Making Power Affordable for Africa and Viable for Its Utilities

Examination of the financial viability of power sectors in 39 countries in Sub-Saharan Africa shows that only two countries have a financially viable power sector, and only 19 cover operating expenditures. Quasi-fiscal deficits average 1.5 percent of gross domestic product. If operational inefficiencies can be eliminated, power sectors in 13 countries become financially viable. In the remaining two-thirds of the countries, tariffs will likely have to be increased even after attaining benchmark operational efficiency. Analysis of power tariffs in another 39 African countries shows that about half of them have small first blocks with low lifeline rates. Data from national household expenditure surveys in 22 African countries show that the subsistence level of grid electricity is affordable to the vast majority of the population in many countries with low rates of access. However, benefits of progressive tariffs are compromised by the widespread practice of multiple connections, prompted by high costs of grid connection. Examination of the sex of the head of household shows that female-headed households are not disadvantaged in electricity use once income and the place of residence (urban or rural) are taken into account. However, female-headed households tend to be poorer, making it all the more important to focus on helping the poor. Installing individual meters and cross-subsidizing installation, encouraging prepaid metering, and reformulating lifeline blocks and rates as ap


Masami Kojima and Chris Trimble
Africa Renewable Energy and Access Program (AFREA)  Many households are unable to afford connection fees and tariffs, thus limiting expansion of access to grid electricity. Some households share electricity meters to offset connection fees, which can exceed monthly income many times over. In Ethiopia, where the lowest connection charge represents 130 percent of average monthly household income, there are two and a half times as many grid-connected households as utility customers. Under current tariff structures, subsistence consumption of 30 kilowatt hours a month is affordable to the vast majority of the population in most countries, including those with low rates of access to electricity. But shared meters deprive low-income electricity users of the benefits of lifeline tariffs, because the combined consumption of multiple households places them in higher-priced tariff brackets.

Polic y considerations
Improving operational efficiency should be the first area of policy focus in reducing quasi-fiscal deficits. If utilities could reduce combined transmission, distribution, and bill collection losses to 10 percent of dispatched electricity (the level considered in this study for benchmark utility efficiency) and tackle overstaffing, an additional 11 countries could see their utility deficits disappear, bringing the total number of deficit-free countries to 13. International experience suggests that unmetered consumption is disproportionately concentrated in large consumers and others who are able to pay cost-reflective tariffs. By targeting better-off, large-volume customers first, significant loss reduction is possible with little loss of welfare.
Most countries may need to increase tariffs. In two-thirds of the countries studied, the funding gap cannot be bridged simply by eliminating operational inefficiencies; tariffs will have to be increased even after achieving benchmark operational efficiency. The pressure to increase tariffs could be eased somewhat by optimizing the power generation mix and reducing costs further, but doing so could take many years and require substantial investment.
Political economy considerations can inform the design of tariff increases: ⦁ Raising tariffs while outages continue unabated is bound to invite a backlash. Utilities need to focus on achieving an acceptable level of service quality to launch a trajectory toward cost recovery in tariff revenues. They could reduce costs by phasing out operational inefficiencies, implementing short-term measures to reduce the duration (if not the frequency) of outages, and addressing customer service quality in general. Information systems are fundamental to attending to customer complaints, accelerating service restoration after outages, regularly measuring system reliability, and providing better commercial service.
⦁ Small, frequent tariff increases may find wider acceptance than infrequent large increases. To eliminate fuel subsidies, India and Thailand raised fuel prices by small fixed amounts every month, announced in advance, until cost-recovery levels were reached.
Predictability coupled with small and manageable increases can go a long way toward achieving acceptance.
ix E x ECU T IVE SU MM A Ry ⦁ A period of low oil prices may be a good time to introduce an automatic fuel price passthrough mechanism. An automatic price adjustment mechanism, such as that in place in Kenya, can depoliticize one element of tariff adjustments. The best time to introduce such a mechanism is when input prices are low so consumers do not equate the mechanism with large tariff increases.
⦁ Targeting tariff increases to customers who account for the bulk of consumption and can afford to pay more would limit adverse effects on the poor. As with commercial losses, it would make sense to focus tariff increases first on large-and medium-size customers, for whom affordability is not as significant a challenge as for small-consumption households. Although the political sensitivity of tariff increases to better-off consumers cannot be ignored, neither should it be overemphasized. In the face of large utility deficits and low access rates, there is no compelling reason to subsidize those who can afford higher tariffs. Indeed, they could be asked to cross-subsidize low-income consumers more, as long as the latter's total consumption is a small fraction of the total electricity

sold. Successful examples of power sector reforms in emerging countries in other regions
show that middle-and high-income consumers in all tariff categories usually accept cost-reflective rates, provided the quality of electricity services is good.

Prepaid meters can help both utilities and customers.
For low-income households with cash flow constraints, the ability to pay in small increments helps align electricity payments with income flows. Households on prepaid plans do not risk disconnection for failure to pay and avoid reconnection fees, which can be considerable in some countries. But prepaid meters should not be made mandatory if grid electricity is unreliable, lest customers pay cash in advance for electricity they cannot get when they need it. For the utility, prepaid meters improve revenue collection.
The additional subsidies needed over and above current lifeline rates to enable the poor to purchase the subsistence level of electricity are modest. If households can be metered individually and accurately, the additional subsidies needed to make the subsistence level of grid electricity affordable to every urban household in the countries studied would be less than $5 million in 19 countries and less than $1 million in 15 countries. Subsidies needed to make electricity affordable to every rural household will be greater; this topic is outside the scope of this study, because rural electrification will play a considerable role.

Sharper targeting of cross-subsidies for the poor can help increase affordability and speed
up access expansion. The first priority in increasing access to electricity is to make the initial connection affordable. Consideration may be given to including assets associated with new connections in utilities' regulatory assets and recovering costs from all customers. A small first block with a lifeline rate can be introduced if the utility has not already done so, its size can be optimized, cross-subsidies can be increased, and a moving average rather than monthly consumption can be used as the basis for the lifeline rate.

Old Problem, New Context
M ore than 1 billion people globally still lack access to electricity. For many more, their electricity is unreliable, with blackouts forcing them to turn to expensive self-generation, suffer business losses, make do with inferior lighting from kerosene-or spend hours in darkness.

Access in Africa continues to lag rest of the world
Africa 1 lags all other regions of the world in installed generation capacity, per capita electricity consumption, and household access to electricity. The total electricity generation capacity in the region, which has a total population of nearly 1 billion, is less than 100 gigawatts-less than the total generation capacity in Spain with its population of 46 million-and halved if South Africa's access lags at the same level of development. Infrastructure development mirrors economic development. Generally, the higher the gross domestic product (GDP) per capita, the better the roads, power system, railways, and telecoms. But when access to electricity is examined at comparable levels on economic and poverty indicators, Africa significantly lags the other regions. Figure 1 illustrates this observation. Countries broadly fall along a downward-sloping line, except African countries, many of which lie far below the trend line. As to the poverty gap itself, the situation in many African countries in 2012 closely resembled that of the rest of the world in 1990. When the poverty gap and access data from 2012 in Africa are overlaid with those from 1990 in other regions, Africa still falls below the global trend   Access to electricity, % of population Poverty gap in % at $3.10/person/day 500 5,000 50,000 Per capita GDP in $ at purchasing power parity

1b. Comparison of access in Africa in 2012 with access in other regions in 1990
Africa has been increasing at a mere 5 percentage points every decade-rising from 23 percent in 1990 to 35 percent in 2012. Yet another consequence is that the Doing Business indicator for supply reliability and tariff transparency is, on a scale of 0 to 8, only 0.9 in Africa-by far the lowest in the world and less than half the second lowest, 1.9 in South Asia (World Bank 2016b).
Past data indicate full cost recovery unlikely. These problems were documented in the Africa Infrastructure Country Diagnostic (AICD), a wide-ranging knowledge program that undertook extensive data collection in the 2000s (primarily from 2001 to 2008) in all key infrastructure sectors, including grid electricity. The AICD's analysis of the electricity sector pointed to inadequate generation capacity, stagnant and inequitable access to unreliable electricity services, prevalence of backup generators, increasing use of leased emergency power generation capacity, and concurrently high and inadequate electricity tariffs. External events-drought, oil price shocks, and civil conflict-exacerbated an already precarious situation. The AICD estimated that a total annual investment of $40 billion would be needed in the sector across the region, more than triple the $12 billion a year being spent at the time. Although gross operational inefficiencies in poorly functioning utilities offered scope for cost savings, the funding gap could not be bridged entirely by eliminating the estimated $8 billion a year of inefficiencies (Eberhard et al. 2011). Combined examination of tariff structures in 27 countries and of household survey data found that recovery of historic electricity production costs, tariffs consistent with future electricity production costs, affordability of tariffs to low-income households, and distributional equity of tariff subsidies would be almost impossible to achieve in the context of the high-cost, low-income environment characterizing much of the region in the mid-2000s (Briceño-Garmendia and Shkaratan 2011).

Changed contex t sharpens the challenge for access
The context for electricity access and utilities' financial performance has changed in important ways since the AICD program was under way.
Global commitment to universal access has strengthened. In 2011, the United Nations launched the Sustainable Energy for All initiative, bringing diverse partners-including 106 governments and the European Union-together to achieve three dimensions of sustainable modern energy by 2030, including universal access to electricity. Africa is likely to be the last region of the world to reach this target, and will benefit from any increase in external assistance to that end. Fresh look at access issues, focusing on full cost recover y and t arif f af fordabilit y Some African governments have attempted to expand access to electricity by subsidizing its use, the cost of connection to grid electricity supply, or both. However, in the face of a falling fiscal balance, there is little room for large-scale government support. This restriction underscores the need to improve utility performance, adjust tariffs to achieve cost recovery, and design subsidy support in a way that maximizes the benefits to the neediest.
The present study draws on more recent data than was available to the AICD to reexamine issues in the African electricity sector. The study looks at the literature on power sector subsidies and their reform with an emphasis on the political economy of such reform, asks about the degree of cost recovery with and without operational inefficiencies, assesses residential tariff structures, estimates the affordability to households of the subsistence level of electricity consumption, and computes how much additional assistance might be needed to enable the poor to use electricity. The study's coverage is largely confined to grid electricity, and does not capture the cost of system expansion or potential cost savings from system optimization based on a least-cost development plan and expansion in cross-border electricity trade.
The next section covers issues related to the electricity sector's financial viability; following that is a discussion of affordability issues from the consumer point of view based on household survey data. A brief discussion of reliability is included, because unreliable electricity harms rather than contributes to economic development, and access is increasingly defined as access to reliable electricity. The report's final section summarizes the main findings and policy recommendations emerging from the study. Two appendixes present the sources, methodology, and assumptions underlying the study. For full details, see Trimble et al. (2016); Kojima et al. (2016); Kojima, Bacon, and Trimble (2014); and companion Excel files (www.worldbank.org/ affordableviablepowerforafrica).

Sector Financial Viability
S imply put, a utility that does not cover its costs will struggle to deliver reliable electricity in sufficient quantity. To determine the financial viability of Africa's electric utilities, this section explores these questions: ⦁ How much of the total cost of electricity supply does cash collected by the utilities cover?
⦁ How much can revenue shortfalls be reduced if operational efficiencies are increased? What are the priority areas in individual countries for reducing operational inefficiencies?
⦁ How do fluctuations in rainfall and fuel prices affect electric utilities' financial viability?
Broadly, there are two approaches to interpreting cost recovery and financial viability: cash needs and full cost recovery (Kojima, Bacon, and Trimble 2014 However, input and other types of subsidies covering unbilled expenditures may not be captured.
The main difference between the two approaches is that full cost recovery includes coverage of future investment projects for significant replacement or upgrading of existing capacity as well as capacity expansion. Essentially, then, the cash needs approach incorporates only known present cash considerations, while full cost recovery includes provisions for future capital costs and adequate returns on investment.
The question of which costs should be recovered through revenue collection influences tariff setting. International experience suggests that the degree of cost recovery from consumer payments depends largely on the level of development of the electricity sector. The electricity sector in Africa is far from mature. The medium-to long-term target is to reach tariff levels that fully cover all reasonably and prudently incurred costs. If, revenues fall far short, the first priority is to meet all cash obligations. The cash needs approach reduces the short-term impact of capital expenditure on tariffs, requiring the utility to seek concessional financing (grants, low-interest loans, long grace periods, partial risk guarantees) for major capital expenditures.
The reality is that available concessional financing to meet future demand for electricity is far from sufficient. The hope is that future economic growth will increase consumers' ability to pay and eventually enable full cost recovery.
This study makes use of both approaches but, due to data limitations, with some modification.
In using the first approach, this study excludes debt obligations from cash needs and refers to total cash requirements as operating expenditures. In the second approach, capital expenditures are confined to new replacement values of current assets, amortized over the economic life of each asset; capacity expansion is excluded (appendix A). The remainder of this report concerns recovery of operational and capital expenditures as understood in this modified sense.
Using utility financial data available in 39 African countries, the following examines in detail the extent of cost recovery and the scope for increasing cost recovery by improving operational

Current cost recover y
To what extent are costs covered by cash collection? Figure 2 compares the cash collected from bills sent out with the total costs of supply-broken down into operational and capital expenditures-per kilowatt hour (kWh) billed. The findings show that only the Seychelles and Uganda fully covered both operational and capital expenditures. Cash collected in 19 countries covered operational expenditures, leaving 20 with insufficient cash to cover these. Among the top 10 countries with highest unit costs, all but Guinea and Sierra Leone are highly dependent on oil-based generation, which is expensive. Logarithms of unit cost and unit cash collected are statistically significantly correlated, with a correlation coefficient of 0.91, suggesting that utilities with high unit costs tend to have high tariffs and collect a higher share of revenue billed. A study for the AICD by Briceño-Garmendia and Shkaratan (2011) using a smaller set of countries similarly found that no country was covering its total costs of supply.

What is the difference between collections and revenue?
The quasi-fiscal deficit is the difference between the net revenue of an efficient electricity sector covering operational and capital costs and the net cash collected by the utilities. The results of this computation are shown in table 1. South Africa had the largest quasi-fiscal deficit in absolute terms, $11 billion.
As a proportion of GDP, São Tomé and Príncipe had the highest deficit at 6.1 percent, followed by The Gambia at 5.8 percent and Zimbabwe at 5.2 percent. As a proportion of cash collected, Ethiopia had the largest deficit (408 percent), followed by Comoros (397 percent) and Guinea (354 percent). At the opposite end of the spectrum, the Seychelles and Uganda had no deficit.
The quasi-fiscal deficit as a percentage of cash collected is positively correlated with capital cost per kWh billed using a 1 percent significance test (which ensures that the probability of mistakenly concluding that there is a statistically significant non-zero correlation-when there actually is not-is 1 percent or smaller); this suggests that the higher the capital cost, the higher the deficit relative to cash collected. Data available from 16 countries enabled estimation of how quasi-fiscal deficits were distributed among different customer classes using simplifying assumptions. Each consumer category's share of the quasi-fiscal deficits broadly tracks kWh of electricity consumed, except for households and industry. Households consumed 38 percent of the total volume of electricity billed but accounted for 42 percent of the deficits, whereas industry consumed 33 percent of electricity billed but accounted for 26 percent of the deficits; this suggests marginal cross-subsidization of households by industry.

Cost recover y at benchmark per formance
To understand what benefits can be achieved through improvements in operational performance, quasi-fiscal deficits were also estimated at efficient operation. Efficient operation-also referred to as benchmark performance-is defined here as follows: ⦁ Transmission and distribution losses (both technical and commercial) of 10 percent of dispatched electricity or lower ⦁ 100 percent bill collection ⦁ The same staffing level as in well-performing, comparable utilities in Latin America Conversely, inefficient utilities are said to suffer from transmission and distribution losses, bill collection losses, and overstaffing. The deficit at benchmark performance is identical to the magnitude of underpricing in the absence of generation mix optimization and other cost-reduction measures such as greater imports of cheaper electricity (appendix A). Figure 3 shows how much quasi-fiscal deficits can be reduced by achieving benchmark performance. The results are shown as percentages of GDP in the reference year. An additional 11 countries, besides the Seychelles and Uganda, eliminate quasi-fiscal deficits altogether, bringing the total number of countries at full cost recovery to 13. It does not, however, automatically follow that these countries can lower tariffs upon achieving benchmark performance, because losses and overstaffing cannot be eliminated at no additional cost.
What drives each country's quasi-fiscal deficit? Large differences in the quasi-fiscal deficit are found between current and benchmark performance. To understand the drivers of the quasi-fiscal deficit in each country, the deficit is decomposed into four components or hidden   Note: Lesotho, Nigeria, and Sudan lacked data for staff-level analysis.
bill collection rates in excess of 95 percent and 10 had collection rates lower than 80 percent; Comoros, with a 58 percent bill collection rate, ranked lowest.

Is there a trade-off between quasi-fiscal deficits and access?
Would countries with ambitious electrification programs run large deficits? Do poorer countries tend to have high deficits because they lack funds, or do they tend to have low deficits because they cannot afford high quasi-fiscal deficits in the first place? To probe these questions, six measures of quasi-fiscal deficits-deficits at current and benchmark performance as percentages of GDP, revenue, and cash collected-were examined with respect to their relationships to GDP per capita, access rates, and the poverty gap. 1 None of the latter variables was statistically correlated with any measure of quasi-fiscal deficits using a 5 percent significance test. There is thus little indication of a trade-off between quasi-fiscal deficits and access, or a relationship between the level of economic development or the depth of poverty and deficits.

Impac t of oil price and drought
Recent trends in oil prices and rainfall have the potential to improve or further harm utilities' financial viability. The study looked at the potential effects of these phenomena on quasi-fiscal deficits. The effect of oil price on quasi-fiscal deficits depends on the price of oil in the reference year and how much electricity generation is sourced from oil. What would have been the magnitude of savings possible if the collapse of the world oil price since the end of 2014 had occurred in the reference year of analysis? As to rainfall, not every country has hydropower in its power mix. The effect of drought is a function of the share of electricity generated from hydropower, and how the loss will be replaced. For simplicity, this study assumes that the electricity loss from drought would be replaced with emergency leased power generation capacity at a cost of $0.20 per kWh. This assumption is similar to conditions in southern Africa in 2015 and 2016, where drought forced Zambia and Zimbabwe to resort to expensive short-term leased electricity generation capacity to make up for loss of hydropower. How much more cash will utilities need if increasing climate variability leads to more drought?
The results in countries where respective effects of drought and the oil price collapse are more than 10 percent of the current quasi-fiscal deficits are shown in figure 6. The average petroleum product prices in 2015 and a 30 percent reduction in hydropower generation were selected for this illustration and applied to each country's reference year. All utilities were assumed to have been paying for diesel and fuel oil at prices linked to those on the world market. Had any utility been benefiting from fuel price subsidies, the savings accruing to it would be smaller. Note: The bars shown are the quasi-fiscal deficit at current performance, the deficit re-evaluated at the 2015 oil price, and the deficit re-evaluated assuming a 30 percent reduction in hydropower generation, all in each country's reference year. Where both oil and hydropower effects are considered, the impact of oil is evaluated first, and the effect of lower hydropower generation is added on next. The difference in quasi-fiscal deficits as a percentage of GDP between hydropower and oil is the isolated effect of hydropower reduction alone on the deficit. Data on oil-based generation in Madagascar were not available. more than doubled if 30 percent of hydropower is replaced with leased electricity generation in the Central African Republic, Ghana, Lesotho, Mozambique, Uganda, and Zambia.

Overall trends in quasi-f iscal def icit s
Data for three or more years are available in 23 countries. In general, quasi-fiscal deficits follow consistent patterns over time in individual countries (improving, worsening, or remaining steady). Most of the countries with low quasi-fiscal deficits see them decline or stabilize over time; most of those with high quasi-fiscal deficits see them remain steadily high. Tariff increases combined with performance improvement helped reduce quasi-fiscal deficits in the Seychelles, Tanzania, and Uganda. Comparison across countries does not suggest declining quasi-fiscal deficits overall in the region.
In summary, only two countries were covering all costs in the electricity sector in the reference year of the analysis, and many were not even able to cover operational expenditure. Absent improvement in financial performance, the utilities cannot afford to maintain or introduce subsidies to make electricity affordable to the poor unless they significantly restructure subsidies for much sharper targeting, increase cross-subsidization for the poor, increase quasi-fiscal deficits, or implement some combination of these measures. Against the backdrop of rising demand, loss reduction alone is unlikely to eliminate outages in many countries, but a lack of generation capacity calls for investment in new capacity; capital spending would be difficult for utilities running such large deficits. In particular, the high risks associated with the financial activities of these utilities would result in a very high cost of capital. Outages caused by the inability to pay for fuels-as in Ghana and Nigeria for natural gas-would be similarly difficult to resolve as long as cash collected is not sufficient to cover fuel bill payments.

Household Use of Electricity
H ousehold use of electricity in Africa mirrors low consumption of electricity in the economy.
Access rates are low for a variety of reasons-households find the initial connection costs, if not monthly bill payments, too expensive; poor reliability makes grid electricity unattractive; and many live in areas where there is no grid electricity. To overcome the challenge of affordability, more subsidies may be needed, raising in turn the question of the affordability of such subsidies to utilities. To quantify these issues, this section addresses the following: ⦁ How is access to electricity affected by income and location of residence? ⦁ How affordable are grid electricity tariffs to households? Are affordability and access strongly correlated? Are there factors that make affordability calculations based on tariff schedules misleading? What does it take to make grid electricity affordable to all households?
⦁ How do the observed spending patterns compare to monthly bills for subsistence consumption of electricity?
⦁ Is there evidence that female-headed households are more or less likely to use electricity than male-headed households?
To answer these questions, this study uses tariff schedules in effect as of July 2014 in 39 countries (33 of which were examined in the quasi-fiscal deficit analysis) and national household expenditure surveys undertaken since 2008 in 22 countries. Data sources and methodology are discussed in appendix B.

Access and af fordabilit y
Survey questionnaires vary from country to country, as does the level of detail available from which to draw conclusions about the type of electricity (grid-connected, self-generation, solar panels) used by surveyed households. Rates of access by location, income quintile, and poverty status are shown in table 2, using the most expansive definition of access (level 5; see appendix B). Because income data are lacking, total household expenditure, which is used to calculate official poverty statistics, serves as a proxy for income in this study. The findings about access are consistent with widely accepted observations: the percentage of the population using electricity remains low in Africa, electricity use is far more prevalent in urban than rural areas, and adoption of electricity and quantities consumed rise sharply with income.
How much would it cost to make electricity affordable to all people? The answer to this question depends on how much electricity and what is considered affordable. This study bases affordability of electricity on having to spend no more than 5 percent of household income on 30 kWh a month, which is considered the subsistence level (box 1). Table 3 shows the results of calculations estimating the amount of subsidies required to enable every household in the country to spend no more than 5 percent of income to purchase 30 kWh a month. These hypothetical subsidies assume a perfect world in which precise targeting of subsidies individually tailored to each household is possible. The countries are listed in order of decreasing grid-electricity poverty gap. Source: World Bank staff analysis of household surveys.
Note: Bottom and top 20 percent refer to the bottom and top income quintiles. Access to electricity as defined here includes those reporting grid connection; those using electricity from the grid, generators, solar panels, or solar energy as their primary energy source for lighting or cooking; those owning generators or solar panels; and those reporting non-zero expenditures on electricity.

Box 1 Subsistence consumption and grid-electricity poverty
The framework developed by the Sustainable Energy for All initiative to define and measure access to energy considers 30 kWh a month to be the subsistence level for grid electricity; the framework considers the electricity affordable if a household does not have to spend any more than 5 percent of its total monthly income to purchase it (World Bank and IEA 2015). Accordingly, this study defines the grid-electricity poor as those who live in households that find monthly consumption of 30 kWh unaffordable. For example, if it costs $5 to buy 30 kWh, then the minimal monthly household income needed to afford electricity is $100, and all people living in households with a combined income of less than $100 are classified as electricity poor.
The grid-electricity poverty headcount is the proportion of a country's population that is grid-electricity poor. The headcount tells who is poor but not how poor they are. To determine depth of poverty, the grid-electricity poverty gap is calculated: it measures the extent to which 5 percent of household income falls below the monthly bill for electricity as a proportion of that bill. The sum of the grid-electricity poverty gaps across all households (multiplied by the monthly bill for 30 kWh) is the subsidy needed to enable everyone to consume the subsistence level of electricity, if subsidy transfers could be perfectly targeted.
So, for example, if every household in a country earns $60 a month, a $5 monthly expenditure on electricity would not be considered affordable (5 percent of $60 is $3). Everyone is electricity poor, and the poverty headcount is 100 percent. But the poverty gap is 40 percent, obtained by dividing $2 (the difference between $5 and $3) by $5, the monthly bill for 30 kWh. The poverty gap would be 100 percent only if every person had no household income. The hypothetical subsidy needed to make electricity affordable to the entire population would be $2 per household, multiplied by the total number of households in the country.
Predictably, the ranking of the grid-electricity poverty gap and the grid-electricity poverty headcount are similar. Access tends to increase with decreasing grid-electricity poverty metrics, but the relationship is not strong; it is weak at low levels of the grid-electricity poverty gap and headcount. Taking the bottom 10 countries in the table as an example, monthly bills for 30 kWh are affordable for more than 97 percent of all people, but some countries-Ethiopia, Malawi, Mali, and Tanzania-have low rates of access, ranging from 22 percent down to 9 percent of the population.
The additional subsidies needed to make the subsistence level of electricity affordable to every household in the country are surprisingly small at the current applicable tariffs. They are no more than 6 percent of utilities' cash collections. In practice, current tariffs are almost certain to be too low to extend grid electricity to every rural household. The costs per kWh delivered would be much higher than in the current system, and could also be higher than minigrids and other off-grid arrangements. In Africa, where 63 percent of the total population is estimated to be rural, the customer base needed to cross-subsidize rural households is not sufficiently large.
For these reasons, grid-electricity poverty indicators and subsidy estimations based on current tariffs are less likely to be applicable to rural areas, many of which are better suited for off-grid electrification (although the level of service is much lower in off-grid systems than in grid networks). Once the fact that many households will be served by off-grid electrification is taken into account, the subsidy requirements for grid electrification would decline further, although the subsidies needed for off-grid electrification can be substantial. A lower bound on subsidies is that needed to make grid electricity affordable to every urban household. The additional annual subsidies to that end would be less than $1 million in 15 countries and less than $5 million in 19. Even the maximum is less than 1 percent of the cash collected by utilities.
In the top 10 countries in   Note: Population weights are used for grid-electricity poverty metrics and access, and household weights for the remaining variables. Subsidies are the amounts needed to make up the difference between the monthly bill for 30 kWh and 5 percent of total household income for every household where the difference is positive. Where there are several tariff schedules applicable to 30 kWh, the lowest possible monthly bill-inclusive of all taxes and applicable charges-is selected. Access is the rate of household access to grid electricity or, where information on grid connection is not available, to electricity excluding batteries, solar energy, and standby generators (if separately accounted for in the survey). HH = household; QFD = quasi-fiscal deficit; -= not available.
a. Annual subsidy required in current $ million in the survey year, or if the tariffs came into effect later, the first year of tariff implementation. customers more and cross-subsidize needy households. All other countries will continue to experience underpricing even after attaining benchmark performance in operational efficiency.
The current quasi-fiscal deficits are not small in the top three countries: 2.2 percent of GDP in Madagascar, and 1.0 percent in Burkina Faso and Rwanda. These findings highlight the challenges facing countries with high grid-electricity poverty gaps and low access rates.
What does it cost to be connected to the grid? An equally important consideration in affordability is the cost charged to households for initial connection to the grid. Figure 7   computations give an idea of how many people can afford to buy varying amounts of electricity and the subsidies needed to make these amounts affordable to all households if perfect targeting is possible.
Which measure of grid-electricity poverty, if any, is best correlated with access? While grid-electricity poverty metrics at 30 kWh a month might be expected to address affordability most directly, the considerable variance in access rates at low grid-electricity poverty levels suggests otherwise. Access rates range from 9 percent to 87 percent at grid-electricity poverty indicators below 3 percent, making a meaningful relationship between the two unlikely. Examination of the relationships between access and the two measures of poverty shows that the poverty gap for 250 kWh and the poverty headcount for 100 kWh are the best predictors of access. Even so, the predictive power is not anywhere near that of the poverty gap measured at $3.10 per person per day.
High grid-electricity poverty indicators signal income being too low relative to the tariffs in effect. In the face of the inability of so many households to pay for electricity, could regulatory authorities keep tariffs relatively low compared to where they should be, even if so doing increases quasi-fiscal deficits? One way of probing this question is to examine the relationships between grid-electricity poverty indicators and quasi-fiscal deficits. Correlation statistics show that the grid-electricity poverty indicators for 30 kWh are not well correlated with quasi-fiscal deficits, suggesting that making the subsistence level of electricity affordable does not drive deficits. Underpricing of tariffs (quasi-fiscal deficits at benchmark performance) had the highest correlation coefficients with grid-electricity poverty indicators, but at consumption volumes far above the subsistence level: the highest correlation coefficients are for the grid-electricity poverty headcount for 250 kWh, followed by the grid-electricity poverty gap for 250 kWh and the grid-electricity poverty headcount for 100 kWh. In every case, the sign was negative, as expected: the smaller the degree of underpricing relative to GDP, bills sent out, or cash collected, the more unaffordable electricity becomes with increasing consumption.
How do monthly bills for 30 kWh compare to actual household spending? Figure

Female -headed households
The share of female-headed households in Africa has been rising over the last two decades (Milazzo and van de Walle 2015). If female-headed households face unmeasured economic disadvantages-such as having greater difficulties accessing credit or not having title to landtheir ability to spend cash at the same total household spending as defined in this study will be constrained. Is there evidence that female-headed households face greater challenges in gaining access to electricity, or that they use less of it once they have access? To investigate this question, the study looked at access to and spending on electricity for male-versus femaleheaded households.
Regression analysis shows that once income (per capita and household) and place of residence (urban and rural) are separately accounted for, female-headed households were not any less likely to connect to electricity than their male counterparts; they tended, in fact, to be more likely to do so. The results on spending on electricity similarly point to no apparent disadvantage for female-headed households. However, female-headed households tended to be poorer.  Note: Household weights are used in the calculations. The three sets of data are confined to those households that reported positive expenditures on electricity, including solar and standby generators but excluding batteries or lamps fueled by kerosene or liquefied petroleum gas. Those reporting values for free electricity are excluded. The poor are those classified as officially poor by the respective government.
These findings suggest that focusing on making electricity affordable for the poor will go a long way in increasing electricity use among female-headed households.

Tarif f schedules and multiple connec tions
This study collected information on the residential tariff schedules in effect as of July 2014  Many countries have lifeline rates-discounted rates based on household consumption of electricity intended to help the poor by cross-subsidizing low-consumption households. In the 39 African countries studied, the most common lifeline block size is 50 kWh a month (8 countries), followed by 25, 75, and 100 kWh (3 countries each). Eight countries have lifeline blocks up to 40 kWh, and five are 25 kWh or smaller-that is, below the monthly subsistence level of 30 kWh.
All five of these latter countries except Benin have increasing block tariffs. South Africa (more specifically, Johannesburg) may appear to have an exceptionally large first block, followed by The Gambia. However, there are special provisions in South Africa that allow the poor to receive free grid electricity for the first so many kWh-including 25, 50, 60, 100, and 150 kWh a month-depending on eligibility criteria (for example, prepaid customers consuming less than so many kWh a month based on the average over the last 12 months), which in turn differ by municipality. Fixed charges punish low-consumption households. For example, unit tariffs (price per kWh) are higher for monthly consumption of 30 kWh than 50 kWh in 15 countries due solely to fixed charges.
Subsidized lifeline rates are limited only to those consuming less than the cap in nine countries (Benin, Cabo Verde, Cameroon, Gabon, Ghana, Mozambique, Nigeria, São Tomé and Príncipe, and Togo), shifting households to higher tariffs in the next tier for the entire consumption if the cap is exceeded. This has the same effect as volume-differentiated tariffs in Gabon, Mozambique, and Nigeria. Not having access to the lifeline rate by exceeding the limit by even 1 kWh makes it more difficult for the poor when the block size is relatively small (for example, less than 50 kWh). The median increase in the effective unit energy charge for consuming more than the limit on the first block is 65 percent. There is large variation across countries, however: the increase ranges from 4 percent in The Gambia and 7 percent in Senegal to 340 percent in Madagascar and 450 percent in Kenya and Zimbabwe.  7). "Informal fees" for the initial connection add to the affordability challenge (box 2).

What scope exists for further cross-subsidizing low-income households?
To achieve universal access in Africa, the poor will have to be cross-subsidized for the foreseeable future to make the connection fee and subsistence consumption of electricity affordable. Cross-subsidies are easier to implement if utilities are not suffering from serious underpricing, and consumption by poor households makes up only a small fraction of total consumption. Of the 39 countries in which this study collected information from utilities, 11 provided information by consumer category, isolating residential consumers from others. Another five isolated low-voltage from higher-voltage consumers. The ratios of residential tariffs to other tariffs-and, in the absence of information about residential customers, of low-voltage to higher-voltage tariffs-give an indication of how much residential tariffs are already cross-subsidized.

Box 2 Bribery for connection to grid electricity
Corruption in the form of bribes adds to the cost of using electricity. Malawi and Nigeria asked households connected to the grid if they had to pay an "informal fee" over and above the official connection charge to get the connection. Bribery was more prevalent in Nigeria, where more than half of the households reported having paid a bribe. In Malawi, there were too few households in the bottom half of income groups for meaningful statistics, but in Nigeria there were enough households connected to the grid across all income groups. The percentage of households reporting informal payments was lowest in the bottom 40 percent and highest in the top 40 percentthat is, those who had greater financial means were also more likely to be approached for a bribe and to pay it.
The share of total kWh billed to residential customers is another indication of the room available for further cross-subsidization. The residential share of electricity consumption in Africa tends to be high. For comparison with a mature electricity sector with detailed information and a relatively high share of residential consumption, the United States was selected for comparison. Table 5 summarizes the findings.
Cabo Verde, Gabon, Liberia, Niger, and Uganda have no or negative underpricing (figure 5), suggesting some scope for cross-subsidizing the poor. However, the high share of total kWh billed to residential (Niger) and low-voltage (Gabon) customers increases the unit cost of supply and poses a challenge to adding more low-tariff, low-volume residential consumers, who are costly to connect because of lack of economies of scale. In Malawi and Guinea, both of which suffer from significant underpricing, residential (Malawi) and low-voltage (Guinea) consumers are already heavily subsidized. Not surprisingly, Malawi has a high grid-electricity poverty gap and headcount (table 3). In such cases, the electricity sector might consider targeting access expansion first to those able to afford higher tariffs. Guinea faces the additional challenge of three-quarters of all consumers belonging to the low-voltage category. Mali and Mauritania are two other countries with low rates of access and low-voltage consumers accounting for Note: Calculations are based on values and kWh billed. R = residential; C = commercial; I = industrial; NR = non-residential; LV = low voltage; MV = medium voltage; HV = high voltage; n.a. = not applicable.
a. The access rate in Lesotho is for 2010, the year of utility data.
b. The median is that for the first 11 countries reporting residential consumption separately. more than half of total consumption. In at least Niger and Mauritania, these small customers are already well cross-subsidized. At the opposite end of the spectrum is Lesotho, where the statistics are comparable to those of the U.S. electricity sector. However, Lesotho suffers from some underpricing (0.15 percent of GDP), suggesting that the tariff levels need to be raised across all consumer categories. The median statistics in table 5 suggest that in general residential customers tend to be cross-subsidized more in Africa than in the United States and in other high-income countries, leaving less room for cross-subsidization.

Reliabilit y
Poor reliability of electricity services is arguably as serious a problem in Africa as low rates Improving reliability is one of the most critical complements to tariff reform, because no customer wants to be asked to pay more for continuing bad service. The pervasive lack of measurement of SAIDI and SAIFI is itself an indication of how far there is to go in tackling poor reliability. Without systematic monitoring, it will be difficult to measure progress and the effectiveness of steps taken to improve reliability.

Conclusions and Implications
I t is useful to highlight the limitations of this study-in terms of scope and data availabilitybefore summarizing its key findings and potential policy implications.
⦁ The study does not cover off-grid rural electrification. More than three-fifths of the population in Africa is still rural. Universal access through grid expansion alone would be too costly and impractical, because of the low density of customers coupled with the low consumption of electricity in many rural areas. Off-grid electricity, especially from renewable sources, would be cheaper than grid expansion, although still costly. The net cost to the electricity sector of reaching rural areas-not captured in this report-will be substantial. These exclusions are all important areas for future study and more in-depth country-level analysis. It is important to bear these limitations in mind in interpreting the following findings and policy implications; these are summarized in the following table.
The key findings of the study and policy implications are summarized below.

Policy implication and response option Anticipated benefit
Cost recovery and utilities' financial sustainability ⦁ Absent systemwide optimization, very few African countries have full cost recovery.
⦁ Give high priority to systemwide optimization through a least-cost development plan.
⦁ Future costs are minimized.
⦁ Eliminating operational inefficiencies can potentially achieve cost recovery in about one-third of African countries.
⦁ Reduce network and bill collection losses as top priority in all countries.
⦁ Install prepaid meters as a way of improving revenue collection.
⦁ Costs are lowered with little welfare loss.
⦁ Utility revenue is increased.
⦁ In the remaining twothirds, both system optimization and tariff increases are needed in addition to eliminating operational inefficiencies.
⦁ Focus tariff increases first on the better-off accounting for large shares of total consumption.
⦁ Consider small, frequent tariff increases instead of large, rare ones.
⦁ Communicate clearly and transparently in advance the timetable for tariff increases, and implement according to announced timetable.
⦁ Tariff increases generate revenue with minimal loss of welfare.
⦁ There is wider public acceptance of tariff increases.
⦁ Utilities are held accountable more.
⦁ External events outside electricity sector's control-oil price and exchange rate volatility, rainfall variabilityhave large effects on financial sustainability.
⦁ Take advantage of low world oil prices today to introduce automatic pass-through of fuel price and exchange rate fluctuations.
⦁ Plan ahead for long periods of drought by optimizing generation mix.
⦁ Automatic pass-through is not equated with price shocks, making it more acceptable to the public.
⦁ Reliance on expensive emergency leased power is minimized, lowering costs.
Improved utility management ⦁ Network system and bill collection losses account for a majority of utility quasi-fiscal deficits in half of African countries.
⦁ Give high priority to reducing unmetered consumption and bill collection losses among medium-and largesize consumers.
⦁ Install commercial information management systems.
⦁ Revenue is increased with minimal loss of welfare.
⦁ Metering, billing, bill collection, connection, and disconnection are all improved, as is customer service quality.
⦁ Service reliability is poor, and seldom measured systematically.
⦁ Reduce service interruption duration and improve customer service for bill payment first.
⦁ These short-term measures "buy time" for longer-term steps to improve service quality.
⦁ Public support is won for other reform measures.

Policy implication and response option Anticipated benefit
Improved utility management ⦁ Measure service reliability statistics systematically at end-user level.
⦁ Service reliability is improved because where to take corrective steps becomes clearer. ⦁ Consider pros and cons of mandating prepaid meters if service quality is poor.
⦁ Not mandating until service quality is improved may mean customers do not pay in advance for electricity not delivered when needed.
⦁ Where billing is based on estimated consumption and there is overbilling for undelivered electricity due to frequent outages, prepaid meters reduce payments.
Affordability of grid electricity ⦁ Connection costs are not affordable in many countries.
⦁ Optimize technical and financial arrangements for all aspects of electrification and move away from charging customers separately for new connection.
⦁ Connection charges can be tailored to customers' ability to pay; connection fees collected may be transferred to a ring-fenced electrification fund and used to accelerate electrification. ⦁ Practice of multiple household connections to a single meter is widespread.
⦁ Benefits of lifeline tariffs for the poor are negated.
⦁ Eliminate practice of multiple household connections, replacing them with individual meters.
⦁ Service quality is improved.
⦁ The poor benefit fully from lifeline rates.
⦁ About half of African countries have small first blocks with low lifeline rates.
⦁ Monthly electricity bills are unaffordable for some in the current setup.
⦁ Sharpen progressive tariffs to increase affordability, and concurrently develop more targeted social protection measures.
⦁ Subsidies become more efficient.
⦁ Install prepaid meters. ⦁ Poor customers can pay when disposable cash becomes available, not when they are billed monthly.
⦁ Poor customers do not risk disconnection. ⦁ In most countries, only small additional subsidies are needed to make grid electricity affordable to every urban household.
⦁ Install more meters; affordability depends critically on metering each household individually accurately.
⦁ Electricity is affordable to more people and subsidies needed are reduced.
⦁ Poor financial state of many utilities makes it difficult to cross-subsidize the poor further.
⦁ A financially viable, well-operating electricity sector is essential for making electricity affordable to more people.
⦁ Universal access is achieved earlier.

Key f indings
Africa lags all other regions with respect to household access to electricity, even after taking into account its level of economic development. At the same income level, South Asia had much higher rates of access. At present, the likelihood of having access to electricity is reduced simply by living in Africa.

Utilit y management
Poor management plagues the electricity sector in many countries. Where transmission and distribution losses are high, bill collection losses also tend to be high, suggesting that operational inefficiencies are not confined to isolated segments of the supply chain but permeate the sector. Transmission, distribution, and bill collection losses combined accounted for more than half of the quasi-fiscal deficits in 21 countries, and more than three-quarters in 13.
Service reliability is poor, and seldom measured systematically. In some countries, households interviewed for national expenditure surveys reported daily blackouts, and poor reliability is supported by Doing Business indicators. Aside from inconvenience and discomfort, the cost to the economy of unreliable electricity service is significant-sometimes even driving investment in the economy away to other countries or regions. Indeed, two out of every five firms surveyed in Africa cites electricity as a major or severe constraint to doing business (World Bank 2016d). And yet few utilities are systematically measuring reliability using internationally accepted metrics: SAIDI and SAIFI statistics could be found in the annual reports of only four utilities; even in these cases, only national averages were reported. National averages by definition do not distinguish between areas with high reliability and those suffering frequent outages, making it difficult to assess utility performance. Nor are SAIDI and SAIFI measured at the level of individual customers. The absence of such statistics in itself could indicate to investors that the sector is not yet at a stage where even basic statistics are being collected, further discouraging investment in the economy.

Utilities' f inancial sust ainabilit y
The financial sustainability of electric utilities in many African countries is precarious. Of the 39 countries studied, only the Seychelles and Uganda were fully recovering their costs of supply, before taking system expansion into account. Quasi-fiscal deficits exceeded 100 percent of the cash collected by the utilities in 11 countries. Twenty countries were not even covering operational expenses; of the remaining 19, only 5 were covering half or more of their capital expenditures.

Improving operational efficiencies can potentially put the electricity sector on a sustainable
path in about one-third of African countries. If transmission, distribution, and bill collection losses combined could be reduced to 10 percent (the level considered for benchmark utility efficiency) and overstaffing could be addressed, an additional 11 countries might see their quasi-fiscal deficits disappear, bringing the total at full cost recovery to 13. In practice, achieving loss requires both time and financial resources, leaving the possibility of residual quasi-fiscal deficits even as losses are being reduced to benchmark levels.
In the remaining two-thirds of African countries, system optimization, tariff increases, ormost likely-both are needed for financial sustainability. Just as did the AICD, this study finds that the funding gap cannot be bridged entirely by eliminating operational inefficiencies in the remaining 26 countries. The median of the quasi-fiscal deficits at benchmark performance in these countries is 0.7 percent of GDP, which is not small. Among the highest such deficits is 2.8 percent of GDP in South Africa, which also happens to be the only country in the sample that provides free electricity and free connection to those meeting eligibility criteria. In all these countries, absent optimization of generation mix and cross-border trade, tariffs will need to be raised. The subsistence level of grid electricity is affordable on paper to the vast majority of the population in many countries with low rates of access. Electricity is affordable if 30 kWh a month costs no more than 5 percent of household income. If the tariffs in effect could be perfectly enforced (which means accurate metering of each house connected to the grid) and perfect targeting is possible, the grid-electricity poverty headcount and poverty gaps would be less than 5 percent in more than half of the 22 countries for which household survey data are available. Bridging the gap between the monthly bill for 30 kWh and 5 percent of household income for every urban household-for which grid extension would likely be the least-cost optionwould cost less than $1 million annually in 15 countries and less than $5 million in 19, and be no more than 1 percent of the cash collected by utilities in all 22 countries. More detailed country-specific analysis of this aspect of affordability would be useful.

High connection charges and informal payments for connection are further barriers to
increasing household access to electricity. Connection is free for some households in Mozambique, Senegal, and South Africa, but can be multiples of household monthly income in others.
High connection charges lead to households choosing not to connect as well as to shared connections. Seeking connection to the grid also provides an opportunity for unscrupulous utility staff to demand informal payments (bribes) for timely or even not-so-timely connection. For the poor, such malpractice reduces the affordability of electricity further.

Polic y implications
Within the electricity sector, political economy considerations point to the importance of prioritizing reform steps according to political risk. Measures that reduce the magnitude of requisite tariff increases, noticeably improve the quality of customer service, and avoid undue hardships on those with affordability challenges can help win public acceptance and expand access while moving the sector toward greater financial sustainability. In interpreting what follows, it is important to bear in mind that the exact sequence of steps and the design of a roadmap in each country will be guided by the specific circumstances of the electricity sector, government ownership and interest, the public's experience with the sector, and other country-specific factors.

Improving utilit y management
There are measures that can improve service quality in the short run. Improving the quality of service requires time and investment, but experience in other regions shows that early gains are possible by diligently attending to customers' complaints for outages and other incidents, restoring electricity supply in the shortest possible time, and making it easy to pay bills (through mobile phones, automatic teller machines, supermarkets, and other easily accessible locations with extended hours of service). Outage (or incident) management systems help deal with customer complaints about incidents related to the quality of supply; and commercial information management systems help improve metering, billing, bill collection, connection, and disconnection, and enable utilities to give full attention to customers. Service quality improvement does not call for significant investment, but requires incorporation of these information systems together with organizational arrangements for their proper use-for example, creation of call centers and use of a website and social networks to receive complaints and respond to them-and a more customer-oriented mindset (Antmann 2009). Global experience shows that, through such initial improvements, utilities can buy time (say two to three years) to achieve longer-lasting solutions through execution of investment projects to rehabilitate, upgrade, and expand the electricity infrastructure.
Loss reduction merits high priority. Reducing losses from unmetered consumption and inefficient bill collection-and sustaining such loss reduction over time-is much more important than addressing overstaffing, particularly against a backdrop of increasing demand and the need for rapid capacity expansion in every country.
Reliability statistics need to be measured at the end-user level. Few utilities report SAIDI and SAIFI, and those that do record incidents only in the high-and medium-voltage segments, ignoring what happens in low-voltage connections and computing SAIDI and SAIFI accordingly.
But interruptions experienced by customers matter. If properly measured, SAIDI and SAIFI values would be much higher, because low-voltage segments are a significant source of interruptions. In the absence of systematic measurement of service quality, it is not possible to assess utility performance or compare tariffs across countries in a meaningful way.
Shared connections degrade the quality of service to the multiply connected. Multiple connections to a single meter with low capacity can mean demand rising above the design value of the installed capacity, worsening the quality of service to those so connected. This is another reason to tackle the widespread practice of shared meters.
Accurately metering each individual customer can bring double dividends. Accurate metering of all customers-and especially medium-and large-size customers-is an integral component of any revenue protection program for a utility. Utilities with progressive tariffs will lose some revenue if meter sharing is eliminated, but that small loss of revenue is unlikely to be an important consideration. A lack of individual metering and tolerance of widespread (illegal) connections creates a more permissive environment for other malpractices, such as electricity theft and bribes demanded by utility staff. Individual metering means low-income customers can benefit fully from progressive tariffs, while offering the utility a chance to increase the efficiency of progressive tariffs further.

Winning suppor t for cost recover y
Sequencing is important in setting the electricity sector on a more financially sustainable path. Raising tariffs while outages continue unabated is bound to invite a backlash. Any perception that the better-off are unfairly benefiting-through large price subsidies offered to "strategic" industries or, worse, exploiting informal discretionary power of corrupt utility employees or executives who collude with large customers to reduce their bills-should be addressed as a matter of urgency. A revenue protection program should first focus on sales to large-and medium-size customers, who usually account for the bulk of commercial and bill collection losses. Cost reduction helps limit the tariff increases needed to attain cost recovery.
The first step is to eliminate unnecessary losses. The public may be more willing to accept tariff increases if priority is given in the near term to improving service quality and billing and collection efficiency. Reducing reliance on diesel-based electricity generation-including emergency leased power-and shifting to less expensive forms of generation overall could ease the pressure on tariff increases. The pursuit of sectorwide long-term optimization is crucial.
Low access rates may make it easier to address underpricing, provided the quality of service is improved. The bulk of total sales and revenues comes from those who can afford electricity-industries, businesses, and middle-and upper-income households. In most African countries, access rates among the poor are low; to the extent the poor are connected to the grid, they consume little. The financial viability of utilities therefore depends on charging tariff rates that would enable recovery of costs of efficient service delivery to better-off consumers accounting for the bulk of electricity sold. Although the political sensitivity of tariff increases to these consumers cannot be ignored, neither should it be overestimated. In the face of large quasi-fiscal deficits and low access rates, there is no compelling reason to subsidize those who can afford to pay more. If anything, they could be asked to cross-subsidize low-income consumers more, as long as the latter's total consumption is only a small fraction of the total electricity sold. Successful examples of power sector reforms in emerging countries in other regions show that middle-and high-income consumers in all tariff categories usually accept cost-reflective rates, provided the quality of electricity services is good. Therefore, in addition to reducing transmission, distribution, and bill collection losses, utilities' top priority should be to achieve an acceptable level of service quality to enable a trajectory toward cost recovery in tariff revenues.
Small, frequent increases may find wider acceptance than large price shocks. Quite a few countries freeze tariffs for years on end, only to find mounting deficits unsustainable and are thus forced to implement large, one-off tariff increases, followed by another long period of tariff freezes. One option for avoiding such large increases is to allow small, frequent tariff adjustments. Indeed, several African countries conduct regular and frequent tariff reviews, although adjustments are not necessarily implemented due to professed socioeconomic considerations.
A useful analogy is how India and Thailand handled fuel price subsidies: increase prices by small fixed amounts every month, announced in advance, until cost recovery levels are reached (Kojima 2016). It is important to communicate clearly and transparently in advance all the steps and the timetable for tariff increases toward cost recovery, and implement the steps according to the announced timetable. Predictability coupled with small and manageable increases could go a long way in finding greater public acceptance than rare, large, ad hoc tariff hikes.

Prepaid meters can help both utilities and customers.
For low-income households with cash flow constraints, the ability to pay in small increments overcomes the classic problem of indivisibility of electricity bill payments: making relatively large payments once a month can be far more challenging than making several small payments during the month. By not risking disconnection for payment failure, prepaid meters also help households avoid reconnection fees, which can be considerable in some countries. For the utility, prepaid meters improve revenue collection. On the downside, depending on how unreliable electricity service is and how customers are billed, prepaid metering could mean customers pay in advance for electricity that is not delivered when needed. It may be unfair to consumers to make prepaid metering mandatory in the face of unreliable electricity service. However, if monthly bills for some customers are based on estimated consumption and there is frequent overbilling for electricity not delivered due to widespread power outages, prepaid meters help reduce payments. For the utility, prepaid meters do not guarantee payment for electricity consumed, in that they can be bypassed by those determined to steal electricity.

Governments could take advantage of current low world oil prices and introduce automatic
pass-through of fuel price changes. Government control of energy prices-and passing through of input cost increases in particular-is politically sensitive. For diesel and fuel oil prices, low world prices offer a good opportunity to introduce an automatic price adjustment mechanism without provoking large tariff increases. As an example, the tariff structure in Kenya allows automatic transmission of oil price and currency fluctuations. There is no better time than now to take this step.

System optimization through a least-cost development plan becomes all the more important
in the face of large quasi-fiscal deficits. Although not covered in this study, system optimization can substantially reduce costs compared to uncoordinated infrastructure investments. In a region where investment needs have far exceeded available funding and cost recovery is a long-term goal for many, maximizing technical and cost efficiency is crucial. Cost reduction will make it easier to achieve financial viability. A financially viable, well-operating electricity sector in turn is essential for a successful grid-based access expansion program.

Making grid elec tricit y af fordable for the poor
The widespread practice of multiple household connections to a single meter decreases affordability. Progressive tariffs are intended to require the rich to pay more. In the absence of information about each customer's household income, the standard practice in developing countries is to use monthly consumption as a proxy for income. When several low-income households connect to a single meter, they appear to the utility as a single "rich" household consuming a lot of electricity. The average unit tariff of multiple connected households could be considerably higher than if each household was individually metered, and poor households could end up paying much more for consuming the subsistence level of electricity. On the one hand, multiple connections mean that limited financial resources can be pooled to enable connection to the grid; also, if one household cannot pay one month, others may compensate to avoid disconnection, assuming their eventual reimbursement. On the other hand, if the officially registered household fails to pay-for whatever reason-everyone could be disconnected, including those who have been paying promptly and fully. And the difficulty of determining who consumed how much may create a free-rider problem, whereby some households pay less than they should at the expense of others.
The approach to setting connection fees merits further examination. High connection fees contribute to the practice of multiple connections. Isolating installation of new connections from other components of an electrification project (low-, medium-, and high-voltage networks) is economically inefficient. There is no compelling argument for separating network assets and service connections. It is much more efficient to build all infrastructure needed to connect new users in a single project, optimizing technical and financial arrangements. Doing so still offers the option of collecting connection fees from users in installments according to their ability to pay, and the proceeds can be transferred to a ring-fenced fund for the specific purpose of accelerating electrification programs. If the user pays fully for the assets associated with connection (as in urban Peru, where asset ownership is assigned to the user), these assets should be excluded from the regulatory asset base of the utility for tariff determination.
Refining progressive tariffs, and concurrently developing more targeted social protection measures, can deliver more efficient results. The additional subsidies needed to make the subsistence level of electricity affordable to all households living in areas that can be potentially connected to the grid are not large. Findings suggest that it may be possible to achieve tariff affordability by refining lifeline rates and cross-subsidies. There are trade-offs in setting the lifeline tariff and block size. Increasing block tariffs benefits the rich and the poor alike, making the subsidies inefficient. Aside from problems associated with shared meters, volume-differentiated tariffs can be punishing for the poor because exceeding the block size by even 1 kWh would catapult the household into the next block, which could have a much higher unit tariff.
One option is to apply the block size to moving averages over the previous several months, but that would require modernization of metering and billing. Keeping the block size small-say 30-50 kWh a month-will ensure that the rich do not benefit disproportionately from heavily subsidized unit tariffs. To protect low-income families from unexpectedly high electricity bills, there could be a buffer in the form of a small second block that does not entail a large increase in unit tariff. Volume-differentiated tariffs in particular may merit such consideration. However well designed, consumption is not necessarily an effective proxy for income. Over the medium to long run, shifting subsidies to a comprehensive and integrated social protection program delivering cash to meet the basic needs of the poor-of which electricity is but one component-is the goal.
Technological advances present an opportunity and a challenge to the poor. Thanks to recent advances, the subsistence level of electricity consumption has been falling. The AICD considered 50 kWh a month a reasonable level; a decade later, it is 30 kWh. Using the most efficient appliances, even 15 kWh a month may be sufficient to burn four light-emitting-diode bulbs for four hours every night, charge a cell phone, and power a small efficient television and a large fan. However, efficient appliances are more expensive, and cash to pay the higher upfront costs is precisely what the poor lack. This is yet another dimension of affordability, and the resources needed to address this challenge arguably extend beyond the electricity sector. Utilit y charac teristic s Table A.2 provides utility data characteristics by country. Figure A.1 shows how utilities and other actors are positioned in the electricity sector of each country.

Methodology
A quasi-fiscal deficit is the difference between the net revenue of an efficient utility (R benchmark ) and the net current revenue (R current ). Let capex designate benchmark capital expenditure, opex designate benchmark operating expenditure, and Q designate dispatched kWh. The tariff at benchmark performance, tariff benchmark , in this study is (capex + opex)/0.9Q and the revenue of an efficient utility is tariff benchmark x0.9Q, where 0.9 accounts for combined transmission, distribution, and billing losses of 10 percent (the level considered for benchmark performance). The net revenue of an efficient utility is R benchmark − cost = R benchmark − (capex + opex) = 0, signaling that the revenue fully covers cost. The revenue at current performance by contrast is tariff current xQx(0.9 − TDL)x(1 − BL), where TDL is transmission and distribution losses in excess of 10 percent and BL is bill collection losses, while the current cost is capex + opex + overstaffing cost.
At benchmark performance, the last three terms are zero, leaving only underpricing.
This study estimates operating costs using the financial statement of the main utility listed in This methodology means that the sector structure has a significant effect on the computation of capital and operating costs. Figure A.1 shows how the structures are categorized into five groups. Various parameters used in the analysis are explained below.
⦁ Revenues (amounts billed) are taken from utility financial statements and comprise only those directly related to electricity sales that are retained by electric utilities. Subsidies in the form of direct transfers from the government or international donors are excluded.
Revenues not directly related to the sale of electricity are excluded, such as those earned from the sale of water for utilities that provide both services.
⦁ Operational expenditures include all fixed and variable operational and maintenance costs, and taxes that are not rebated such as corporate income tax. All costs deemed to be related to capital costs are excluded because they are replaced by calculated annualized capital costs for existing assets. All loan repayments-interest payments typically recorded on income statements and principal payments typically recorded on cash flow statements-are considered to be for capital costs. Other exclusions include depreciation, losses on foreign-denominated debt, costs not directly related to electricity sales (such as for providing water services), and costs from extraordinary activities. For full details, see Kojima et al. (2016).

Monthly payment s and connec tion charges
To assess affordability, bills to be paid by households for monthly consumption of 30, 50, 100, (where the connection fee is zero but new customers have to buy the materials needed for the initial connection), and São Tomé and Príncipe. If information on when the connection charges came into effect was not available, the dates of effectiveness are assumed to be the same as those for tariff schedules. Table B.1 summarizes monthly payments, connection charges, and adjustment factors by country.

Def ining access
The household expenditure surveys do not enable systematic analysis of access because the questions asked are not uniform. All but one survey (South Africa) asked about the primary source of energy for lighting. Sixteen asked about connection to the grid, and Togo's survey asked whether the household had spent money on grid electricity in the previous two months.

G rid-elec tricit y pover t y gap and subsidies
The poverty gap for grid electricity is computed as Required monthly payment − 5% of monthly expenditure for household i Required monthly payment where the required monthly payment is the monthly bill inclusive of taxes and other charges that a household has to pay to consume the corresponding amount of electricity, P is the total population living in households for whom the monthly payment exceeds 5 percent of total monthly household expenditures (inclusive of freely acquired food and other items), and N is the total population of the country. Where the monthly electricity bill exceeds the 5 percent share, electricity is deemed unaffordable; the degree of unaffordability for a household is the size of the gap between the bill and the 5 percent share when this is positive, zero otherwise.
Although monthly consumption of 30 kWh is the basis for defining affordability, the poverty gap is also computed for 50, 100 (multitier framework tier 4), and 250 kWh (multitier framework tier 5) a month to see how many people can afford higher consumption. This study also takes the numerator in the above equation and aggregates the affordability gap (where it is positive) across all households using household weights. The sum is the amount of subsidy needed to enable every household to keep spending on electricity at or below 5 percent of total household expenditure.

Female -headed households
This study carries out simplified regression analysis to see if, after accounting for total expenditures and location (urban or rural), female-headed households are any more likely to use electricity than male-headed households, and whether spending on electricity shows any differences. Probit regressions (for the first three variables listed below) and ordinary least squares (for the last) are carried out for urban and rural households separately on the following dependent variables: ⦁ Expenditure dummy for positive expenditures on electricity (1 if positive, 0 if zero or missing) ⦁ Electricity dummy for citing electricity of all forms, including generators and solar energy, as the primary source of energy for lighting or cooking (1 if electricity was used for lighting or cooking, 0 otherwise) ⦁ Grid dummy (1 if the household has access according to levels 1 or 2, 0 otherwise) ⦁ Logarithm of expenditures (log expenditure) on electricity for those households that reported positive expenditure, and repeating the regression with the sample confined only to those connected to the grid according to level 1 The following explanatory variables are tested using a 5 percent significance test (that is, the probability that the coefficient for the independent variable is actually zero when the regression shows a non-zero value is less than 5 percent): ⦁ Logarithm of household expenditures per capita ⦁ Logarithm of household size ⦁ Dummy for female-and male-headed households (1 for female, 0 for male)