty and inequality. The implicit universal subsidy embedded in below-market prices transfers significant value to many households that do not need it. Those at the bottom of the income distribution receive less benefit in absolute terms, but what they do receive is proportionally much larger in relation to their incomes. In Serbia, for example, 56% of the benefits linked to low prices go to the top half of the income distribution. Yet the transfer of value represents 14.5% of disposable income for households in the first decile, compared with just 5.7% on average.3 In addition, implicit subsidies tend to reduce inequality, because even though the better off consume more electricity, they do so less than proportionally, therefore the transfer of value reduces inequality compared to market incomes (see Chapter 5).
Simulated electricity price increases lead to increased levels of both monetary and energy poverty. Building on microsimulation methods, this report analyses the impact of energy subsidies on people across the income distribution, as well as the impact on poverty and energy poverty (see Chapter 5). In North Macedonia, a 30% rise in electricity prices (as modelled above) is estimated to increase the absolute poverty rate by 2 percentage points (based on the USD 6.85 per day threshold), a sizeable increase starting from a baseline of 11.6%.4 A 70% price increase leads to a much sharper rise, with the poverty rate increasing by 5.2 percentage points (Petreski, forthcoming[13]).
Electricity price increases have particularly pronounced effects on energy poverty. When measured as the share of households spending 10% or more of their income on energy, energy poverty rates are 7% in Albania, 39.2% in Serbia, and as high as 84.7% in North Macedonia (although data in North Macedonia have serious deficiencies in this respect).5 In North Macedonia, a 10% price increase leads to a 4.6% rise in energy poverty, while a 70% increase results in a 16.6% rise (Petreski, forthcoming[13]). In Serbia, the impact is even more severe: a 10% increase raises energy poverty by 4.4%, and a 70% increase pushes it up by 33.1% (Vladisavljevic and Zarkovic, forthcoming[14]). In Albania, using slightly different simulation parameters, a 24% price increase raises energy poverty by 16%, a 40% increase by 32%, and a 77% increase leads to a 65.1% rise (Zhabjaku Shehaj, forthcoming[15]).
Compensation mechanisms will be necessary to prevent increases in both monetary and energy poverty. In North Macedonia, a 17% electricity price increase would generate an additional EUR 99 million in revenue for the electricity sector. Redirecting a small portion of this (EUR 14 million) toward social assistance through the Guaranteed Minimum Assistance (GMA) programme, by increasing benefit levels and adjusting the income threshold to reflect real income declines, would nearly offset the rise in extreme poverty (reducing it to just 0.02 percentage points). A slightly larger allocation of EUR 21 million would be sufficient to fully mitigate the impact on moderate poverty as well (Petreski, forthcoming[13]). In Serbia, by contrast, significantly larger adjustments would be required. To offset the effects of a 70% electricity price increase, the budgets for the Subsidy for Energy Vulnerable Consumers SEVC or the Financial Social Assistance (FSA) programmes would need to rise by approximately EUR 128 million and EUR 190 million, respectively, to adequately protect those at risk of poverty6 (Vladisavljevic and Zarkovic, forthcoming[14]).
Increased fiscal resources can support the green transition and finance compensatory measures
Energy market reforms have the potential to generate significant fiscal revenues. If energy market reforms drive price shifts comparable to those simulated, they would generate sizeable incomes for the budget. In all baseline scenarios, tax revenues are estimated to increase. As modelled, the increase in revenues from indirect depends on the aggregate impact on demand. The existence of special lower VAT rates for electricity (as is the case in Kosovo) could drive indirect tax revenues up even with constant aggregate demand. In addition, the increased profitability of energy sector firms also drive up tax revenues.
The simulated increase in tax revenues is substantial. It averages 1.1% for the lower bound scenarios and 3.2% for the upper-bound scenarios. In relation to 2023 tax revenues, these amount to EUR 10 million in Montenegro to EUR 569 million in Serbia in the lower bound scenario (Figure 4.4). These estimates may be biased upward by the fact that the model enters all taxes as ad valorem taxes so that price increases automatically increase all contributions.
Impact of price reform on total tax revenue of lower bound and upper bound price scenarios (in EUR million)
Note: lb: lower bound; ub: upper bound.
Source: Authors’ calculations.
Price reforms also generate significant revenues for the energy sector. In the POWER-CGE model, as is standard practice in CGE models, firms are subject to a zero-profit condition. Therefore, the model cannot directly generate estimates of firm income (additional income in paid in capital returns to the firms’ owners, the households). To complement the POWER-CGE results, the outcome of individual simulations in terms of equilibrium prices and demand by the individual segments (“grids” in the terminology used in the model) is used to calculate the additional earnings that the energy sector firms would gain. The analysis is circumscribed to the incumbent SOE producers or suppliers. In some cases, implicit subsidies are channelled through control of wholesale prices, as is the case with the bulk supply agreement between state-owned Korporata Energjetike e Kovovës (Kosovo Energy Corporation, [KEK]) and the USS in Kosovo and the price-setting policy for sales from the state-owned power utility company (Korporata Elektroenergjitike Shqiptare [KESH]) to the USS in Albania. In those cases, the increase in wholesale prices automatically increases the supplier margin, which is included in the estimated additional earnings. While the model automatically includes increased fiscal revenues from corporate taxes on these firms, the size of the estimated additional earnings suggests that they could be tapped in additional ways (Figure 4.4). The additional income has two main components: the increase in revenue from serving regulated demand, which increases despite the price increase, and the ability to sell additional electricity in the open market given the fall in demand from regulated segments.
The proceeds of reform can be used to mitigate the social consequences of reform and to finance necessary investment to sustain the green transition. Increased profitability for the energy sector would eliminate the need for direct support to such companies. It also underlines the importance of improving corporate governance in those sectors. Additional earnings can be used to finance necessary investment in the electricity sector. Although the analysis assumes earnings go to the producer or supplier, necessary investments in the grid would be passed on through network charges and thus change the distribution of earnings rather than the total amounts. In addition, governments may want to mobilise those funds to benefit wider development objectives. In particular, in all cases, the respective firms are state-owned. First and foremost, these earnings negate the need for direct support, as companies can make normal profits in the market. In addition, profits can be tapped through dividend policy to finance other development priorities.
Additional income for the energy sector and funds needed to compensate vulnerable populations (EUR million)
Note: Compensation to vulnerable populations refers to the poorest 20% in Albania, Kosovo, those at risk of poverty in Bosnia and Herzegovina, Montenegro and Serbia. In North Macedonia, it corresponds to the necessary increase in guaranteed minimum income to prevent an increase in poverty.
Source: Authors’ calculations.
Additional earnings are much larger than the necessary funds to compensate the most vulnerable categories of the population. The estimates of additional income can be compared with estimates of the size of mitigation measures. These measure the size of financial compensation needed to protect the most vulnerable from the price increase. Compensation needs are estimated on the basis of microdata and correspond to the increased electricity consumption cost for the vulnerable population without accounting for a demand response. In other words, those households would be able to purchase the exact same bundle if given such compensation – even though they would in practice substitute away from electricity and be better off. In all cases except North Macedonia, these correspond to total amounts transferred. In North Macedonia, displayed amounts correspond to the necessary increase in the budget of the main social assistance programme (GMA). This incorporates the ability of GMA targeting to reach the poor, and therefore results in higher estimates. Excluding North Macedonia, the necessary compensation is, on average 14% of total additional income. Therefore, if resources can be mobilised and channelled to those in needs, only a fraction of proceeds is necessary.
The deregulation of retail electricity markets has progressed at different pace across economies in the Western Balkans. Across the region, households can choose to be served the USS, and the overwhelming majority are served by the USS at regulated prices. There are greater differences across economies with respect to firms. Across the region, small non-household customers can choose to be served by the USS. In Albania, medium voltage customers were served by the supplier of last resort (SLR) and have been gradually transferred to the open market which now serves those connected at 10 kV, 20 kV and 35 kV while the USS serves all those with low-voltage connections. At the other end of the spectrum, in North Macedonia, 40% of the electricity was supplied at non-regulated prices as of 2023, and in Serbia, only about 20% of electricity to non-household low-voltage customers is supplied at regulated prices. Finally, in Montenegro, prices are deregulated, although all customers are served by the incumbent SOE utility.
The degree to which SMEs benefit from regulated prices varies. In most Western Balkan economies, non-household regulated customers (including MSMEs) pay higher prices than households for the same basic electricity service, which effectively means non-household customers cross-subsidise household customers. To the extent that the tariff schedule represents total procurement costs, non-household regulated tariffs are closer to procurement costs. In practice, certain firms may benefit only marginally from regulated prices, although specific subsidies also exist to certain sectors (e.g. bakeries in Albania have lower prices than other SMEs). There also benefits to SMEs from remaining under regulated prices. Regulated prices for non-household customers are also shielded from variability in international prices and provide greater stability. In addition, some SMEs may not have the necessary sophistication to benefit from a wider set of price schedules on offer in the open market.
While good reasons exist to shield SMEs from high and volatile prices, systematically subsidising electricity is inefficient. SMEs are typically less energy-intensive than larger firms; as such, they are less exposed to the impacts of changes in energy prices.7 However, SMEs are more vulnerable to shifts in input prices as their capacity to cope with negative impacts is lower. This has been a consideration for energy subsidy reform in developing countries (Tambunan, 2013[16]). It also led many OECD governments to support SMEs during the energy crisis by putting in place measures such as price caps and energy subsidies. OECD guidance suggests that such measures should ideally be targeted to otherwise viable enterprises that are genuinely vulnerable to an energy price shock – and should be temporary. Finally, they should be coupled with measures to improve energy efficiency (OECD, 2023[17]; Hemmerlé et al., 2023[18]).
Simulating progressive deregulation in retail electricity markets
To assess the impact of further deregulation, the model is shocked with a reduction by half of the share of electricity served by the regulated grid across sectors. Data on electricity consumption by sector and voltage level is not available across the region, so that the sectoral distribution of firms affected by a reform such as the further deregulation of medium-voltage firms in Albania cannot be precisely parametrised. In practice, the model includes a vector of parameters that distributes the value of electricity used by each productive sector between the unregulated supplier and the non-household regulated supplier. This vector is divided by two in the simulation.
Halving the share of electricity procured by firms in regulated markets (% change relative to baseline)
Source: Authors’ calculations.
Progressive deregulation has positive aggregate impacts on the economy. In Bosnia and Herzegovina, North Macedonia and Serbia, the simulated deregulation results in increases in both output and household consumption (Table 4.3). The main mechanism bringing about this result is the fall in the domestic unregulated electricity price. As the regulated grids no longer have to supply electricity at fixed prices to those customers, wholesale volumes are transferred to the unregulated grid, where demand increases by a smaller amount driving prices down. The fall in prices and the increase in aggregate output drive household consumption up.
Impacts vary across sectors, which can negatively affect employment. The sectors most widely covered by regulation are the ones most negatively affected. The accommodation and food services sector, for example, suffers significant falls in output across simulations (2.2% in Bosnia and Herzegovina, 1.8% in North Macedonia, and 1.3% in Serbia). The energy sectors and electricity production sub-sectors and see output falls due to the fall in prices. In contrast, more energy intensive sectors benefit from the reform, so that manufacturing output increases (by 0.9% in Bosnia and Herzegovina, 0.9% in North Macedonia, and 1.2% in Serbia). Since some of the sectors with higher regulation shares are labour intensive, the reform leads to a small fall in aggregate employment of the order of a basis point (0.02% in Bosnia and Herzegovina, 0.03% in North Macedonia, and 0.04% in Serbia).
The size of the simulated impacts is sensitive to the parametrisation of regulated segments. As detailed in the specific economy chapters (Chapters 7 to 12), the share of regulated electricity in electricity input by sector is approximated by the share of SME value added across sectors. Where this share is significantly higher than the share of regulated electricity inputs across the economy, alternative parametrisations are run to check the robustness of results. For example, in the case of Serbia, the same simulation undertaken assuming that no electricity input to manufacturing is regulated. In that case the impacts are qualitatively similar but smaller, with an impact on output of 0.07%.
Support to fossil fuels commanded EUR 3.1 billion of budgetary resources in the Western Balkans in 2018-22 (Chapter 3). Most of these resources were supporting electricity or heat produced (in whole or in part) from fossil fuels. However, a sizeable share (about EUR 686 million) was allocated directly to fossil fuels. Given the limitations of the source data and the model, this section focuses on subsidies to fossil-fuelled electricity and includes the case of natural gas supply. However, it does not include subsidies to petroleum products, in particular for transport, as these cannot be adequately identified in the datasets constructed for this report.
Direct support to fossil-fuelled electricity distorts markets and incentives
Direct support to fossil fuels for electricity production is particularly distortive. Direct support to coal mining and coal use for electricity production distorts the costs of coal production and procurement. In energy systems in which producers of coal-based electricity are regulated, the distortion is passed on through the value chain. As a consequence, prices do not reflect the full cost of production. With the exception of Montenegro, Western Balkans economies do not have carbon pricing mechanisms in place. Continued use of public funds to support fossil fuels is at odds with findings (from multiple simulation studies) that carbon pricing would have positive impacts on economic growth and on GHG emissions reduction (World Bank, 2023[19]).
Systematic subsidies to fossil fuel undertakings, particularly for coal mining activities, blunt incentives to increase productivity. Certain mining operations in Bosnia and Herzegovina have overly large workforces, with excessively low ratios of production workers to administrative workers. On average, coal mines in the Western Balkans have lower productivity (1 632 tonnes per full-time equivalent workers [t/FTE]) than the EU average (4 730 t/FTE). This low productivity stems, in part, from geological factors and modes of exploitation, with underground coal mines having significantly lower productivity. Organisational factors and ageing equipment also play important roles, with state-owned firms exhibiting lower productivity (Ruiz et al., 2021[20]). One long-running mode of support in the Western Balkans has been allowing the accumulation of arrears from mines. In Bosnia and Herzegovina, mines have accumulated arrears both to the electricity supplier Elektroprivreda Bosne i Hercegovine (EPBiH) (and its parent company) and to the state, particularly to the social security fund. This debt makes restructuring more complex.
Direct support to fossil-fuelled electricity has limited macroeconomic impact but distorts the energy mix
Direct support to fossil-fuelled electricity and to gas has limited macroeconomic impact. The impact of schemes supporting coal and gas production and distribution in the region is analysed by simulating their removal. The selected schemes include in particular those identified in the Inventory of Energy Subsidies and Support Measures in the Western Balkans that are larger in size or have been implemented in continuous fashion. As an exception, the coal royalty exemption granted to coal extraction during the COVID-19 recovery period is also simulated to showcase how different subsidy mechanisms can have differential effects. The schemes simulated have relatively modest size, and therefore relatively modest impact (Table 4.4). All instances of support to coal production have negative impact on GDP, and their removal has modest expansionary impacts ranging from 0.01% to 0.06% basis points. This negative impact is largely due to the use of fiscal resources, which the model otherwise attributes to government expenditure. In contrast support to gas supply in Bosnia and Herzegovina and Serbia is found to have a positive impact on output, albeit also a very small one.
Simulated impacts of removing schemes providing direct support to fossil fuels
Source: Authors’ calculations.
Direct support to coal sustains a fossil-fuelled based electricity sector. In contrast to the modest macroeconomic impacts, the schemes analysed have sizeable effects in the production structure. For example, they support production of electricity from coal by 3% in Serbia through direct support to coal production. Explicit and implicit tax incentives are found to have particularly large impacts. This amounts to a 6% increase in electricity production from coal in the case of Kosovo’s recovery programme, a 7.8% shift in electricity supply from coal in the case of the simulated tax to reflect unpaid taxes and contributions from coal mines in Serbia in the period 2018-20. The relative impact of tax incentives compared to direct transfers is also a consequence of the CGE framework used. Indeed, since there are no capacity constraints in the model and all sectors are subject to a zero-profit condition, the impact of a direct transfer is to increase the size of a sector, but it has no incidence on modelled behaviour. In contrast, taxes impact prices and quantities in equilibrium and their impacts are better estimated in a CGE framework.
Social impacts from direct support concern mainly employment in mining sectors
Direct support is also found to have relatively large social impacts through employment. By supporting activity in coal mining, direct transfers also support employment. Simulations suggest that the removal of direct transfers to coal mining would lead to falls in employment in mining and in the electricity sector. For example, in the case of transfers in Serbia to the Resavica mining company, they are estimated to sustain 1.3% of labour demand, with similar results in other cases. The quantitative estimates should be taken as indicative, since the model allows for firms to adjust production to match equilibrium market prices, but does not account for capacity constraints or fixed costs, so that it cannot account for situations where production costs are too high for mines to participate in competitive markets.
In contrast, impacts on consumers through prices are relatively modest. One possible channel for direct subsidies to energy production to affect households is through electricity prices. Support for coal production can lower procurement costs for firms and thus drive lower electricity prices for consumers. In the POWER-CGE model, electricity from one source is treated as an imperfect substitute of electricity form another source (including foreign sources). Therefore, in practice, falls in production from a given source are readily substituted, which limits the price impact. Impacts are larger where the simulated policies affect incentives (such as the case of taxes and tax incentives), but even in the case of the royalty exemption in Kosovo, they remain modest (1.7% price increase for a 12% tax incentive).
Energy market reform promises positive macroeconomic impacts but requires complementary social measures
This chapter presents simulations of changes in prices that could result from market opening and deregulation in the Western Balkans. Simulations of the impact of prices converging towards market-consistent prices show that moderate price increases would have positive macroeconomic impacts. On the other hand, shifts that fully reflect the cost of electricity in wholesale spot markets would harm competitiveness and output. In addition, even price shifts that have positive impacts on output drive a significant fall in household consumption expenditure ranging from 0.65% to 2.3% in the scenarios converging to pre-energy crisis market prices. Therefore, it is necessary to combine energy market reforms with compensatory mechanisms that protect those in need.
Energy market reform can pay for the necessary compensatory measures. Simulations of price changes show how the implicit subsidies identified in the Inventory presented in this report would be reduced under market prices. This would significantly increase earnings in the energy sector. Moreover only 17% of those additional earnings would be necessary to compensate the most vulnerable population. Market reforms also drive significant increase in tax revenues which could finance both compensatory programmes, as well as broader development objectives.
There are multiple complementary entry points for energy market reform
Electricity prices in the Western Balkans remain well below where they would be if markets were fully open. The market-consistent prices calculated as part of the construction of the Inventory of Energy Subsidies and Support Measures in the Western Balkans illustrate where prices would be if wholesale procurement took place at average annual international day-ahead markets. Prices for households in particular are significantly below this benchmark (by 40% and more before the energy crisis, almost half the benchmark in 2023), although there is significant variation across economies in the region. Reference prices as of 2023 should be considered as an upper bound of where open market prices would be. Indeed, in practice, long term contracts in wholesale electricity markets reduce the impact of transitory changes in market conditions.
While simulations of price shifts presented in this chapter focus on the levels of regulated prices, multiple approaches to reform can drive prices to converge towards market prices. The results of simulations suggest that cross-subsidies are inefficient and should be phased out. Removing the scope for cross-subsidies in network and energy tariffs can be driven by regulation. On the other hand, the low procurement prices of regulated firms (both USS and in some cases SLRs) do not necessary result from regulation. Rather, politically-influenced pricing policy on the part of SOE incumbent producers and suppliers drives low retail prices. In addition, contractual obligations, such as in the case of the bulk supply agreement in Kosovo, also sustain low retail prices in the region. Gradual increase of regulated prices so that they cover all supply costs and are reflective of the market can contribute to increase efficiency, improve incentives for energy efficiency and energy saving, and generate financial resources for investments in the energy transition. However, if energy sector reform seeks to encourage the formation of open, competitive markets with market-determined prices, allowing market signals to play a much greater role in price determination is of particular importance.
Progressively expanding deregulated retail electricity markets has positive aggregate impacts. Simulations presented in this chapter show that reducing the size of the non-household regulated market by half generates positive output and consumption outcomes. While it does harm sectors that are still largely served under regulated prices, the overall impact on employment is found to be particularly small.
The lack of maturity of retail markets in Western Balkan economies is an important reason why, small- and medium-sized enterprises (SMEs) are served through universal service supply. In principle, firms across the region could be served by incumbent suppliers at market prices. When incumbents hold a dominant position, however, it is less likely that markets can produce prices that lead to efficient allocation. In some economies (e.g. Albania or Kosovo), only large customers connected to medium and high voltage lines participate in the open market and their needs and degree of sophistication are very different from that of most MSMEs.
A necessary complementary approach to expand the share of electricity traded in open markets is to alter public service obligation provisions. A key mechanism that leads to low retail electricity prices in the Western Balkans is the obligation of incumbent producers to cover most (or all) of the demand for electricity of USS and SLR suppliers, in several cases at below-market prices. Since USS and SLR suppliers are regulated, low procurement costs translate into low retail prices. In addition, this obligation generates significant costs and risks for producers, who sell at low prices when they have sufficient capacity and are obliged to purchase it in the market otherwise.
Reducing the share of regulated demand served by incumbent producers and allowing wholesale prices to reflect costs and market conditions can help foster open electricity markets in the Western Balkans. Allowing a greater share of market procurement in the USS would force regulated prices to be partially responsive to market signals, as procurement costs would vary with market prices. The development of organised markets in the region in the past years helps provide reference prices and an avenue for procurement, although in practice, USSs would be expected to arrange for delivery through longer term bilateral contracts. Reducing the stringency of public service obligation provisions, on the other hand, would allow incumbent producers to behave as market actors, and be subject to market discipline, which can be beneficial in the need to improve corporate governance.
Direct support to fossil-fuelled electricity has limited macroeconomic impacts but distorts the energy mix
Direct transfers and tax expenditures to sustain coal production for electricity and gas supply have limited macroeconomic impact but distort the energy mix. Support mechanisms to sustain coal and gas, when simulated to represent their non-crisis levels, have limited macroeconomic impact. Most of the measures are found in simulations to have small negative impacts on output, driven by the use of fiscal resources. This is in contrasts with the measures applied to sustain the energy sector during the energy crisis, which contributed to sizeable increases in public deficits in the region. However, despite their relatively small size, direct support to coal alters the electricity production mix, increasing the share of coal-fired electricity compared to a simulated situation where these subsidies would be removed.
Subsidies to fossil fuels also reduce incentives to invest in renewable energy sources. As production prices appear lower than their real economic costs, trade-offs are slanted in the benefit of fossil fuels. Certain renewable technologies are already more cost-effective than coal-fired power plants, given the coal sector’s low productivity. Auctions for solar power parks in Albania have resulted in record-low winning prices of 24.90 EUR/MWh (for the Karasta Photovoltaic Park) and 28.89 EUR/MWh for the Spitalië Park (OECD, 2022[2]; EBRD, 2021[21]). In contrast, average costs of production from TPPs in the FBiH were 53 EUR/MWh prior to the energy crisis (and rose to 66 EUR/MWh in 2022). Average prices are far from the only consideration for further investment in energy sources, given the need to ensure sufficient baseload capacity and balancing reserves. Notably, the first solar power plan in Southeast Europe to be built on an old mine (Oslomej 1) is operated by Elektrani na Severna Makedonija (ESM), which prior to the energy crisis had received significantly fewer subsidies than most incumbents in recent years.
To enable the phase-out of coal subsidies in the Western Balkans, policies will need to consider the energy security and social concerns that justify them. The coal sector is an important employer in the region, generating an estimated 33 782 direct jobs in coal mining and 7 728 direct jobs in coal-fired TPPs, in addition to some 61 000 indirect jobs. Typically, TPPs are located near coal mines, which means closing or restructuring coal mines has significant impacts for the local economy. The availability of significant proven reserves of brown coal and lignite in Bosnia and Herzegovina, Kosovo and Serbia acts as an incentive to maintain the capacity to produce and use coal, despite certain operations not being economically profitable. In light of commitments to emissions reduction and to phasing out of fossil fuel use, the closure of most coal mining operations should be planned for appropriately – and may, in fact, be accelerated by subsidy phase-out. Progress has been made in planning for a just transition in some coal regions, particularly with support such as the Initiative for Coal Regions in Transition in the Western Balkans and Ukraine (EC, 2023[22]).
References
[31] Abril-Pla, O. et al. (2023), “PyMC: a modern, and comprehensive probabilistic programming framework in Python”, PeerJ Computer Science, Vol. 9, p. e1516, https://doi.org/10.7717/peerj-cs.1516.
[6] Aguiar, A. et al. (2022), “The Global Trade Analysis Project (GTAP) Data Base: Version 11”, Journal of Global Economic Analysis, Vol. 7/2, pp. 1-37, https://doi.org/10.21642/jgea.070201af.
[30] Bradbury, J. and et al. (2018), “JAX: composable transformations of Python+ NumPy programs”.
[8] Burfisher, M. (2020), Introduction to Computable General Equilibrium Models, Cambridge University Press, https://doi.org/10.1017/9781108780063.
[25] Center for Global Trade Analysis (2019), “Global Trade Analysis Project (GTAP) Database Version 10 (database)”, https://www.gtap.agecon.purdue.edu/databases/default.asp.
[7] Château, J., R. Dellink and E. Lanzi (2014), “An Overview of the OECD ENV-Linkages Model: Version 3”, OECD Environment Working Papers, No. 65, OECD Publishing, Paris, https://doi.org/10.1787/5jz2qck2b2vd-en.
[9] Corong, E. et al. (2017), “The Standard GTAP Model, version 7”, Journal of Global Economic Analysis, Vol. 2/1, pp. 1-119, https://doi.org/10.21642/jgea.020101af.
[24] Dixit, A. and E. Stiglitz (1977), “Monopolistic Competition and Optimum Product Diversity”, American Economic Review, Vol. 67/3, pp. 297-308, https://www.aeaweb.org/aer/top20/67.3.297-308.pdf.
[21] EBRD (2021), VISP: OSHEE COVID-19 Response, European Bank of Reconstruction and Development, London, https://www.ebrd.com/work-with-us/projects/psd/52360.html (accessed on 4 May 2023).
[22] EC (2023), “Coal regions in the Western Balkans and Ukraine”, Energy, Climate change, Environment, https://energy.ec.europa.eu/topics/carbon-management-and-fossil-fuels/coal-regions-western-balkans-and-ukraine_en.
[11] ERC (2024), Annual report 2023, Energy, Water Services and Municipal Waste Management Services Regulatory Commission of the Republic of North Macedonia, https://www.erc.org.mk/odluki/ANNUAL%20REPORT%20for%202023%20-ERC.pdf.
[23] European Commission (2008), NACE Rev 2 - Statistical classification of economic activities in the European Community, https://ec.europa.eu/eurostat/documents/3859598/5902521/KS-RA-07-015-EN.PDF.
[3] Eurostat (2025), “Energy intensity”, https://ec.europa.eu/eurostat/databrowser/view/nrg_ind_ei/default/table?lang=en (accessed on 30 April 2025).
[10] Eurostat (2008), NACE Rev. 2 Statistical classification of economic activities in the European Community, Eurostat, European Commission, Luxembourg., https://ec.europa.eu/eurostat/documents/3859598/5902521/KS-RA-07-015-EN.PDF.
[18] Hemmerlé, Y. et al. (2023), “Aiming better: Government support for households and firms during the energy crisis”, OECD Economic Policy Papers, No. 32, OECD Publishing, Paris, https://doi.org/10.1787/839e3ae1-en.
[32] Herce, C. et al. (2023), “A methodology to characterize energy consumption in small and medium-sized enterprises at national level in European countries”, Clean Technologies and Environmental Policy, Vol. 26/1, pp. 93-108, https://doi.org/10.1007/s10098-023-02606-z.
[33] LEAP4SME (2021), Mapping SMEs in Europe. data collection, analysis and methodologies for estaimating energy consumptions at country levles, LEAP4SME, https://leap4sme.eu/wp-content/uploads/2021/07/LEAP4SME-D2.1-SME-energy-and-economic-mapping-in-Europe.pdf.
[5] Lustig, N. (ed.) (2022), Commitment to Equity Handbook: Estimating the impact of fiscal policy on inequality and poverty. Second edition, https://commitmentoequity.org/wp-content/uploads/2023/04/CEQ-Handbook-Volume-1-.pdf.
[17] OECD (2023), “SME policy responses to the 2022/2023 energy crisis: Policy highlights and country experiences”, OECD SME and Entrepreneurship Papers, No. 43, OECD Publishing, Paris, https://doi.org/10.1787/80012fbd-en.
[1] OECD (2023), Tax Policy Reforms 2023: OECD and Selected Partner Economies, OECD Publishing, Paris, https://doi.org/10.1787/d8bc45d9-en.
[2] OECD (2022), Multi-dimensional Review of the Western Balkans: From Analysis to Action, OECD Development Pathways, OECD Publishing, Paris, https://doi.org/10.1787/8824c5db-en.
[4] OECD (n.d.), Green growth indicators (database), https://stats.oecd.org/Index.aspx?DataSetCode=GREEN_GROWTH (accessed on 31 May 2024).
[28] Pearson, K. (1997), “GTAP Book - GEMPACK Documentation”, Center for Global Trade Analysis, https://www.gtap.agecon.purdue.edu/resources/res_display.asp?RecordID=408.
[13] Petreski, M. (forthcoming), “The Distributional Impact of North Macedonia’s Taxes and Social Spending and of Government Electricity-related Measures during the Energy and Food Price Crisis of 2022-23”, OECD Developent Centre Working Papers, OECD.
[26] Robinson, S., A. Cattaneo and M. El-Said (2001), “Updating and Estimating a Social Accounting Matrix Using Cross Entropy Methods”, Economic Systems Research, Vol. 13/1, pp. 47-64, https://www.tandfonline.com/doi/pdf/10.1080/09535310120026247.
[27] Rosenthal, R. (2007), “A gams tutorial”, Vol. 5/26, https://www.gams.com/docs/pdf/Tutorial.PDF.
[20] Ruiz, P. and H. Mederac (eds.) (2021), Recent trends in coal and peat regions in Western Balkans and Ukraine, https://doi.org/10.2760/81752.
[12] SERC (2024), SERC Annual report 2023, State Energy Regulatory Commission of Bosnia and Herzegovina, https://www.derk.ba/en/godinji-izvjetaji-derk-a/godinji-izvjetaj-2023.
[16] Tambunan, T. (2013), Fossil-fuel subsidy reform and small and medium-sized enterprises (SMEs): the impacts and possible responses, GSI/ISSD.
[29] Theano Development Team (2016), “Theano: A Python framework for fast computation of mathematical expressions”, http://arxiv.org/abs/1605.02688.
[14] Vladisavljevic, M. and J. Zarkovic (forthcoming), Distributional impact of energy subsidies and energy subsidy reform in Serbia, background paper for this report.
[19] World Bank (2023), Testing Resilience, The World Bank, Washington, DC, https://openknowledge.worldbank.org/server/api/core/bitstreams/fde84e0e-1b8f-4889-a971-06ff64d1f1c6/content (accessed on 31 May 2024).
[15] Zhabjaku Shehaj, E. (forthcoming), The distributional impact of Albania’s fiscal policy and government electricity-related interventions during the 2021-23 energy and food price crisis, OECD Development Centre, Paris.
The POWER-CGE is a Computable General Equilibrium (CGE) model developed for the Western Balkans as part of this report to assess the impacts of different energy-related policies on the various sectors of the economy. The model consists of a set of 93 equations, 93 variables, and 128 parameters, which are integrated to represent the dynamics of national economies. For each economy, a specific version of the model is calibrated using its respective social accounting matrix (SAM); therefore, the corresponding equilibrium and values obtained reflect their economic realities.
Computable General Equilibrium (CGE) models are a class of economic model that capture the interactions among different sectors, agents, and markets within an economy. Rooted in general equilibrium theory, a CGE model describes how supply and demand in multiple interconnected markets adjust simultaneously to reach an equilibrium where all markets clear. These models are "computable" because they use numerical methods to solve the system of equations representing the behaviour of agents – such as households, firms, and the government – based on microeconomic principles and calibrated to observed data.
At the core of a CGE model lies a SAM, a dataset which provides a comprehensive snapshot of all economic transactions in an economy during a specific period. This includes the flows of goods and services, factor incomes (including labour, capital and energy), and transfers between institutions. The model incorporates behavioural functions that describe how agents optimise their decisions – households maximise utility, firms maximise profits, and the government follows policy rules – subject to constraints. These behaviours are governed by functional forms (e.g. CES production functions) and elasticities that determine how responsive agents are to changes in prices and income.
CGE models are widely used for policy analysis and forecasting, particularly in evaluating the impact of reforms such trade liberalisation, taxation, climate policies, and external shocks. Their main strength lies in their ability to capture the indirect and economy-wide effects of a policy intervention, including feedback loops across sectors and markets. Their flexibility and grounding in economic theory make them a valuable tool in both academic and applied policy settings.
For the development of the POWER-CGE model, the OECD ENV-Linkages model was used as a base, with three major modifications (Château, Dellink and Lanzi, 2014[7]):
Greenhouse gas emissions were not considered in the model.
The model is static: time dynamics are not modelled to focus exclusively on the general equilibrium effects.
The electricity supply sector is divided into three parallel grids: regulated grid for households, unregulated grid for non-households, and regulated grid for non-households.
The POWER-CGE uses stylised "grid" sectors to represent the electricity supply sector. The stylised “grid” sector aggregates outputs from all energy wholesale suppliers into a uniform product. This composite energy is then resold to producers of consumption goods, who, as a result, demand only a standardised energy product rather than differentiated forms of wholesale energy.
The POWER-CGE model is structured around four core sectors: energy production, industrial production, households, and government. The supply side consists of energy and industrial production, while all sectors are interlinked through production, demand, and market-clearing equations that the model solves to determine equilibrium. The model also incorporates an international component for the supply side, capturing exports and imports with exogenously determined world prices under the small open economy assumption. A diagram illustrates the interactions among these sectors (Annex Figure 4.A.1).
Source: Authors’ elaboration.
The industrial production and energy production processes are modelled using three factors of production: capital, labour, and electricity. Capital and labour are supplied by households, while the final electricity good is supplied by an energy grid. Firms combine these three factors with raw materials or intermediate goods to produce the goods they supply.
Industrial production
POWER-CGE divides industrial production into 26 sectors. These are based on the 23 sections defined in NACE Rev. 2 (European Commission, 2008[23]), with further disaggregation of the mining sector into: (1) coal, oil, and gas extraction; (2) extraction of other products; and (3) support services. This detailed breakdown reflects the model’s focus on energy-related policy analysis. Annex Table 4.A.1 presents the details of the sectors used by the model. The production process follows a standard CGE setup (Burfisher, 2020[8]) with a CES production function for the three factors of production.
The production functions use a traditional CGE set-up, where each firm produces a final good that can be used as a consumption and investment good. The final goods are produced with a Leontief production function that reflects fixed-proportions technology using two inputs: a “value-chain” component () and a “value add” component (). The Leontief production function has the form:
Where represents recombination of final goods from each of the 26 industrial firms used as intermediary inputs, and is modelled by the following equation
Where J corresponds to the set of all the 26 sectors, and corresponds to a weighting parameter.
The “value-add” production process combines three-factor inputs - capital, labour and electricity - subjected to a nested CES production function of the form:
Where corresponds to a combination of electricity and capital.
Energy
The energy sector in POWER-CGE comprises two components: energy producers and grid operators that aggregate and sell energy. Introducing a distinct grid component is a key innovation of the model, enabling analysis of policy impacts within complex regulatory environments – particularly relevant for electricity markets, where specific segments are often subject to targeted regulations.
Each energy-producing firm in POWER-CGE generates a specific type of raw energy. In line with the Social Accounting Matrix (SAM) structure based on NACE Rev. 2 (European Commission, 2008[23]), energy production is categorised into seven distinct sources, along with a separate category for energy distribution. The energy sources include: coal, gas, hydro, nuclear, oil, solar, wind, and other. For each energy type, an energy firm is modelled, which is subject to a traditional optimisation problem given by demand and a production function.
The output of each energy-producing firm is not sold directly to other sectors of the economy. Instead, it is routed through three intermediary grid operators, which bundle and distribute the energy under distinct regulatory frameworks. These grid firms serve as commercial intermediaries, enabling the model to reflect the regulatory diversity typically found in real-world energy markets. The aggregation process of the grid follows the aggregation method first introduced by Dixit and Stiglitz (1977[24]). These intermediary grids do not make any profit and act merely as aggregators that connect the demand with the supply of the different energy firms. To represent the complex regulatory environments, three different grids are used:
Regulated grid that provides electricity to households.
Regulated grid that provides electricity to firms and the government.
Unregulated grid that provides electricity to firms and the government.
Grid pricing depends on the regulatory regime. For regulated grids, prices are set exogenously, and both grid operators and energy producers must meet all demand at the given price. For unregulated grids, prices and quantities are determined endogenously through market equilibrium.
Energy supply to industrial sectors draws on both regulated and unregulated grids, following economy-specific patterns. Following real-life regulatory patterns, each sector of industrial production is served by both regulated and unregulated energy grids. The percentage of energy obtained from each kind of grid for each individual sector is defined using economy-specific data.
Energy production firms also employ a Leontief production function composed by a value-chain component () and a value-add component (). The production function has the form:
Where the value chain has the functional form of a Leontief function:
and the value-add component has the form of a constant elasticity of substitution (CES) function that bundles a combination of capital and labour (), and raw material (), with the functional form:
Finally, electricity grids use the following Dixit-Stiglitz energy bundling equations where the electricity output is given by:
This functional form reflects the fact that electricity produced by different sources, both domestic and foreign, are imperfect substitutes and assumes that the elasticity of substitution is identical between all varieties.
Like most CGE models, POWER-CGE is constructed and calibrated using economy-specific SAMs. These SAMs have a data structure that automatically encodes the equilibrium assumptions of the CGE model. A SAM is an n x n matrix, written in a double-entry bookkeeping style. All row accounts (expenses of agents) add up to all column accounts (incomes of agents). This ensures that all market-clearing conditions will be automatically satisfied with no residuals.
The SAMs for all the economies are obtained primarily from the GTAP dataset (Center for Global Trade Analysis, 2019[25]). However, in the cases of Bosnia and Herzegovina, Kosovo, Montenegro, and North Macedonia, the data is only available aggregated in the category “rest of Europe.” For these cases, to obtain the data, the following process is applied:
To obtain a base SAM, the Rest of Europe SAM provided by GTAP is split proportionally to the economies’ GDP.
This base SAM is adjusted using updated information from National Accounts published by Central Banks and National Authorities.
The new SAM is rebalanced to guarantee the CGE equilibrium conditions.
The rebalancing of the new SAM was done using a cross-entropy (CE) method. The CE method is an iterative technique commonly used for rebalancing or updating SAMs when new, often incomplete or inconsistent, data become available. The CE method offers a statistically grounded approach to do this by minimising the divergence between the updated matrix and a prior or “base” SAM, subject to newly imposed constraints derived from fresh data (Robinson, Cattaneo and El-Said, 2001[26]).
At the heart of the CE method lies the Kullback-Leibler divergence, a concept from information theory that quantifies the distance between two probability distributions. In the SAM context, the method minimises the relative entropy between the original and the adjusted matrix, meaning it updates the matrix entries in a way that remains as close as possible to the base data while strictly respecting the new macroeconomic aggregates or microdata constraints. The objective function is set up as a sum over all matrix entries of the product of the adjusted value and the logarithm of the ratio between the adjusted and base values. Subject to linear constraints (e.g. new row and column totals), this optimisation problem is typically solved using convex programming techniques.
Most existing CGE models are built with the use of pre-existing software packages like GAMS (Rosenthal, 2007[27]) or GEMPACK (Pearson, 1997[28]). One of the main innovations of the POWER-CGE model is that it was developed alongisde a new Python package, cge_modeling for defining and solving CGE models. The new cge_modeling is developed on top of the pytensor library. pytensor, a fork and continuation of the theano project (Theano Development Team, 2016[29]), a library for creating and optimising static computational graphs. In POWER-CGE, the CGE model is represented as a computational graph and compiled into optimised JAX code (Bradbury and et al., 2018[30]). While this approach forgoes the specialised solvers available in GAMS, it enables GPU acceleration, automatic algebraic simplification, memory optimisation, and seamless integration with the broader Python scientific ecosystem. Crucially, implementing cge_modelling in pytensor allows the user to embed CGE models within probabilistic frameworks using PyMC (Abril-Pla et al., 2023[31]). This integration makes it possible to incorporate parameter uncertainty into policy experiments, impute missing data, and share information across economies to improve parameter estimation.
← 1. The estimated value of induced transfers through price interventions in Chapter 3 also include sales at below-market prices to network operators to compensate for losses. Those are excluded from this calculation.
← 2. An alternative lower bound could be constructed on the basis of supply costs, including full cost recovery values for the share of electricity supplied by producers. Such data are not readily available across the region. Indeed, in the case of vertically integrated firms, the price of inputs to electricity production (in particular coal) is also subject to price interventions.
← 3. The distributional incidence is estimated on the basis of SILC and HBS data combined. Since the data does not include details on consumption volumes and patterns, this calculation does not account for block tariffs.
← 4. The absolute poverty rate is estimated to be 11.6% when accounting for direct and indirect taxes and transfers, corresponding to the measure of consumable income in the Commitment to Equity (CEQ) framework (Lustig, 2022[5]).
← 5. In North Macedonia, 65% of respondents to the household budget survey (HBS) report zero electricity consumption, due to the short recall period not capturing the billing frequency for electricity. If this problem is ignored, 25.1% of households are found to be energy poor. However, if households reporting zero consumption 84.7% of households are found to spend more than 10% of their income on energy.
← 6. The reference budgets for FSA and SEVC do not necessarily correspond to official figures. The FSA budget is estimated consistently with microdata from SILC, and is therefore likely to be underestimated. The SEVC budget is simulated according to programme rules and may be overestimated if take-up is less than perfect.
← 7. Energy intensity of SMEs depends greatly on the overall sector composition of SMEs. Generally, despite constituting the overwhelming majority of firms, SMEs account for a small share of total energy consumption (ranging from 9% in Portugal to 27% to 29% in Italy) (Herce et al., 2023[32]; LEAP4SME, 2021[33]).