70557 Asmara Power Distribution and Rural Electrification Project Pre-electrification Survey Analysis Report prepared by Kyran O’Sullivan, World Bank and Kevin Fitzgerald (Consultant), April 28, 2006 June 2005 Exchange Rate: Nakfa 15.375 = 1 USD Main Findings  Households in the project area spend as much on dry cell batteries as they do on kerosene for lighting—both readily substituted by grid electricity. Those with the lowest incomes spend 76 Nakfa/month (US$ 5/mo) and those with the highest incomes 102 Nakfa/month (US$ 6.6/mo)on average  These observed expenditures on these energy sources provide minimum estimates of household ability and willingness to pay for the higher quality services provided by grid electricity.  Even the poorest households appear to be well-positioned to afford the monthly cost of electricity—at the national tariff, the average 76 Nakfa they currently spend on lighting fuel and batteries would buy 33 kWh/month (two 60 Watt bulbs used 4 hours daily would consume less than 15kWh/month and cost 37 Nakfa/mo or US$ 2.40 at the national tariff).  Comparing low income households that currently buy electricity from a neighbor to those that light with kerosene shows that electrification could boost lighting levels more than twenty fold while cutting lighting fuel and battery expenditures, now consuming 10.2% of total cash outlays (US$ 5), by more than half (to roughly US$ 2).  Costs incurred by households that operate gasoline or diesel generators, commonly at very low load factors, and irrigation pump sets rival and exceed other household expenditures. Grid electrification may substantially cut the costs of these productive end uses (this finding is indicative only).  Several options for improving project evaluation during the second round survey are reviewed. All options are viable, but only those with a control group will be able to formally evaluate the impacts of the project itself. o Default option – re-survey original households after electrification. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 1 of 24 o Default + control group – the preferred option for the second round: re- survey the original households as well as households in a comparison group from outside the Project Area. o Post-electrification formal evaluation – the preferred option only if the original sample does not represent the population in the Project Area: survey newly drawn households from within and outside the Project Area. o Suggestions for strengthening the survey include adding survey sections on housing type (material), water supply and sanitation, health measures, education levels achieved, biomass fuels sources and use, land cultivated and under irrigation, and distances to water supply, schools, clinics, major roads, and markets—all factors that have been found to distinguish households when explaining the impacts of electrification. Introduction During supervision of the project (supervision mission May 9 – 19, 2005) it was agreed with the Ministry of Energy and Mines that in order to evaluate the impacts of electrification in the four project areas (59 villages in the Keren, Barentu, Dekemhare and Adi Keih areas), an ex-ante survey would be conducted prior to electrification of the villages and an ex-post survey will be conducted in late 2008 or early 2009 prior to project closing (electrification will be carried out during 2006 and 2007 at a cost of approximately $11.5 million). Accordingly, the Ministry of Energy and Mines, designed and implemented a survey that was conducted during July to September, 2005. The survey covered 18 villages out of 59 targeted villages and 398 sample households were interviewed (there are approximately 4,200 households and 22,500 people living in the 59 villages). The survey was designed to characterize energy use in existing institutions, commercial enterprises and households before electrification and, thereby, set a baseline for monitoring and evaluation of the impacts of the project going forward. The report prepared by the Eritrea Department of Energy and Mines in October 20051, presents findings at the village and household level. In addition to presenting a preliminary analysis of socio-economic characteristics and energy use in 398 sampled households, it summarized existing electricity (gensets and solar systems) and motive power sources (diesel pumps), institutions and commercial enterprises in each of the 18 villages. It documents income generating activity and social infrastructure services (schools, clinics, etc) and their use of energy. The current inquiry complements the report of the Dept. of Energy and Mines. The objectives are twofold: (i) to the extent possible, make an ex-ante estimate of the benefits to households of electrification under the project and; (ii) review the survey 1 Poverty and Social Impact Analysis in the Rural Electrification Component of the Asmara Power Distribution and Rural Electrification Project: Pre-electrification Survey Report, Eritrea Department of Energy, October, 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 2 of 24 methodology and suggest modifications to the survey instruments for the 2nd phase of surveying to be conducted ex-post (after the villages have been electrified) 1. Estimation of Benefits using Household Survey Data The pre-electrification survey of households included 398 households in 18 of 59 villages to be electrified under the project. The sample frame, presented in the Appendix, was stratified by income group. Caveats: An initial exploration of the data showed that the sample frame itself was purposive: designed to characterize existing patterns of energy use, but also to make sure existing sources of electricity and motive power were not overlooked. For example, of nearly 1,800 households in Tokombia, 7 have private generators and 3 of these were included in the set of 33 households surveyed (DOE 2005). Since households with generators and irrigation pump sets were apparently over-sampled, relative to their frequency in the population, survey results that include them would bias results. This analysis does include these households, but expenditures for generating electricity and for irrigation pumping are reported separately. Moreover, it is unclear if the sample frame is self-weighting by income group. Put differently, are 20%, 25% and 55% of households in the project area in higher, medium and lower income groups? Because the answer to this question is unknown, this analysis is done separately for sampled households in each income group and results pertain only to households that have the general characteristics of those surveyed. No generalizations are made to households in the sampled villages or in the Project Areas. During data cleaning, 5 households were eliminated from the analysis due to extremely high expenditures on kerosene or electricity generation. 393 of 398 sampled households are included in the analysis below. Family size TABLE 1. HOUSEHOLD COMPOSITION Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Household size 6.80 80 6.36 97 5.26 216 5.84 393 Number of adult household members 2.80 80 2.60 97 2.24 216 2.44 393 Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 3 of 24 Number of household members less than 5 1.18 80 1.11 97 .93 216 1.03 393 Number of household members between 6 2.83 80 2.65 97 2.09 216 2.38 393 and 8 No of children in the household that attend 2.83 80 2.44 97 1.94 216 2.24 393 school Asmara RE Project pre-electrification survey 2005. Family size in Table 1 shows the expected correlation with income, but otherwise little variation across income groups with a mean of 6 members per household. The average family has a young child and two of school age. One of the chief benefits of electrification is provision of better and more reliable light for studying at night. Because the sampled families in the project villages have, on average, two children in school and one of pre-school age, they can be expected to benefit from the higher quality light that electricity provides. The benefits of electrification at the household level have been examined using statistical methods in other countries. Such studies statistically compare similar households, with and without electricity, to draw valid inferences about the benefits of electrification. Since the survey fielded for this project characterized households in areas before grid electricity is available, this kind of approach to estimate the benefits of electrification will not be possible here. But it will be possible to examine existing fuel use for the chief end uses that grid electricity may provide more cheaply and reliably as a basis for drawing conclusions. After reviewing overall income and expenditures in sampled households, current expenditures on end uses efficiently powered by grid electricity are examined. Namely: grid electricity substituting for lighting kerosene, candles and batteries and grid electricity displacing the fuel and operating costs of private generators and irrigation water pumps. Household income Following years of civil strife, Eritrea is now one of the poorest countries in the world. Latest figures for 2004 showed a total population of roughly 4.4 million with an average annual per capita income of US$ 210. Roughly 70% of the population live in rural areas, most engaging in subsistence agriculture. Table 2 shows income sources for sample households. For reference, all Income and Expenditure Tables presented in the body of this report are reproduced in the Appendix using 2005 US Dollar equivalents. At an average per capita income of Nakfa 4,432/yr (US$ 288 USD), only the Higher Income households surveyed in the project area have incomes above the 2004 national average. Sampled households in Medium Income and Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 4 of 24 Lower Income categories reported annual per capita incomes averaging only Nakfa 2,315 and 1,054 (US$ 151 and US$ 69). As observed in DOE 2005, incomes may have been consistently under-reported by surveyed households, particularly in the Lower Income group. For this reason, expenditures are assumed to provide a more accurate picture of household cash resources. Nonetheless, by any measure, the sampled households in Medium and Lower Income groups are extremely cash poor. Farm income makes up nearly half of total income across all income groups. Government aid and wages are very important in the Lower Income group, each making up roughly 1/3 of total income, on average. Higher Income households are much more likely to own a business, making, on average, nearly 1/3 of total income from trading. TABLE 2. ANNUAL HOUSEHOLD INCOME (Nakfa/year) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 30,139 78 14,725 94 5,545 202 12,981 374 Income/cap 4,432 2,315 1,054 2,223 Farm income 16,459 80 6,700 96 2,182 210 6,264 386 Non-farm income 14,678 78 8,411 95 3,667 205 7,131 378 Wage income 2,698 78 3,749 95 1,941 206 2,550 379 Trading 9,426 78 3,587 95 902 208 3,317 381 Remittances 1,027 78 245 95 89 210 319 383 Suwa or injera 10 78 408 95 318 208 277 381 Home business 172 78 20 95 88 208 89 381 Government aid 2,212 79 2,609 97 2,389 216 2,408 392 Other 652 78 397 96 348 207 422 381 Asmara RE Project pre-electrification survey 2005. TABLE 3. HOUSEHOLD EXPENDITURES (Nakfa/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 2,511.59 78 1,227.07 94 462.07 202 1,081.79 374 Farm income 1,371.59 80 558.33 96 181.80 210 522.04 386 Monthly expenditures 1,978.46 79 938.72 96 746.88 213 1,045.11 388 Food 797.56 80 477.18 97 384.74 214 492.14 391 Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 5 of 24 Clothing 152.18 80 121.55 97 81.12 214 105.68 391 Schooling & health 80.84 80 49.90 97 38.99 214 50.26 391 Firewood & biomass 52.03 80 30.48 97 29.60 213 34.42 390 Commercial fuel 557.63 79 154.69 97 115.53 216 214.32 392 Travel 316.10 80 76.51 97 46.73 216 108.91 393 Rent 29.48 80 20.67 97 10.91 214 17.13 391 Other 158.64 80 25.74 96 36.91 214 59.13 390 Asmara RE Project pre-electrification survey 2005. Household expenditures Table 3 shows average monthly expenditures by income group. Lower income households report substantially higher monthly expenditures than their reported incomes would support. But at Nakfa 747/month, low income households still spend, on average, under $50/month—half of which goes to purchase food. Poor households in the sample paid, on average, US$ 2/month on firewood and other biomass fuels. This is a small but not insignificant amount for these households. As most rural dwellers in the developing world collect wood and agricultural residues, some times from considerable distances, paying cash for these commodities indicates scarcity. Average expenditures on commercial fuels are more troubling. Taken by themselves, they would seem to indicate that sample households in Higher, Medium and Lower Income groups spend, on average, 28%, 17%, and 16% of all cash outlays on commercial fuels. If accurate, this would signal severe hardship. But these averages are misleading— they are biased upwards by the few households that generate electricity and those that use gasoline or diesel water pumps. These productive uses are treated separately below. Energy expenditures Table 4 breaks down expenditures for each fuel and major end use. Not all households pay for firewood and biomass fuels, but the share of those who buy at least some of their firewood increases with income. Across all households surveyed, the amount spent on biomass fuels is, on average, roughly equivalent to the amount spent on kerosene for cooking, boiling water and ignition.2 Nearly all surveyed households use at least some kerosene for purposes other than lighting. Across the sample, households spend a substantial share, roughly 6%, of total expenditures on fuels for cooking and heating. 2 Each household was asked to estimate the share of kerosene used for lighting, cooking, boiling water and fire ignition. These responses were used to allocate expenditures on kerosene to lighting and to other end uses. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 6 of 24 This is nearly as much as they spend on lighting fuels. This analysis assumes that electricity will not displace kerosene or biomass fuels in cooking and water heating end uses in rural households. Lighting & battery expense is the sum of expenditures for kerosene used for lighting, candles, dry cell batteries and car battery recharging, if used. While it is not surprising that almost all households use kerosene and candles for lighting, the widespread use of dry cell batteries is unusual. Moreover, the substantial share of expenditures to purchase batteries rivals the amount spent on kerosene for lighting. Dry cell batteries are almost universally used to power radios and flashlights (only one surveyed household used a battery powered lamp). Lighting and battery uses are all substitutable by grid electricity. As such, the amount of cash that a sampled household now pays for these energy sources can be taken to indicate the household ability to pay for grid electricity. Each household may be willing to pay substantially more for reliable grid electricity and may in fact consume more services (light, radio, television, etc.) once connected to the grid, but it is a safe assumption that the amount now spent on lighting and battery powered end uses can be taken as a proxy for the household minimum ability and willingness to pay for grid electricity. TABLE 4. ENERGY EXPENDITURES (Nakfa/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N ALL HOUSEHOLDS Income 2,511.59 78 1,227.07 94 462.07 202 1,081.79 374 Farm income 1,371.59 80 558.33 96 181.80 210 522.04 386 Monthly expense 1,978.46 79 938.72 96 746.88 213 1,045.11 388 Firewood & biomass 52.03 80 30.48 97 29.60 213 34.42 390 Kerosene for cooking, etc. 58.32 80 38.41 97 25.53 216 35.38 393 Lighting & battery expense 102.86 80 83.76 97 76.42 216 83.61 393 Kerosene for lighting 45.76 80 34.94 97 39.75 216 39.79 393 Candles 6.43 80 4.08 97 3.52 216 4.25 393 Dry cell batt 47.89 80 43.36 97 32.80 216 38.48 393 Storage batt 2.78 80 1.39 97 .35 216 1.10 393 Light/batt expense share 5.2% 8.9% 10.2% HH electric expense 85.61 80 4.97 97 8.02 216 23.06 393 Irrigation expense 623.26 80 53.71 97 2.12 216 141.30 393 Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 7 of 24 USERS ONLY firewood cost 143.52 29 109.52 27 128.67 49 127.85 105 light+batt cost 102.86 80 83.76 97 77.13 214 84.04 391 electric cost 402.86 17 80.28 6 108.28 16 232.37 39 irrigation cost 1,994.45 25 868.38 6 114.58 4 1,586.56 35 Asmara RE Project pre-electrification survey 2005. Sampled households in the Lower, Medium and Higher income groups spend, on average, 76, 84, and 103 Nakfa/month, on average, on lighting kerosene, candles and batteries. This amounts to 10%, 9% and 5% of total cash outlays by households in these groups, respectively. These current expenditures for lighting kerosene, candles and batteries by sampled households in the Project Area will be taken as proxies for minimum spending power on grid electricity. Cash outlays by sampled households for electric bills, electric generation and water pumping are discussed in the next section. In 2004, the residential tariff for customers outside of Massawa and Assab, was composed of a fixed standing charge of 6 Nakfa/month + 2.15 Nakfa/kWh (Tariff Study Update, Table 8.13). Median household consumption in 2004 (mostly urban households) was 60 kWh/month with a substantial share of households consuming 20 kWh or less per month (Tariff Study Update, Figures 8.1 and 8.2). Households in the project area that are similar to the households sampled in the survey would likely be able to afford the monthly cost of electricity at the tariff rates cited above. At 2.15 Nakfa/kWh, households consuming 20, 40 or 60 kWh/mo would have monthly costs of 6+43=49, 6+86=92 and 6+129=135 Nakfa/mo, respectively. These costs compare very favorably to existing expenditures 76, 84 and 103 Nakfa/month now spent by Lower, Medium and Higher income households, particularly when considering that newly electrified lower income rural households elsewhere have been found to rarely consume more than 30kWh/month during their first few years. Even households in the Lowest Income group appear to be well-positioned to afford the monthly cost of electricity: at the national tariff, 76 Nakfa currently spent on lighting fuel and batteries, on average, would buy 33 kWh. Note that two 60 Watt bulbs used 4 hours daily consume less than 15kWh/month (37 Nakfa/mo or US$ 2.40 at the national tariff). Less than 20% of the entire sample reported spending under 37 Nakfa/month on lighting fuels and batteries. As such, the vast majority of households surveyed in the Project Area could afford electricity for at least two 60 Watt bulbs at the national tariff. 3 Tariff Study Update, Asmara Power Distribution and Rural Electrification Project, Interim Report, PB Power, November 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 8 of 24 This indicative analysis uses current substitutable expenditures to proxy for ability to pay for electricity at the national tariff. If in addition to these recurrent charges, households were required to pay a substantial connection charges, those charges would likely limit connection rates, particularly in lower and medium income groups. The extent of the barrier that a substantial connection fee might place on actual connections cannot be estimated from this study. The use of current expenditures as a proxy for electricity spending power does not require complete electric substitution for these other fuels. Experience in other countries shows that once rural households are electrified, they build up their electricity use slowly and do not stop using kerosene and candles for lighting. Depending upon the number and wattage of fixtures and appliances, loads for most lower income rural households rarely exceeds 30 kWh/mo in the first few years of electrification. The typical pattern is for newly electrified rural households to reduce their use of kerosene and candles gradually over several years as the household obtains lighting fixtures and appliances that use grid electricity. Existing tariff rates for municipal systems in several towns in the sample frame as well as flat rate charges for some private mini-grids are reviewed in DOE 2005. Tariffs for most of these existing systems are higher than the national tariff cited above. Accordingly, the current tariff rate in Akrur (page 17) and in Korbaris (page 25) is reported at 2.75Nakfa/kWh (standing charge not reported). The Municipal generator in Afabet charges households 20 Nakfa/month/lamp and 80 Nakfa/month/refrigerator. Finally, some households in the peri-urban areas around Tokombia are served by private generators and pay a fixed charge of 60 Nakfa per lamp and 20 Nakfa per radio per month. These isolated systems commonly run only a few hours each day, operate at very low load factors and even with charges higher than the national tariff, appear to run at a substantial financial loss. Twenty-six of the households that were surveyed in the project area, in all income groups, use electricity and pay a monthly bill. All of these pay a flat charge based on number of bulbs/appliances. The average monthly bill is Nakfa 50/month (US$ 3.25/mo) (see Table 6 below). Most of these households that have already revealed their willingness to pay for electricity would benefit from more reliable grid electricity, particularly if charged the national tariff. Lighting substitution potential Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 9 of 24 Lighting levels delivered by kerosene lamps were estimated for each household using survey responses and technical characteristics of standard kerosene lamps.4 Average liters consumed and lighting delivered by lamp type are shown in Table 5. Simple wick lamps are not much brighter than candles. More than half the lighting kerosene is consumed in regulated wick lamps, commonly called hurricane lamps. Because hurricane lamps are more efficient than simple wick lamps, they deliver more than 75% of the light from kerosene in sampled households of every income group. TABLE 5. KEROSENE LIGHTING Users Only (liters and klmh/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Lighting kero (l/mo) 4.59 76 3.43 96 3.95 212 3.95 384 Simple wick share .33 75 .34 96 .47 210 .41 381 Regulated wick share .67 75 .66 96 .53 210 .59 381 Petromax share .00 75 .00 96 .00 210 .00 381 Kerosene light (klmh/mo) 6.34 76 4.68 96 5.34 211 5.37 383 Simple wick (klmh/mo) 1.30 76 1.07 96 1.29 211 1.24 383 Regulated wick (klmh/mo) 5.05 76 3.61 96 4.05 211 4.14 383 Petromax (klmh/mo) .00 76 .00 96 .00 211 .00 383 Asmara RE Project pre-electrification survey 2005. To set these kerosene lighting levels in context, similar estimates were derived for the 26 sampled households that pay a monthly bill for electricity from a municipal system or a neighbor with a generator. Table 6 shows average estimated electric use and lighting delivered in households that use electricity by income group.5 All of these households pay a flat monthly charge based on number of bulbs and appliances. The figures in Table 6 are not meant to predict electricity usage after electrification—the number of households is too few and the quality of service to different to be predictive. Table 6 merely illustrates the quality of service and lighting levels delivered in the sampled households that now pay an electric bill. 4 Technical characteristics of standard lamps are presented in the Appendix and are from Nieuwenhout, FDJ, PJNM van de Rijt, and EJ Wiggelinkhuizen, 1998, Rural Lighting Services: A comparison of lamps for domestic lighting in developing countries, Energieonderzoek Centrum, Netherlands. 5 Households that generate electricity for own use or for sale to neighbors were excluded from this Table. Household responses about the total installed wattage of various appliances and average hours of daily use were employed to estimate total monthly kWh consumed in each type of device. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 10 of 24 Roughly half of the current electricity users in Table 6 report their source of electricity is available 30 days per month, with the rest reporting 15 days per month of service, on average. Most of these sources deliver electricity less than 5 hours per day. Even though the observations are few, the lighting levels delivered by incandescent and fluorescent lamps are illustrative. The 13 households in the Lower Income group use electricity almost exclusively for incandescent lighting. They consume, on average, less than 10kWh/mo overall and enjoy about 145 klmh/mo.6 TABLE 6. ELECTRIC USE and LIGHTING Bill Payers Only (Nakfa, kWh & klmh/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Days per month the household get electricity 21.50 8 21.20 5 22.92 13 22.15 26 from gen-set Gen-set service hours per day 3.88 8 4.60 5 4.46 13 4.31 26 monthly elec bill (Nakfa) 49.38 8 39.00 5 49.62 13 47.50 26 Total est. elect kWh/mo 24.16 8 14.17 5 9.68 13 15.00 26 incandescent (kWh/mo) 13.63 5 7.27 5 11.83 9 11.11 19 fluorescent (kWh/mo) 10.35 4 4.50 1 2.00 5 5.59 10 radio (kWh/mo) . 0 .00 1 2.80 1 1.40 2 color TV (kWh/mo) 13.20 1 .00 1 6.60 1 6.60 3 B&W TV (kWh/mo) 9.00 1 .00 1 . 0 4.50 2 refrigerator (kWh/mo) 30.75 2 30.00 1 . 0 30.50 3 Total est. light klmh/mo 412.74 8 141.26 5 144.46 13 226.39 26 incandescent (klmh/mo) 163.58 5 87.26 5 142.00 9 133.28 19 fluorescent (klmh/mo) 621.00 4 270.00 1 120.00 5 335.40 10 Asmara RE Project pre-electrification survey 2005. Using Lower Income households as an illustrative example of the benefits of electrification: they pay, on average, 50 Nakfa for 10 kWh delivered on 23 days each month. At the national tariff, these households’ monthly cost would be cut roughly in half to 27.5 Nakfa for 10kWh and it should be available more than 23 days each month. Moreover, they would enjoy 145 klmh of electric lighting and would likely be able to 6 Households in medium and Higher Income groups use more fluorescent lighting. Since fluorescent lamps are roughly 5 times as efficacious as incandescent lamps, they enjoy more light per kWh. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 11 of 24 displace the bulk of their dry cell battery needs (for powering a radio) for under 30 Nakfa each month. Meanwhile, the average non-electrified Lower Income household would still be getting only 5.34 klmh (Table 5) for 39.75 Nakfa from kerosene and paying 32.80 Nakfa for dry cell batteries each month (Table 4). This example indicates that making electricity available to Lower Income households in the project area could boost lighting levels more than twenty fold (145 klmh / 5.34 klmh = 27) and cut lighting fuel and battery expenditures, now consuming 10.2% of total cash outlays (76.4 Nakfa/month (US$ 5), Table 4), by more than half (to roughly 30 Nakfa/mo (US$ 2) as above). Dramatic improvements in lighting are immediate benefits that are commonly observed in rural electrification programs. Cost reductions usually take several years as households build up electric lamps and appliances. But it may be reasonable to expect that substantial outlays for dry cell batteries may be reduced very quickly by electrified households in the Project Area. Electric and irrigation expenditures Table 7 shows household expenditures for electricity, generation and for operating irrigation pump sets. As stated at the outset, in Caveats, it appears that households owning and operating generators and pump sets may have been over-sampled relative to their frequency in the population at large. Moreover, it is clear from the figures reported in Table 7 that expenses for electric generation and irrigation are unique to these households and may not have been included in overall household expenditures. In addition, relying on mean expenditures across all households can be misleading when only a few sample observations have a generator or pump set. As such, electricity use, generation and water pumping are analyzed separately in this section. Electricity consumption Electricity use in the 26 sample households that reported paying a monthly bill have been analyzed above. These electric expenses were not included in the lighting and battery expenses reported above. But since so few households in the sample use electricity, the means used as proxies for spending power on grid electricity are biased downwards only slightly, by 3 Nakfa / month, on average. Electricity generation Fifteen households, almost all of which are in the Higher Income group, reported owning and operating a generator. Table 7 shows average monthly fuel and maintenance costs for these generators. These costs have been allocated evenly to the number of households served when the generator is operated. Even so, monthly expenses are substantial—more than 500 Nakfa/month/household served (US$ 32). The amount of electricity generated by these self-generators is unknown. However, even with costs allocated evenly across Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 12 of 24 the number of households served, they are an order of magnitude above the average electric bill reported by other households in the sample frame. There are several possible explanations for these very high reported costs. These generators may be grossly inefficient. Old equipment and low load factors are commonly cited for the systems reviewed in DOE 2005. Another explanation may be that they are supporting demands that are more in line with higher income urban household consumption, including refrigeration and possibly even air conditioning. Perhaps both explanations are operative here. If the reported costs are accurate, however, the households that now generate electricity for own consumption and for their neighbors would likely benefit substantially, in terms of reduced costs, from grid electrification. TABLE 7. ELECTRIC & IRRIGATION EXPENDITURES (Nakfa/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N HH electric expense 85.61 80 4.97 97 8.02 216 23.06 393 monthly elec bill 4.94 80 2.01 97 2.99 216 3.14 393 genset oper cost 54.97 80 2.84 97 4.15 216 14.17 393 genset maint cost 25.70 80 .12 97 .88 216 5.75 393 monthly elec bill 49.38 8 39.00 5 49.62 13 47.50 26 genset oper cost 399.77 11 275.00 1 298.78 3 371.25 15 genset maint cost 186.92 11 11.67 1 63.69 3 150.59 15 Irrigation expense 623.26 80 53.71 97 2.12 216 141.30 393 elec for irrig cost 27.91 80 .00 97 .00 216 5.68 393 irrig pump fuel cost 595.35 80 53.71 97 2.12 216 135.62 393 irrig monthly cost 1,994.45 25 868.38 6 114.58 4 1,586.56 35 elec for irrig cost 744.33 3 . 0 . 0 744.33 3 irrig pump fuel cost 1,905.13 25 868.38 6 114.58 4 1,522.76 35 Asmara RE Project pre-electrification survey 2005. Irrigation water pumping Thirty-five households reported operating a pump set for irrigation. Irrigation is a substantial productive use for motive power and pump sets are a primary target for electrification. Many other countries, where irrigation has been shown to dramatically increase yields and allow multi-cropping, have made electrification of irrigation pump sets a policy priority and have even premised their rural electrification programs on agricultural returns to irrigation. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 13 of 24 This survey, because it was intended to characterize the baseline, did not go into agricultural yield details. As such, it is not possible to formally assess the economic returns to irrigation through electrification of pump sets using this data set. But it is clear from the costs reported in Table 7 that irrigation costs can easily dwarf household energy costs and even normal household expenditures. Average irrigation costs across all income groups (1,600 Nakfa/mo) exceed mean total household expenditures across the entire sample by 50%. Irrigation costs for households in the Higher Income group that irrigate are roughly equal to mean household expenditures by all households in that group. Water pumping for irrigation, drinking water and sanitation are major productive end uses that merit further study, particularly owing to arid conditions in the Project Area and predominance of rain fed agriculture. Water pumping for irrigation and public water supply may provide an overall economic rationale for this and future rural electrification projects in Eritrea. 2. Recommendations for ongoing monitoring and evaluation of the project The Poverty and Social Impact Analysis, DOE 2005, amply summarized village level and household level data from the pre-electrification survey of enterprises and households in 18 of the 59 villages to be electrified under the Asmara Power Distribution and RE Project. That report was comprehensive, identifying institutions and enterprises for electrification at the village level as well as giving an overview of findings at the household level. The team that fielded the survey and quickly turned it around in a few short months into the preliminary analysis contained in that report should be commended. The current report is intended to supplement the DOE 2005 report. The survey instrument itself is well-designed for characterizing a baseline of energy use patterns and practices in enterprises and households that might be substituted by grid electricity. It goes into considerable detail on lighting kerosene use and lamps, electricity consumption and productive end uses—namely private electricity generation and irrigation. Most sampled households use kerosene for lighting and the questionnaire was sufficiently detailed to support an estimation of lighting levels achieved by each household. The few surveyed households that consume electricity were not sufficient to undertake a statistical evaluation of the benefits of electricity substitution for lighting kerosene. But the survey was detailed enough to enable an indicative analysis to be completed. Research design for assessing the impacts of electrification Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 14 of 24 When the impacts of any treatment, in this case electrification, are formally assessed it is common practice to use multivariate regression tools to control for the many differences between observations (in this case households) and, thereby, isolate the effects of the treatment (in this case electrification) on lighting, farm productivity, education achieved, incomes, health measures and other indicators of household welfare. Such analysis requires rich data sets that contain observations that get the treatment and those that don’t. This overall approach is called quasi-experimental design. Consider assessing the impact of drip irrigation on farm productivity. The researcher would properly apply the treatment (drip irrigation) to one part of the field, leave the other as rain fed and observe the difference in yields. Assessing the impact of rural electrification or any other project is no different. To properly asses the impact of the project, the survey should include households in the Project Area as well as those in a control group outside the Project Area. For the present purposes, this could be done during the second fielding of this survey, after electrification. It would suffice to survey public institutions, enterprises and households that are similar to those in the Project Area. Energy use and expenditures by members of such a comparison group could then be compared to similar members within the Project Area using statistical methods to evaluate the impacts of electrification. Survey instrument The questionnaire served its purpose well—to characterize expenditures on and use of substitutable fuels. For the second round, however, several modules could be strengthened that will serve to enhance project evaluation. Some of the suggestions below may already be contained in the village-level survey, but most are particular to each household. It has been found that highest levels of education achieved in the household, basic health indicators for household members, the type of housing structure and the quality of water supply or sanitation are all primary indicators of household welfare. Including some basic questions on these measures could be very helpful in evaluating the differential impacts of the project, particularly when trying to explain which households connected and which ones did not. In addition, it is important to ask distances to certain facilities, such as the closest school, health clinic, water supply, major road, market for selling farm goods water supply, and source for collecting wood fuels, if any. More generally, the survey would benefit from asking a more detailed set of questions about the sources and consumption of wood fuels and agricultural residues. Even though these fuels are not substituted by electricity, it would allow a fuller examination of family resources (including time) devoted to meeting energy needs overall. Finally, if economic impacts of lower cost irrigation power are a high priority to the project team, it would be prudent to ask detailed questions about farm holdings, livestock, Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 15 of 24 land under cultivation, land irrigated and yields from each harvest. In all fairness, such a farm and irrigation productivity study is usually well beyond the scope of even a thorough household energy survey, so the project team should consider this only if it is a high priority. Sample frame Eighteen villages were selected to be surveyed out of the 59 to be electrified. It is unclear whether these 18 villages were selected randomly or purposively. If not randomly, data from these villages may not be representative of the Project Area as a whole. Even if the 18 villages surveyed are, in fact, representative of all 59 villages, the sample was stratified by income groups. It is unclear whether the shares of sampled households in each income group reflect the proper population proportions. On other surveys, it has been observed that when enumerators are allowed to select households, they tend to be clustered near each other as that is more convenient to interview. The enumerators’ manual did, however, caution against this. Finally and as noted above, households with generators and possibly those with irrigation pump sets appear to have been over-sampled. While there may be good reasons for over- sampling a particular group (to make sure there are enough observations to properly characterize their energy use), it is important to set them aside so that they do not bias the general evaluation of the overall project. All of these potential problems are fairly common. Moreover, they can all be resolved. At this pre-electrification stage, no attempt was made to generalize findings to the Project Area as a whole. The indicative findings reported here appear to be sufficiently robust to overcome most concerns with the sample frame. But attention should be paid to these matters in the second round when a more formal evaluation of project impacts on public institutions, enterprises and households will likely be more important. The Bottom Line The pre-electrification survey and research design served its purpose well. The current level of substitutable expenditures in sampled households are more than enough to indicate a strong ability to pay existing tariff rates for electricity. The magnitude of benefits, in terms of increased lighting and reduced costs, from electrification at the household level have been illustrated. Moreover, households that currently use electricity from community or private generators and those that irrigate with non-electric pump sets would likely be able to cut costs substantially if grid electricity were made available at the national tariff. Options Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 16 of 24 The observations and suggestions above aim to strengthen the research design, survey instrument and sample frame to be employed during the second round, after electrification. Making at least some of these modifications will improve the ability of the monitoring and evaluation team to complete a formal evaluation of the impacts of the Asmara Power Distribution and RE Project. 1. Default option. Re-survey the same villages and households after electrification— compare conditions before and after electrification and attribute changes to electrification. This would be the least-cost approach. It may be satisfactory if the original sample frame can be said to be representative of the population in Project Area. It would be a within treatment design—no control group. The evaluation team may be able to identify systematic differences between households that connect to the grid and those that do not. But it will not be able to evaluate project impacts generally, i.e., it won’t be able to identify substantive differences between electrified and non-electrified villages and households that are otherwise similar. At any rate, the team should consider making some of the modifications to the survey instrument mentioned above. 2. Default option with added control group. Re-survey the same villages and households after electrification—but expand the sample frame to include a control group. As with the default option, this is premised on the original sample frame being representative of the population in Project Area. The advantage of this option is that the control group would allow a more formal evaluation of impacts of the project itself on villages and households, rather than just identifying differences between households that connected and those that did not. As above, the team should consider strengthening/adding survey sections on housing type (material), education, water supply and sanitation, health, biomass fuels, land cultivated and under irrigation, and distances to water supply, schools, clinics, major roads, and markets. These measures have been found to be important in explaining the impacts of electrification in other countries. 3. Bootstrapping option. Re-survey the same villages and households after electrification—but supplement the pre-electrification survey by administering it to a freshly drawn control group this summer. Such a control group should be drawn from outside the Project Area but contain otherwise similar villages and households, including some that have already been electrified. When combined with data from the pre- electrification survey, data from such a control group may allow a more formal evaluation of the anticipated impacts of the project itself. Of course, the viability of this option also requires that the existing sample represents the population in the Project Area. This option may be most costly as it would involve fielding the survey again, before the second round. If this approach is taken, the questionnaire should be modified as above. 4. Post-electrification evaluation. If the current sample is not representative of the population in the Project Area and cannot be weighted to make it so, the impact of the project may be assessed after electrification by drawing a fresh random sample, in a dozen or so project villages and half as many villages outside the project area that have Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 17 of 24 been electrified for some time, and enumerating the same questionnaire, modified as suggested above. If this option is select, the team may still choose to re-survey the same villages and households after electrification. But formal evaluation of impacts due to the project would be done on the basis of the random or stratified random samples drawn from inside and from outside of the Project Area. RECOMMENDATIONS The default option with added control group (Option 2.) is the recommended approach in the case that the original sample frame is representative of the population in Project Area. Post-electrification evaluation (Option 4.) is the recommended approach otherwise. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 18 of 24 APPENDIX ANNUAL HOUSEHOLD INCOME (US Dollars/year) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 1,960 78 958 94 361 202 844 374 Income/cap 288 151 69 145 Farm income 1,071 80 436 96 142 210 407 386 Non-farm income 955 78 547 95 239 205 464 378 Wage income 175 78 244 95 126 206 166 379 Trading 613 78 233 95 59 208 216 381 Remittances 67 78 16 95 6 210 21 383 Suwa or injera 1 78 27 95 21 208 18 381 Home business 11 78 1 95 6 208 6 381 Government aid 144 79 170 97 155 216 157 392 Other 42 78 26 96 23 207 27 381 Asmara RE Project pre-electrification survey 2005. HOUSEHOLD EXPENDITURES (US Dollars/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 163.36 78 79.81 94 30.05 202 70.36 374 Farm income 89.21 80 36.31 96 11.82 210 33.95 386 Monthly expense 128.68 79 61.05 96 48.58 213 67.97 388 Food 51.87 80 31.04 97 25.02 214 32.01 391 Clothing 9.90 80 7.91 97 5.28 214 6.87 391 Schooling & health 5.26 80 3.25 97 2.54 214 3.27 391 Firewood & biomass 3.38 80 1.98 97 1.93 213 2.24 390 Commercial fuel 36.27 79 10.06 97 7.51 216 13.94 392 Travel 20.56 80 4.98 97 3.04 216 7.08 393 Rent 1.92 80 1.34 97 .71 214 1.11 391 Other 10.32 80 1.67 96 2.40 214 3.85 390 Asmara RE Project pre-electrification survey 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 19 of 24 ENERGY EXPENDITURES (US Dollars/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 163.36 78 79.81 94 30.05 202 70.36 374 Farm income 89.21 80 36.31 96 11.82 210 33.95 386 Monthly expense 128.68 79 61.05 96 48.58 213 67.97 388 Kerosene for cooking, etc. 3.79 80 2.50 97 1.66 216 2.30 393 Lighting & battery expense 6.69 80 5.45 97 4.97 216 5.44 393 Kerosene for lighting 2.98 80 2.27 97 2.59 216 2.59 393 Candles .42 80 .27 97 .23 216 .28 393 Dry cell batt 3.12 80 2.82 97 2.13 216 2.50 393 Storage batt .18 80 .09 97 .02 216 .07 393 HH electric expense 5.57 80 .32 97 .52 216 1.50 393 Irrigation expense 40.54 80 3.49 97 .14 216 9.19 393 Asmara RE Project pre-electrification survey 2005. ENERGY EXPENDITURES Users Only (US Dollars/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N Income 163.36 78 79.81 94 30.05 202 70.36 374 Farm income 89.21 80 36.31 96 11.82 210 33.95 386 Monthly expense 128.68 79 61.05 96 48.58 213 67.97 388 light+batt cost 6.69 80 5.45 97 5.02 214 5.47 391 firewood cost 9.33 29 7.12 27 8.37 49 8.32 105 electric cost 26.20 17 5.22 6 7.04 16 15.11 39 irrigation cost 129.72 25 56.48 6 7.45 4 103.19 35 Asmara RE Project pre-electrification survey 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 20 of 24 ELECTRIC & IRRIGATION EXPENDITURES (US Dollars/month) Income Level Table Total Higher income Medium Income Lower income Mean Valid N Mean Valid N Mean Valid N Mean Valid N HH electric expense 5.57 80 .32 97 .52 216 1.50 393 monthly elec bill .32 80 .13 97 .19 216 .20 393 genset oper cost 3.58 80 .18 97 .27 216 .92 393 genset maint cost 1.67 80 .01 97 .06 216 .37 393 monthly elec bill 3.21 8 2.54 5 3.23 13 3.09 26 genset oper cost 26.00 11 17.89 1 19.43 3 24.15 15 genset maint cost 12.16 11 .76 1 4.14 3 9.79 15 Irrigation expense 40.54 80 3.49 97 .14 216 9.19 393 elec for irrig cost 1.82 80 .00 97 .00 216 .37 393 irrig pump fuel cost 38.72 80 3.49 97 .14 216 8.82 393 irrig monthly cost 129.72 25 56.48 6 7.45 4 103.19 35 elec for irrig cost 48.41 3 . 0 . 0 48.41 3 irrig pump fuel cost 123.91 25 56.48 6 7.45 4 99.04 35 Asmara RE Project pre-electrification survey 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 21 of 24 ERITREA pre-electrification survey sample frame Income Level Table Total Higher income Medium Income Lower income Count Col % Count Col % Count Col % Count Col % Dekemhare Endadeko Number 4 5.0% 5 5.2% 8 3.7% 17 4.3% Godeity Number 3 3.8% 6 6.2% 13 6.0% 22 5.6% Alla Number 14 17.5% 14 3.6% Gaden Number 6 7.5% 6 6.2% 10 4.6% 22 5.6% Wekerty Number 2 2.5% 4 4.1% 10 4.6% 16 4.1% Akrur Number 3 3.8% 8 8.2% 9 4.2% 20 5.1% Degsa Number 3 3.8% 6 6.2% 15 6.9% 24 6.1% Korbaria Number 4 5.0% 6 6.2% 15 6.9% 25 6.4% Adikeih Mendefera Number 3 3.8% 4 4.1% 11 5.1% 18 4.6% Halay Number 3 3.8% 4 4.1% 12 5.6% 19 4.8% Embeito Number 3 3.8% 4 4.1% 9 4.2% 16 4.1% Keren Debresina Number 4 5.0% 6 6.2% 14 6.5% 24 6.1% Afabet Number 8 10.0% 9 9.3% 18 8.3% 35 8.9% Libana Number 5 6.3% 6 6.2% 15 6.9% 26 6.6% Barentu Tokombia Number 5 6.3% 8 8.2% 20 9.3% 33 8.4% Bimbina Number 4 5.0% 6 6.2% 12 5.6% 22 5.6% Mogolo Number 4 5.0% 5 5.2% 15 6.9% 24 6.1% Areda 2 2.5% 4 4.1% 10 4.6% 16 4.1% Table Total Number 80 100.0% 97 100.0% 216 100.0% 393 100.0% Asmara RE Project pre-electrification survey 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 22 of 24 ERITREA pre-electrification survey sample weights Income Level Table Total Higher income Medium Income Lower income Count Col % Count Col % Count Col % Count Col % Dekemhare Endadeko Number 54 1.9% 68 1.9% 108 1.4% 230 1.6% Godeity Number 39 1.4% 79 2.3% 170 2.2% 288 2.1% Alla Number 312 10.8% 312 2.2% Gaden Number 68 2.3% 68 1.9% 113 1.5% 249 1.8% Wekerty Number 31 1.1% 62 1.8% 155 2.0% 248 1.8% Akrur Number 90 3.1% 240 6.9% 270 3.6% 600 4.3% Degsa Number 58 2.0% 115 3.3% 288 3.8% 460 3.3% Korbaria Number 96 3.3% 144 4.1% 360 4.7% 600 4.3% Adikeih Mendefera Number 38 1.3% 51 1.5% 141 1.9% 230 1.6% Halay Number 38 1.3% 51 1.5% 152 2.0% 240 1.7% Embeito Number 20 .7% 26 .7% 59 .8% 104 .7% Keren Debresina Number 73 2.5% 109 3.1% 255 3.4% 437 3.1% Afabet Number 1196 41.3% 1346 38.6% 2691 35.4% 5233 37.4% Libana Number 265 9.1% 318 9.1% 795 10.5% 1378 9.9% Barentu Tokombia Number 271 9.4% 434 12.5% 1085 14.3% 1791 12.8% Bimbina Number 75 2.6% 112 3.2% 224 2.9% 410 2.9% Mogolo Number 120 4.1% 150 4.3% 449 5.9% 719 5.1% Areda 56 1.9% 113 3.2% 281 3.7% 450 3.2% Table Total Number 2899 100.0% 3484 100.0% 7596 100.0% 13979 100.0% Asmara RE Project pre-electrification survey 2005. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 23 of 24 Luminous Efficacy of Standard Lighting Devices Power Light Output Watts lumens klmh/kWh Incandescent bulb 25 Watt 25 230 9.20 40 Watt 40 430 10.75 50 Watt 50 580 11.60 60 Watt 60 730 12.17 100 Watt 100 1,280 12.80 Fluorescent tube 10 Watts 10 600 60.00 20 Watts 20 1,200 60.00 40 Watts 40 1,613 40.33 Non-electric lamps klmh/liter Paraffin Candle 60 12 Kerosene Wick 118 11 0.94 Kerosene Hurricane 198 32 1.57 Kerosene Pressure 1,380 2,040 14.37 Source: Nieuwenhout, FDJ, PJNM van de Rijt, and EJ Wiggelinkhuizen, 1998, Rural Lighting Services: A comparison of lamps for domestic lighting in developing countries, Energieonderzoek Centrum, Netherlands. Asmara Power Distribution and Rural Electrification Project, Pre-electrification Survey Analysis, April 28, 2006 page 24 of 24