Mozambique - Power Tariff Study

¶1) E ESMAP Joint UNDP / World Bank Energy Sector Management Assistance Programme c/o The World Bank * 1818 H Street, N.W. * Washington, D.C. 20433 * U.S.A. CONFIDENTIAL 15150-MOZ MOZAMBIQUE POWER TARIFF STUDY Power Development, Efficiency & Household Fuels Division Industry and Energy Department The World Bank 1818 H Street, N.W. Washington, D. C. 20433 This document has restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without UNDP or World Bank authorization. Wli H ESMAP Joint UNDP / World Bank Energy Sector Management Assistance Programme clo The World Bank * 1818 H Street, N.W. * Washington, D.C. 20433 * U.S.A. MOZAMBIQUE POWER TARIFF STUDY OCTOBER 1995 Power Development, Efficiency & Household Fuels Division Industry and Energy Department The World Bank 1818 H Street, N.W. Washington, D. C. 20433 This document has restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without UNDP or World Bank authorization. Abbreviations GDP gross domestic product GW gigawatt GWh gigawatt hour IV high voltage km kilometer kV kilovolt kVA kilovolt ampere kW kilowatt kWh kilowatt hour I litre LV low voltage MT Meticais MV medium voltage MW meagawatt MWh megawatt hour Acronyms DNE Departamento Nacional de Estatistica DOE Department of Energy EDM Electricidade de Mozambique ESMAP Energy Sector Management and Assistance Program HCB Hidroelectrica de Cahora Bassa RSA Republic of South Africa Currency Equivalents US$ I = MT 6,054 (1994 average) = MT 8,000 (May 1995) US$ I = Rand 3.6 (May 1995) TABLE OF CONTENTS INTRODUCTION 1. EX EC U T IV E SU M M A R Y ........................................................................................................ 1 2. SECTOR CHARACTERISTICS AND PERFORMANCE .......................................... PHY SICAL INFRASTRUCTURE........................................................................................... G RO SS C O N SUM PTION ...........................................................................................................3 SA LES A N D L O SSES ...............................................................................................................4 F IN A N C IA L SITUATION ..........................................................................................................4 3. PRESENT TARIFFS AND COST OF SUPPLY ....................................................................... 6 T A R IF F S ................................................................................................................................6 SU PPLIES FRO M H C B ............................................................................................................ 7 IM PO RTS FROM E SKOM ............................................................................................E...M.8... - 8 P R O SP E TSE.. .. ...C. ... ..... ..TS. .. ..... ..... ..........................................-......-...-..-.................8...............-8 4. D EM A N D FOR EC A STS .........................................................................................................10 5. IN V ESTM EN T AN D LRA IC ..................................................................................................13 PLANN ED INVESTM ENTS .....................................................................................................13 L R A IC ................................................................................................................................18 6. PROPOSALS FOR A NEW TARIFF SCHEME.....................................................................20 A N N E X 1: K E Y D A T A ................................................................................................................23 ANNEX 2: EDM'S TARIFF SYSTEM.........................................................................................3 ANNEX 3: MINIMIZING THE COSTS OF ELECTRICITY IMPORTS FROM ESKOM........35 ANNEX 4: LOAD ANALYSIS SOUTHERN SYSTEM .............................................................39 ANNEX 5: ECONOMETRIC ANALYSIS OF ELECTRICITY DEMAND ...............................43 ANNEX 6: FORECAST OF ELECTRICITY CONSUMPTION.................................................51 ANNEX 7: LONG-RUN AVERAGE INCREMENTAL COSTS OF POWER SUPPLY ..........57 ANNEX 8: BASIC FEATURES OF A REVISED TARIFF SYSTEM........................................72 INTRODUCTION The present study is a continuation of efforts undertaken by the World Bank and ESMAP to assist Mozambique in its efforts to define and implement appropriate policies in the energy sector which is of crucial importance for the welfare and development of the country. This study is cofinanced with funds from the Netherlands and Sweden. The process of this study also contains an element of institutional development as the team responsible for carrying it out proposes to train staff of EDM's Planning Department to carry on this work in the future by providing the appropriate software and some training in its use. This study was prepared by Michel Del Buono (Senior Economist, IENPD) and Witold Teplitz Sembitzky (Energy Economist. Consultant) following a mission to Mozambique that took place in May/June 1995. Comments were received from John Besant-Jones. Robin Bates (IENPD), and Robert Bacon (Consultant, IENPD). Jose Lopez provided some estimates of possible cost savings on the EDM investment program. Manuel Ruas of the Ministry of Mineral Resources and Energy provided information and critical thinking. EDM's Planning Department provided access to data and participated enthusiastically in all aspects of the preparatory work. The authors are grateful for all these contributions but wish to stress that any remaining errors are their responsibility alone 1. EXECUTIVE SUMMARY 1.1 This study suggests a revision of the level and structure of electricity tariffs in Mozambique. The present tariff regime. which has been implemented in 1992, does not reflect the costs of meeting future demand, is over-complicated, and gives wrong signals to consumers. 1.2 In order to arrive at a more efficient tariff system, the study critically reviews EDM's investment program in the light of forecast electricity consumption and peak load. The forecasts are obtained by estimating a demand function and applying the model to different scenarios about GDP growth and tariff increases. The two scenarios selected for investment planning are the case of low GDP growth cum high tariffs and that of medium GDP growth combined with low tariffs. 1.3 This Study's review of EDM's planned investments concludes that they are overly ambitious. too expensive, and difficult to finance and, therefore. need to be trimmed. The study puts forward a revised sequence and volume of investments which, while compatible with forecast demand, are substantially less costly than those proposed by EDM. The revised program is used to compute the long-run average incremental costs of meeting future demand under the two scenarios. 1.4 Depending on the scenarios, total incremental costs range from 6 to 9.1 UScents/kWh.. At the generation (actually, import or purchase) end, under either scenario, incremental costs are about 2.8 UScents/kWh, which is low by regional and international standards. The balance is accounted for by transmission and distribution costs which make-up two-thirds of the planned investments. 1.5 The study recommends to base the new tariff system on long-run incremental costs associated with the medium growth scenarios. The proposed rates are composed of single two-part tariffs (capacity fee plus energy charge) at the HV- and MV-level. At the LV-level, the schedules distinguish between residential and non-residential users, with some of the capacity costs subsumed under the energy charge. 1.6 The tariffs calculated in this study are different from those arrived at in the recently completed financial restructuring study of EDM. They differ because economic analysis does not take into account past (irreversible, sunk) costs which are very much part of financial costs; future costs, however, will need to be accounted for and, even in this case, there are differences between the investment program proposed by EDM and that which is assumed in this study (see pp. 20 ), which imply fairly large cuts and delays in investments, in an effort to keep down costs. For all these reasons, long-run incremental or marginal costs are lower that the financial costs estimated in the recently completed financial restructuring study. -2- 2. SECTOR CHARACTERISTICS AND PERFORMANCE 2.1 In the past, when Mozambique's power sector was perilously exposed to the risks of civil unrest and sabotage, sector development was dictated by the need to keep the system running, more or less without regard to costs and substantially dependent on the concessionaire funds that generous donors provided. As a consequence, most of the investments that were not for replacement and repair, were sunk into projects based on illusory expectations about the return of peace and economic progress and implemented with inappropriately high design and reliability standards. Moreover, with the advent of peace in 1992, EDM not only encountered the challenge of having to graduate from an engineering task force to a commercially-oriented utility, but was saddled with a number of damaged and/or imprudently acquired assets that are costly to repair and maintain, and difficult to operate within a system, thus resulting in a huge financial burden reflecting basically sunk costs. Physical Infrastructure 2.2 EDM's transmission network consists of three independently operated systems (see Map---). The northern grid is tied to a 220 kV line from Songo (Cahora Bassa) to Nampula (1,000 km), extending at 110 kV to Nacala (200 km), designed as a double circuit between Songo and Caia. There is also an incomplete line from Matambo to Chibata which was planned to connect the northern grid with the central system. A weak, temporary interconnection is expected to be completed in late 1995. 2.3 The central system, which was repeatedly sabotaged during the war, comprises about 500 km of 110 kV lines and 113 km of 66 kV lines. It stretches from the Zimbabwe border to the coastal city of Beira. Of the two parallel lines between Mavuzi and Nhamatanda, one has been reactivated recently. The double circuit connection between Nhamatanda and Beira can only be operated as a single circuit system. 2.4 The southern system is composed of a 225 kV line, running from Maputo to Komatipoort (RSA), a 110 kV line coming from the RSA to Corumana and extending to Maputo, and a 110 kV line connecting Maputo with the towns of Chokwe and Xai-Xai through Macia. 2.5 EDM's installed generating capacity amounts to 310 MW. Currently available are about 206 MW, of which 85 MW are hydro, mainly in the central region. The southern system accounts for 206 MW, while the central and northern systems are endowed with 82 MW and 20 MW, respectively (for details, see Table 11, Annexl). There are also a number of isolated units, notably in Pemba, Angoche and Lichinga (North), and in Inhambane (South). In addition to its own capacity, EDM has contractual access to 200 MW of hydro power from Cahora Bassa, currently at very low tariffs, and has the capacity to import about 150 MW from South Africa, also at a very advantageous price. Hidroelectrica de Cahora Bassa (HCB) is a private Mozambican corporation that owns the Cahora Bassa dam, power plant and direct-current line to South Africa. HCB, in turn, is owned by the States of Portugal (82%) and Mozambique (18%). Gross Consumption 2.6 Even though EDM's own available generating capacity would be sufficient to meet most of the domestic load, about 78% of the power supplied in 1994 came from the RSA or was acc i from HCB, with the balance covered by indigenous hydro resources and a few thermal pla operated in isolated areas. For comparison, the share of gross consumption' matched by imports and HCB-acquisitions was only 59% i 1991. The sharp increase in imports (southern system) and HCB-supplies (northern system), which cost less than the operation of EDM's thermal plant, became feasible through improvements in the reliability of the network brought about by the return of peace in 1992. 2.7 Between 1990 and 1994, aggregate gross consumption rose by 5 % a year. Growth was fastest in the southern system with an average rate of 6%, compared to 3.5% in the central plus northern region (able 1, Annex 1). As a consequence. the share of gross consumption accounted for by the southern system increased from 61% in 1990 to 65% in the first quarter of 1995 (compared to 23.3% in the central region and 11.7% in the northern region). While the peaceful conditions prevailing since 1992 were conducive to faster demand growth in the northern (7.4%) and southern systems (6.7%), consumption in the central region remained stagnant during the last two years. 2.8 Currently, the system peak is about 162 MW, compared to 144 MW in 1990. The daily peak tends to shift from about 7 p.m. during the wet season (April-October) to 12 a.m. in the dry season (see Annex 6).There is a deep trough in demand, at night, year round. 2.9 Another change that can be attributed to the end of civil strife is a sharp increase in the system's load factor. During the 1980s, the load factor on average was well below 0.6. Since 1992, however, the figure steadily improved to about 0.67 in early 1995, reflecting the greater reliability of supply achieved by EDM.' 2.10 It is also worth noting in this connection that the load factor moved up in spite of the diminishing share of medium and high voltage consumption (see Table 2, Annex 1). By contrast, the number of residential users as well as their specific consumption has steadily increased during the last five years (except for 1992), and a similar trend is observable for nonresidential small- Gross consumption includes technical and nontechnical losses as well as station use. 2 In 1994, a cyclone caused heavy damage to the network in the northern towns of Nampula and Nacala, resulting in a temporary disruption of supply. Since there is no information available on the coincidence of subsystem peaks, estimates of system wide load factors include some margin of error. In February 1995, the load factor of the central system was in the vicinity of 0.62, compared to 0.68 in the southern system. -4- volun. onsumers (see Table 10, Annex 1).' By early 1995, residential and nonresidential small- volume customers accounted for 53% of final consumption, followed by medium-voltage custon -s with a share of 39.7%, while high-voltage consumption was down to 1.4% (see Table 7, Anne-, Sale 7d Losses 2.11 While the data on gross supply are fairly reliable, the records of electricity sales give a hi5 distorted picture of the level of final consumption and the losses involved in the tran Lon and distribution of power. The sales figures published in EDM's annual reports are label _illed energy ("energia facturada") and reflect what EDM thinks it delivers to final cons :s. A thorough investigation of EDM's accounts in 1994 shows, however, that there is a larg - between invoiced energy and actual (recorded) sales: Out of a gross supply of 948 GWh, 727 1 were considered to be delivered, but only 596 GWh qualified as recorded sales. Hence, tech and nontechnical losses amounted to 33.7% of gross consumption. 2. 1.L Armed with this insight. EDM's financial department has prepared a 1995 forecast that closely approximates the composition of losses facing the utility (see Table 5, Annex 1). Acec to this breakdown, technical losses account for 14% of gross consumption. while 17.2% are r up by nontechnical losses (including public lighting). Moreover, EDM expects to collect only of the revenues accruing from its forecast sales; this would result in overall losses of 40%. 2.13 About 40% of the nontechnical losses are assumed to be theft (e.g. tampering of meter 6% are ascribed to substandard metering equipment, and 18% to false meter reading. As a consu ace, EDM gives top priority to a program aimed at reducing the nontechnical losses by two-, .s over the next three years. Fine -1 Situation 2.14 While the large proportion of non-technical losses has been contributing to EDM's fina, difficulties, it is not the only area of concern. Another source of financial difficulty has been excessive level of investment and the needlessly high technical standards and quality and relia y of service that are not justified by the present uses of electric power. This excessive inve ..:ent is destined to continue in the future, as evidenced by EDM'S ambitious investment proi i. Unlike past investments, future ones will affect incremental costs and hence, tariffs. In sho: .DM'S financial losses stem from basically three sources: High level of losses, especially non-technical; High, not always prudent investment; and s argued below, however, the level of sales reported in the past is subject to measurement errors. While this is not likely to affect the trends underlying the published data on electricity use by consumer group, care should be taken in interpreting the absolute figures. The depreciation of the metrical which has caused large increases in the domestic currency equivalent of obligations in foreign exchange (such as debt service; imported energy, equipment and spare parts; and technical assistance). 2.15 EDM has begun to act on the first problem and will surely obtain some measure of success. The second problem will not be acted upon as it seems to be part of a pohi :al decision to continue rapid investments in electrification. The third problem is beyond EDN tpower but is bound to ameliorate as the economy and macroeconomic policies and variables stabilize. Together with high tariffs, improvements on these three fronts cold help improve EDM's financial situation. However, to the extent that HCB tariffs also are expected to rise, EDM's financial situation may not improve substantially. Consideration must then be given to the possibility of writing-off sunk costs provided that the same process of excessive investment does not persist. However, vigorous enforcement of timely payments by customers in addition to sharply rising tariffs is bound to have a serious impact on household budgets and the reaction of electricity users to tariff levels never experienced before cannot be predicted. To gain an insight in tis process, ESMAP has proposed to carry out a brief survey designed to estimate consumer ability/willingness to pay such tariffs within the limits of present household budgets and to canvass their attitudes towards various forms of energy. -6- 3. PRESENT TARIFFS AND COST OF SUPPLY Tariffs 3.1 With a few exceptions. the rate-setting approach underlying the currently prevailing tariff system closely adheres to recommendations that Coopers and Lybrand made in a tariff study submitted in 1991. The proposals were implemented in January 1992 and have since been adjusted for inflation, but not in structure.' 3.2 A detailed examination of EDM's tariff regime is provided in Annex 2. Its major shortcomings can be summarized as follows: (a) The estimated average costs used as a benchmark for fixing the initial tariff levels did not properly reflect the future costs of meeting (growing) demand. Adjustments in nominal tariffs which have since taken place failed to correct this weakness. (b) The tariffs overcharge for energy, undercharge for capacity, and do not properly account for network losses across different voltage levels. (c) The tariff system is rife with options that are not only economically meaningless, but tend to aggravate the distortions created by its flawed structure. (d) At each voltage level, the tariffs cause cross-subsidies from consumers with a high load factor to those with a lower load factor. (e) The system subjects low-voltage customers (< 19.8 kVA) to discriminatory capacity charges which have no economic justification. 3.3 In particular, HV-, MV- and LV-customers with a maximum demand in excess of 19.8 kVA are given the choice among three two-part tariffs involving a trade-off between payments for energy and capacity.' The schedule is designed to induce consumers to select a lower energy rate associated with a higher capacity charge if the savings from cheaper energy more than offset the additional capacity costs. The main effect of the optional tariffs is, however, that users with a low load factor are charged below estimated average cost while those with a high load factor pay more. It may even happen that the total revenues generated by the scheme fall short of the estimated costs of supply.' The most recent increase in nominal tariffs was on August 1, 1995. 6 The schedule is composed of a short-utilization, medium-utilization, and long-utilization two-part tariff, whereby long-utilization users are those with a load factor of about 0.15 or more. Note that "estimated average costs" have been computed by Coopers and Lybrand and do not reflect economic costs. -7- 3.4 To make things worse, EDM offers advice on how customers should choose among tariff options that are useless in the first place. Moreover, the optional tariffs, while overcomplicating the tariff selection and billing process, have limited practical relevance: There are virtually no short-utilization users. At the HV-levei. 211 customers fall into the long-utilization category. And about 97% of the LV-customers (< 19.8 kVA), which account for less than 6% of total electricity sales, choose the medium-utilizatior. .Oriff. 3.5 Similar arguments apply to the discriminatory capacity charges designed for residential and small-volume nonresidential users. The schedule comprises nine different flat rates for capacity, which, on a per unit basis, increase with the level of maximum demand.! There is, however, no convincing reason why unit capacity costs should rise in direct proportion to the level of load served. Also, five of the capacity charges are tailored to consumption profiles which represent less than 10% of the customers affected by the tariff scheme. Finally, electricity bills are prepared with carefully computed capacity and energy charges even for.the snallest consumers when this doesn t matter as EDM, can neither control, nor meter capacity utilization by individuai consumers. 3.6 Finally, it should be mentioned that while EDM theoretically is entitled to adjust nominal tariffs for inflation and exchange rate depreciation, its latitude to do so has been limited by the need to seek political approval. In the last three years, nominal tariffs were usually frozen over a period of four to six months before EDM could catch up with the general price increases, thus exposing the company to unnecessary financial losses. 3.7 In conclusion, EDM's situation did not improve when it set the tariffs along the lines proposed by Coopers and Lybrand. That tariff system does not internalize the costs of meeting future demand, is overcomplicated, and gives wrong signals to consumers. Rather than tinkering with the tariffs on the basis of revalued assets in order to meet EDM's financial needs, -which seems to be EDM's major concern at the moment -, the system should be overhauled under the guidance of sound economic principles. Proposals for a tariff reform are presented in Chapter 5 of this Study. Supplies from HCB 3.8 EDM is entitled to call upon 200 MW from Cahora Bassa (HCB).' Currently, about 33 MW are injected into the northern grid, while imports from Eskom (by way of a bilateral contract between EDM and Eskom) substitute for the supplies that HCB is unable to deliver to the southern system as long as the line to Apollo (RSA) is out of service. Since there are no load limiters installed, maximum demand charges are assigned in accordance with the level of consumption. 9 Strictly speaking, EDM's share is 220 MW at the Cahora Bassa busbar or 190 MW at Komatipoort (RSA). -8- 3.9 The rates EDM pays for HCB power supplied to the northern system were agreed upon in 1983 and, as they have remained fixed in nominal terms. have since been eroding through inflation. As a result, EDM's northern system obtains HCP power almost free of charge."o 3.10 For the southern system, the rates negotiated in the early 1980s for HCP power wheeled through the RSA are somewhat higher (via a tripartite agreement among HCB, EDM and Eskom which assigns all of Cahora Bassa's output to Eskom, except for a 200 MW entitlement to EDM). Under the old terms, and assuming a load factor of 0.66, EDM would pay about 10.6 Rand/MWh plus transmission fees. In 1989. however, Eskom indicated that it would be prepared to pay a higher price once it has access to HCB-power (for details, see Annex 7).Since the tripartite agreement states that EDM must pay at least the same price as Eskom for HCB power, at these higher rates, EDM's bill would be about 28 RandiMWh (or UScents 0.8 per kWh as of May 1995) plus transmission fees. Imports from Eskom 3.11 As a substitute for HCB power that EDM cannot now obtain, the southern system imports from Eskom. The latest agreement between Eskom and EDM came into effect in January 1995 and will be valid until July, 2000. or three years after the resumption of supply from HCB to Eskom (for details, see Annex 3). 3.12 By international standards, the terms of the contract are advantageous. The rates, which are denominated in US$ and indexed with respect to US-producer prices, currently amount to 0.5 UScents/kWh for firm energy, 3.24 UScents/kWh for emergency energy (i.e. in excess of contracted capacity), and US$8.44 per kW of contracted capacity, charged on a monthly basis. 3.13 Also, the way the rates are structured makes it profitable for EDM to contract less capacity than needed to serve expected maximum demand, and to cover the balance with emergency energy. EDM has quickly realized that there is a potential for cost savings and should be commended for pursuing the correct strategy of reducing the amount of contracted capacity during the first four months of 1995. On average, it paid only UScents 1.97 per kWh imported from Eskom, which is close to the theoretical minimum of 1.76 UScents/kWh." Prospects 3.14 Since HCB is committed and determined to bring the line from Cahora Bassa to the RSA back into operation by March 1997, EDM can count on the availability of HCB-power in the southern system from that time onwards. However, one would be on less safe grounds in assuming that the rates HCB is going to charge will remain at the low level that EDM's northern system 10 EDM pays 4 Rand per MW times the duration (number of hours) of the respective month. For instance, if the load factor is 0.55, EDM pays 7.23 Rand/MWh (= 4/0.55), which in May 1995 was equivalent to 2 USS/MWh. " The theoretical minimum can only be achieved if EDM correctly predicts the southern system's monthly load (duration) curve, which is unlikely (see more details, see Annex 3). -9.- enjoyed in the past. The politically appealing expectation that EDM will enter an era of cheap power is unfounded. 3.15 Reactivating the high-voltage line to Apollo and resuming large-scale power generation at Cahora Bassa has the objective to generate revenues for the repayment of HCB's accumulated debt, not to mention the investment costs that were sunk at the outset. Clearly, with rates of 1 UScent/kWh or less (which South Africa said it might be prepared to pay), this would be a hopeless undertaking. So HCB is going to press for higher rates, and once the line to the RSA is operational, HCB's leverage to do so will be much greater than today. In any case, Escom should be willing to pay higher rates, so long as they remain below its short-term generation cost. 3.16 Indicative of what future rates may look like is Eskom's offer of 1989, which at that time was in the vicinity of 1.5 UScents/kWh. Another benchmark is the recent import arrangement between EDM and Eskom. resulting in an average rate of about 2 UScents/kWh. 3.17 Also relevant in this connection is that EDM expects to have access to additional HCB-power (i.e., >200 MW) when needed. This is only feasible if Eskom releases part of its claim on HCB supplies (as it has done in favor of Zimbabwe).12 Eskom's opportunity costs of giving EDM a cut of its share, however, are the revenues it would forego by not exporting to Mozambique its own power or, what comes to the same thing, power acquired from HCB. In fact, the current agreement between Eskom and EDM leaves Mozambique the choice to continue importing power from the RSA beyond 2000. It can therefore be assumed that any imports or acquisitions that exceed EDM's current 200 MW entitlement will cost at least what EDM pays to Eskom under the present import contract. 3.18 Needless to say, power obtained at 2 to 2.5 UScents/kWh is significantly less expensive than operating EDM's thermal plant and is also an attractive alternative to investments in additional hydro or thermal (coal or gas-fired) facilities built to serve the domestic market. 3.19 In sum, it can be argued on fairly plausible grounds that HCB will seize the opportunity to renegotiate the terms of its agreement with Eskom and EDM once it is in a position to reliably supply the RSA and EDM's southern system. Also, the resumption of large-scale supplies from Cahora Bassa will buttress the trend towards a greater regionalization of power markets and, thus, the equalization of rates at which bulk supplies are traded. These forces will inexorably increase the price of tradable power beyond the level at which EDM's northern system, which geopolitically and in terms of power transactions is a niche, was privileged to buy from HCB. The new prices can be expected to rise to 2 UScents/kWh or even higher. 12 It should be also kept in mind that Zimbabwe is a potential bidder for what is left of Eskom's entitlement to HCB- power. - 10- 4. DEMAND FORECASTS 4.1 A number of electricity demand forecasts have been prepared in the last four years to lay the groundwork for EDM's investment planning and financial projections." The forecasts were based on ad-hoc assumptions or extrapolated past trends into the future. In retrospective, their accuracy was not particularly exemplary. 4.2 In fact, under the circumstances prevailing in Mozambique, forecasting can easily become a hapless exercise. Since past developments were swamped by the impacts of war and/or recording errors. and given that the recent period of negotiating and consolidating peaceful conditions has a transitional character. there is little empirical evidence upon which long-term predictions can be built. In light of these difficulties, the present Study relies on a Bayesian approach to modeling electricity demand using prior information from Mauritius (see Annex 6), and applies the coefficient estimates to different scenarios covering the period 1995-2010. 4.3 The estimated demand function is of the partial adjustment type with real GPD. inflation-adjusted average tariffs, and gross electricity consumption lagged one year, the dependent variable, as arguments. The short-run elasticity estimates obtained are 0.46 for GDP and -0.097 for price, while the long-run counterparts are 1.46 and -0.309 for GDP and price, respectively (for details, see Annex 5).4 4.3 In terms of scenarios, the cases of slow, medium, and high GDP-growth are combined with a low and a high tariff policy. The high-tariff variant assumes that average tariffs rise to 9.5 UScents/kWh by 1998, which EDM thinks is the target level it needs to reach for financial reasons. Alternatively, a lower target level of 7.5 UScents/kWh is considered. In both cases, only modest tariff increases (in real terms) are expected for the period after 2000. 4.4 Regarding the prospects for GDP growth, the least favorable case can be compressed into the assumption that real GDP rises at an average rate of 3% a year. Medium growth translates into average annual growth rates of 4.7% for 1995-2000 and 4.2% for 2000-2010. High growth corresponds to 6.7% a year for 1995-2000 and 5.5% for 2000-2010 (for details, see Annex 6). Forecasts have been conducted by Norconsult, EDF, KfW, and, most recently, by EDM's financial department. 14 While the long-run GDP elasticity may appear to be on the higher end, in the context of the estimated coefficients adjustment is slow so that the long-run tends to be remote into the future and its elasticity only is approached asymptotically. - II - 4.5 The resulting forecasts are presented in Table 1 below. Table 1: Forecasts of Gross Electricity Consumption (GWh) Year Fla Fib F2a F2b F3a F3b 1995 935.03 935.03 942.58 942.58 949.64 949.64 1996 962.38 968.80 984.21 990.77 1003.22 1009.92 1997 982.90 999.57 1021.89 1039.22 1057.80 1075.74 1998 1000.52 1035.64 1058.56 1095.71 1117.08 1156.29 1999 1028.89 1077.99 1109.06 1161.98 1198.81 1256.01 2000 1065.50 1125.64 1170.97 1237.07 1300.75 1374.18 2001 1108.77 1177.77 1243.01 1320.37 1414.76 1502.80 2002 1158.64 1235.36 1327.14 1415.02 1539.26 1641.18 2003 1211.49 1295.03 1417.64 1515.39 1675.08 1790.58 2004 1266.23 1355.62 1510.27 1616.89 1817.69 1946.01 2005 1323.02 1417.91 1605.85 1721.04 1965.94 2106.95 2006 1382.04 1482.24 1705.29 1828.92 2121.44 2275.24 2007 1442.25 1547.25 1807.76 1939.36 2283.85 2450.11 2008 1505.34 1615.24 1915.74 2055.60 2456.71 2636.06 2009 1566.54 1680.78 2023.50 2171.06 2639.00 2831.43 2010 1628.02 1746.93 2133.94 2289.79 2827.83 3034.36 ARG2/: 1995-2010 4.0 4.5 5.9 6.4 7.9 8.4 1995-2000 2.6 3.8 4.2 5.4 6.4 7.6 1/ Fla= low GDP. high tariff: F1b= low GDP, low tariff. F2a= medium GDP, high tariff: F2b medium GDP, low tariff; F3a = high GDP, high tariff, F3b = high GDP, low tariff. 21 Average annual rate of growth (%); least squares estimates. 4.6 Depending on the scenario assumed, gross electricity consumption is predicted to grow at average annual rates ranging from 4.0% to 8.4%. For comparison. in the last three years. consumption grew at about 5.5%. Beyond 2000, forecast consumption tends to increase faster than in the period 1995-2000, mainly because the dampening effect of the tariff adjustments tapers off, while GDP continues to rise, thus having a marked positive impact on demand in the longer term. 4.7 In addition to the long-term projections shown above, a short-term forecast of monthly gross electricity consumption has been conducted on the basis of time series analysis (see Annex 6). The cumulative estimates are 950 GWh for 1995 and 993 GWh for 1996, compared to 899 GWh in 1994. The long-term predictions that come closest to these figures are those made under the medium-growth-cum-low-tariffs scenario. 4.8 Since the data base is too weak to estimate a (nonlinear) peak load function, maximum demand is assumed to be linear in electricity demand. Table 2 shows the forecast peaks - 12 - conditional on forecast gross consumption, assuming that the system load factor remains in the vicinity of 0.66." Table 2: Forecasts of System Peak (MW)" Year Fla Flb F2a F2b F3a F3b 1995 161.7 161.7 163.0 163.0 164.3 164.3 1996 166.5 167.6 170.2 171.4 173.5 174.7 1997 170.0 172.9 176.7 179.7 183.0 186.1 1998 173.1 179.1 183.1 189.5 193.2 200.0 1999 178.0 186.5 191.8 201.0 207.3 217.2 2000 184.3 194.7 202.5 214.0 225.0 237.7 2001 191.8 203.7 215.0 228.4 244.7 259.9 2002 200.4 213.7 229.5 244.7 266.2 283.9 2003 209.5 224.0 245.2 262.1 289.7 309.7 2004 219.0 234.5 261.2 279.7 314.4 336.6 2005 228.8 245.2 277.8 297.7 340.0 364.4 2006 239.0 256.4 295.0 316.3 366.9 393.5 2007 249.5 267.6 312.7 335.4 395.0 423.8 2008 260.4 279.4 331.4 355.5 424.9 455.9 2009 271.0 290.7 350.0 375.5 456.4 489.7 2010 281.6 302.2 369.1 396.0 489.1 524.8 1/ Fla = low GDP, high tariff; Fib = low GDP, low tariff, F2a = medium GDP, high tariff; F2b = medium GDP, low tariff: F3a = high GDP, high tariff: F3b = high GDP, low tariff. 4.9 There is no foolproof recipe for selecting among the different demand scenarios. The approach recommended by this study is to restrict the focus to a forecast interval within which electricity consumption is expected to develop. As a lower band, we choose the low-growth-cum- high-tariffs scenario since its predictions are close to EDM's forecast of May 1995. On the other hand, it can be argued that the scenarios considered are biased towards the proposition that good must prevail, thus disregarding the possibility of long-lasting stagnation. let alone a severe recession. To make up for this bias, we drop the rosy scenarios based on high GDP-growth and define the medium-growth-cum-low-tariffs scenario as the upper band. 4.10 Regarding the distribution of gross electricity consumption between EDM's southern system and the rest of the country it is assumed that the share accounted for by the southern system is 65% in 1995, declines to 62.5% by 2000, and remains at this level until 2010. Since the systems will be interconnected by then, this will not matter very much. 4.11 Needless to say, the forecasts as well as the underlying models should be refined and updated once additional information about the relevant variables becomes available. ' Even though a load factor of 0.66 looks rather high, it reflects the current situation. On the other hand, assuming that the load factor will decrease in the future implies that the forecasted peaks would be even higher than is shown in Table 2. - 13 - 5. INVESTMENT AND LRAIC Planned Investments 5.1 The investment program EDM plans to launch over the next 15 years is rather ambitious. About US$ 350 million will be needed to implement the projects. Given that EDM can supply most of its generation needs from HCB or Eskom. most of the US$ 320 million investment program is earmarked to rehabilitate and extend the transmission and distribution network." More importantly, about two-thirds of the investments are scheduled for the period 1995-2000, and many projects appear to be over designed and expensive judged by the design and engineering standards prevailing in similar countries. 5.2 The program reflects policy promises made by the new government. Policy makers have pledged to increase the number of residential connections by 5000 a year and to extend the grid to remote areas. In response to these commitments, EDM distinguishes between commercial projects on the one hand and. on the other. semi-commercial and non-commercial projects. The latter are supposed to generate less revenues than is needed to recoup investment costs, while the financial returns from ventures labeled non-commercial may not even cover operating costs.17 5.3 From an economic viewpoint, however, the above distinction is irrelevant. If a project is economic in the first place (i.e., shadow-valued benefits exceed costs), it should be undertaken. Also, once the facilities are in place, it is never economic to charge less than it costs to operate them, but it may be desirable to service the debt through income transfers rather than through revenues collected from the direct beneficiaries of the facilities. What EDM seems to have in mind, though, is that investments in so-called non-commercial or semi-commercial projects should be financed in a non-commercial way, preferably through grants (which might also need to cover operating losses). 5.4 Yet the point made by EDM is moot since EDM does not identify the projects that it thinks need concessionaire funding. Nor does it identify the sources where these funds are supposed to come from. Besides, even if cheap (untied) finance were available on larger scale, the opportunity costs of assigning new resources to power, namely the foregone (possibly sizable) benefits from alternative uses, would still need to be considered.'" As a consequence, the economic cost of raising and/or using the funds required by EDM's investment program (including operating 16 See EDM's "Highlights 1994/95", presented in Maputo during a donor conference in June 1995. 1 EDM claims, however, that many of the non-commercial projects are least-cost or even economically viable on account of external benefits. 18 Of the committed concessionaire funds, about US$ 56 million are allocated to ongoing projects that will completed soon and, thus. do not contribute to EDM's incremental investments (i.e. those that need to be considered for tariff purposes). Concessionaire funds under negotiation may affect the planned I10 kV extensions to Inhambane (about USS 25 million) and Gurue (about US$ 5 million) and minor telecommunications projects. - 14- profits) will be captured through a uniform annual discount rate of 10%, and this rate also would apply to the benefits measured in kWh. 5.5 Moreover. even though EDM needs to some extent to deliver on the promises made by policy makers. the program should still meet future demand at least-cost subject to political or other constraints. Yet the program is excessive and far from a sequence of projects that keep costs at a reasonably low level. The only screening EDM provides is a breakdown into first-priority projects which are scheduled for the period 1996-2000, and the remaining projects which are given second-priority because EDM has no choice but to postpone them, reluctantly it seems, to the period 2001-2005. 5.6 It is not the task of this study to devise a least-cost power sector expansion plan that lives up to the standards of sophisticated optimization models. What is feasible, though, is to streamline EDM's investment program against the background of forecast demand so that incremental costs are kept at a level warranted on economic grounds (i.e. the demand to be served; for details. see Appendix 7). 5.7 To this end, we have removed all projects that do not contribute to serving domestic demand and/or improving the reliability of supply. As a result, the only regional projects considered in the revised expansion plan (apart from the HVDC line to Apollo) are the 275 kV line connecting Zombodze (Swaziland) with Matola, which will increase the southern system's import capacity to 350 MW by 2000, and the proposed interconnection between Orange Grove (Zimbabwe) and Chibata, assuming that the works start in 2006 rather than in 2001. Regarding the interconnection between the northern and central system it is assumed that the 220 kV line between Matambo and Chibata/Xiagadora will come on stream by 2001 (in addition to the temporary 110 kV interconnection which may happen much sooner). 5.8 Existing thermal generation plant. which in some locations needs to be overhauled, is used as stand-by. unless it supplies isolated areas until they are connected to the grid. The firm hydro capacity available in the northern-central system is estimated at 50 MW for 2500 hours a year. The Corumana hydro plant is rated at 12 MW for 2500 hours a year and is now operated to shave the peak load of the southern system. Finally, HCB-power can be wheeled to the southern system from early 1997 onwards. However, HCB supplies could be shifted between the three subsystems and EDM's hydro capacity could be used to lower system costs depending on the type of tariff faced by EDM. 5.9 A comparison of the above supply options with the two demand forecasts reveals (see Table 1, 2, 3, and 4, Annex 7) that EDM must seek additional sources of supply either in 2002 (medium demand growth) or in 2007 (low demand growth). Basically, EDM could resort to imports from Eskom (which would be feasible by extending the current import agreement), or from other parties (Southern Africa will, by then be entirely interconnected), count on the possibility of increasing its share of HCB generation, or expand its own generation capacity. It is most likely that - 15 - one of the first three options will be available (at about 2 UScents/kWh) and, therefore the fourth option is too costly and is dismissed as a concrete possibility in this study. " 5.10 Another problem that needs to be addressed is that the risk of line outages is relatively high in the northern-central region. While the planned extensions to Pemba and Lichinga will increase this risk, investments in rehabilitating and reinforcing the grid tend to reduce it. A project to rehabilitate and strengthen the Center-North Line is currently being carried out and will, hopefully, improve line reliability. While additional insurance against disruptions in supply could be bought by investing in backup generation plant. the economics of this option depend on the probability of transmission failures and the value of lost load." Unfortunately, the likelihood of future line outages is not known; nor do we know what different consumer groups are, or will be, willing to pay for greater reliability than that provided by the grid. However, a sizable group of consumers with a comparatively high value of lost load is more likely to emerge under the medium- growth scenario than under slow growth. As a consequence, this Study's revised investment program endorses the planned investment in a 25 MW gas turbine in Nacala in 2002 in the case of medium growth. but not with slow growth. 5.11 In addition, this study suggests to trim the planned investments in transmission and distribution facilities. In some cases, the EDM cost estimates are revised downwards to account for savings than can be achieved by lowering the design standards. In other cases, the planned investments are postponed or stretched over a longer period than is assumed by EDM (for details, see Annex 7). These changes notwithstanding, the revised program considers all proposed/committed projects, except those aimed at power exports. and is consistent with forecast demand. Table 3 provides a summary of EDM's investment program and the changes proposed by this Study. 19 In particular, during the planning horizon there is no need for investing into a hydropower station at Alto Malema. Nor is there any merit in considering a coal-fired plant near Moatize because its generation costs have been estimated at 4-6 US cents per kWh. 20 For instance, if the probability of line outages were equal to 220 hours a year, it would economic to provide those customers with a backup who are willing to pay a premium of US$ 0.80 per expected kWh of lost load (see Annex 7). Some generation along or at the end of a long line also may be required for system stability. - 16- Table 3: Summarv of EDM's and Revised Investment Program (USS million) 00 0 - 3.a 96-5 1.25 - Toa T7 20.0 - - 20.0 9.0 5.0 - 14.0 T 15.0 - - 15.0 6.0 1.0 - 7.0 T4 - 4.0 - 4.0 3.0 - - 3.0 T5 11.4 - - 11.4 11.4 - - 11.4 T6 - 3.0 - 3.0 0.75 1.25 - 2.0 T 10.0 - - 10.0 6.5 6 12.5 Tg 4.0 -- 4.0 2.0 2.0 4.0 T9 20.0 - - 20.0 - 20.0 - 20.0 T10 25.0 - - 25.0 20.0 - - 20.0 Til - 19.0 - 19.0 - 20.0 - 20.0 T12 -4.0 - 4.0 - - 4.0 4.0 T13 - 4.0 - 4.0 - - 4.0 4.0 7.4 10.0 - - 10.0 - 1 - - TIS - 6.0 - 6.0 - - 6.0 6.0 S- 15.0 - 15.0 - - - - 7.0 - - 7.0 7.0 1.0 - 2.0 8.0 - - 8.0 7.0 1.0 8.0 7.0 - - 7.0 3.0 1.0 - 4.0 6.0 - - 6.7 5 0 - 6.2 7.0 - - 7.0 1.0 2.0 - 3.0 212 8.0 - - 8.0 1.9 0.2 - . 7.0 - - 7.0 7.2 - - 7.2 0 3.0 - - 3.0 0 4.0 - 5.0 9 13.7 - - 3.7 - 6.0 - 6.0 6 9.2 - - 9.2 4.5 1.5 - 6.0 1 27.0 - - 27.0 8.5 9.5 - 18.0 .12 8.0 - - 8.0 8.3 - - 8.3 2 -6 7.0 - 7.0 - 7 - 7.0 0 - 0 - 5.0 - 5 - 5.0 - 2.0 - 2.0 - 2.0 - 2.0 D40 4.0 - 4.0 - 4.0 -.7.0 - 7.0 - 7.0 - 7.0 1. - - L.0 1.0 - - 1.0 0.6 --0.6 0.6 -- 0.6 3- - - 3-0 3.0 - - 3.0 - - 110 110 - --- 12.0 --12.0 -16.0 - 16.0 3 30 --3.0 1.5 1.5 - 3.0 20 --2.0 2.1 -- 2.1 S3.0 --3.0 1.5 1.5 -3.0 C33.0 --3.0 3.0 - - 3.0 C4 5.9 - - 5.9 4.6 1.3 - 5.9 Total 250.8 80 110 440.8 131.25 127.05 14 272.3 Source: EDM and mission estimates. - 17- GI = thermal rehabilitation G3 = Alto Malema Hydro, 80 MW G4 = gas turbine Nacala G5 = Chicamba dam rehabilitation TI = 220 kV, rehab Songo-Nampula. split into Stage I and Stage 2 T3 = 110 kV, replacement Nampula-Nacala T4 = 110 kV, rehab of 2nd line Mavuzi-Nhamatanda T5 = 245 kV, Swaziland (Zombodze)-Mozambique (Matola) T6 = reactive power devices Beira T7 = 110 kV + 220 kV, interconnection HCB-EDM (at Chibata) T8 = 110 kV, Alto Molocue-Gurue + substation Gurue T9 = 110 kV, Nampula-Ancuabe-Montepues-Pemba T10 = 110 kV, XaiXai-Inhambane TI = 110 kV, Gurue-Cuamba-Lichinga T12 = 110 kV Corumana-Ximavane T13 = 110 kV, Massingir-Chokwe T14 = 220 kV. Mozambique (Matambo)-Malawi (Blantyre) TI5 = 110 kV Caia-Luabo-Marromeu DI = rehab and extension of Nampula distribution system D2 = rehab and extension of Beira distribution system D3 = rehab and extension of Quelimane distribution system D4 = rehab and extension of Nacala distribution system D5 = rehab and extension of Maputo distribution system D6 = power supply to Buzi D7 = rehab XaiXai, Mampula. Nacala substations D8 = extension Maputo substations SE4, SE5, SE6 D9 = primary and secondary distribution Gurue, Cuamba, Lichinge D10 = rehab electrification Chimoio and Chokwe D1I =rehab electrification south D12 = Maputo substation extension D 13 = overhaul of Matola distribution system D14 = overhaul distribution system XaiXai and Chokwe D15 = overhaul of Angoche distribution system DI 6 = conversion from diesel to electric pumping in Limpopo D17 = new substation Matola D18 = new substation in Tete and Matundo El = urban household energy E2 = emergency program Nacala Cl = rehab control system central region C2 = remote control center Maputo C3 = Telecommunication Phase II C4 = Cahora Bassa study C5 = vehicles, equipment, EDM headquarters building - 18 - 5.12 It should be noted that only one project of the revised program is conditional on demand, namely the gas turbine proposed for Nacala. The reason why investments in transmission and distribution have not be tailored to the different profiles of forecast demand is that we consider the medium-growth scenario at least as likely as the low-growth scenario. With this assumption. the minimum-regret strategy would be to plan as though the system will follow a medium-growth path, and adjust the program downwards if demand happens to increase at a lower rate, or to bring forward investments when demand grows faster than expected. On the other hand, going ahead with the more costly and less flexible program proposed by EDM would only make sense if high growth were more likely to occur than medium growth, a very risky assumption which this Study does not recommend.22 LRAIC 5.13 Table 4 shows the long-run average incremental costs (LRAIC) implied by the revised investment program (for fU.nher detaiis. see Appendix 7). Note that the cost estimates do not account for non-technical losses. As expected. LRAIC vary considerably in direct proportion to the rate at which demand is expected to rise. Under the slow-growth-cum-high tariffs scenario. total average incremental costs amount to 9.13 UScents/kWh. which is roughly in line with the tariff assumption underlying projected demand (EDM's target level of 9.5 UScents/kWh). With medium growth and lower tariffs (7.5 UScents/kWh targeted for 1997), total incremental costs are about 6 UScents/kWh. 5.14 At the generation end, the unit cost differential proves insignificant. This is because the slightly higher costs incurred under the medium growth forecast (Nacala gas turbine) are distributed over larger volumes of consumption. With about 2.8 UScents/kWh, incremental generation costs are fairly low, reflecting the fact that we expect EDM to have access to cheap power from Eskom, HCB. or other interconnected parties. 5.15 On the other hand, there is a sizable gap between the scenario-dependent average incremental costs pertaining to transmission and distribution (5.97 vs. 3.0 UScents/kWh, including expenditures on O+M). This indicates how sensitive average costs are with respect to demand. Also, the high level of network-related average incremental costs, particularly in the case of slow growth, sheds light on the risk of over investing in these facilities. Needless to say, this risk would increase considerably if EDM were to embark on its more ambitious investment program. 22 In particular, the large and irreversible upfront investments called for by EDM's expansion program would saddle the utility with a huge financial burden, leaving little scope for adjustments if demand proves less buoyant than has been hoped, thus resulting in excess capacity and high unit costs. - 19 - Table 4: LRAIC of Power Supplyu Investment Program and LRAIC Low Growth. High Tariffs Medium Growth. Low Tariffs Generation: UScentstkWh 2.795 2.803 Transmission: US$/kW/year 164.93 82.81 UScentsikWh 2.853 1.432 Distribution: US$/kW/year 150.68 75.66 UScents/kWh 2.606 1.308 Total T+D: US$/kW/year 315.61 158.47 UScentsikWh 5.459 2.740 O+M: Transmission Distribution 0.202 0.101 (UScentsikWh) 0.307 0.154 Miscellaneous: UScents/kWh 0.367 0.184 1/ In prices of 1995. 2/ Based on investment costs and expenditures for fuel, maintenance, etc. Source: Mission estimates - 20 - 6. PROPOSALS FOR A NEW TARIFF SCHEME 6.1 In reforming EDM's tariff system, the guiding principle should be to inform consumers of the level and structure of costs they impose on the system. Moreover, tariffs should account for load characteristics, be simple and transparent, easy to implement, and fair. 6.2 As a rule, electricity tariffs should be forward-looking, i.e., based on what it costs to meet future demand. The yardstick referred to in this study is average incremental costs rather than marginal costs. In fact. with the present pattern of load and with hydropower and imports as the main sources of supply, there is no need for establishing a complex tariff system based on marginal costs (see Annex 8). At generation, average costs are not significantly different from marginal costs. Nor are the daily or seasonal changes in load significant enough to justify even simple versions of time-of-use pricing. It is therefore recommended to focus on standard two-part tariffs built on average incremental costs differentiated by voltage-level. Any simple form of peak-load pricing could easily be introduced if the situation at some point warrants it and. given that there is a deep trough at night, year round in the daily load schedule ,a change to a specific, off peak tariff could be implemented as soon as some users emerge that would consume reasonably large amounts of electricity during those hours. 6.3 Moreover, no particularly convincing case can be made for regionally discriminatory tariffs. Once the northern and central systems are interconnected and HCB-power is wheeled to the southern region via the RSA, the average costs of bulk supplies tend to become fairly uniform across the networks. There will be location differences in transmission losses, notably along the line through the northern region, but little would be gained from designing site- specific transmission charges. However, the true cost at any given point along the line should be known so that, in case an investment in some electricity-intensive activity were considered, the decision on location can be made optimally to minimize costs to the system. What provides a stronger argument in favor of regionally distinct tariffs is the fact that about two-thirds of EDM's planned investments are assigned to the northern and central region. But making this point is probably not worth the political quarrel that it will create. 6.4 The basic structure of the proposed scheme is discussed in Annex 8. Table 5 below compares the voltage-specific capacity charges and energy rates implied by EDM's present tariff regime with the estimates obtained under the low-growth and medium-growth scenarios. The figures show that the tariffs in place grossly underestimate capacity costs and that this bias increases in indirect proportion to the voltage level.23 This conclusion continues to hold after the most recent tariff adjustments of August 1995 which led to a 25% rate increase on average. - 21 - Table 5: Comparison of Capacity and Energy Cost Estimates Cooper & Lybrand"EDM LRAIC (low growth) LRAIC (medium growth) Voltage Capacity Energy Capacity Energy Capacity Energy Level (S/kW/month) (S/kWh) (S/kW/month) ($/kWh) (5/kW/month) ($/kWh) HV 8.33 0.018 22.61 0.0188 15.32 0.0157 MV 7.73 0.020 27.78 0.0194 18.03 0.0162 LV 5.98 0.021 39.14 0.0242 24.08 0.0191 1/ Updated Cooper & Lybrand estimate as of May 1995. currently in force at EDM. Source: Mission estimates. 6.5 Table 6 gives an idea of how the proposed rates would translate into (notional) voltage-specific average tariffs. The estimates are based on preliminary assumptions about the composition of demand and typical load factors by voltage level. They suggest that system average tariffs vary by about 3.5 UScentsikWh depending on the demand scenario. Also. LV-consumers would pay between I and 2 UScents more (per kWh) than the average. It should be noted that these tariffs are different from those calculated in the recent Coopers & Lybrand financial restructuring study of EDM. That is because LRAIC do not take into account past or "sunk' costs, and because the investment program adopted is more modest and slower than EDM's. Table 6: Notional Average Costs by Voltage Level and Scenario" LF' Low Growth Medium Growth Generation/Point of Injection 0.66 3.15 2.94 HV 0.65 6.65 4.79 MV 0.58 8.50 5.88 LV 0.54 12.35 8.02 System Average3' 10.43 6.95 1/ In UScents/kWh, this table differs from table 4 because miscellaneous investment costs have been lumped with generation costs. 2/ Assumed average load factor. 3/ Demand is assumed to be composed of 8% HV, 38% MV, and 54% LV. Source: Mission estimates. 6.6 At the LV-level, the distinction between residential and non-residential consumers should be maintained. In the same vein, the new system should continue to provide a (cross-) subsidized schedule for low-income (small-volume) residential users. It is suggested that the subsidy be granted through a discount on the capacity fee, while charging a uniform rate for energy. The subsidy to small residential users would be financed through a slightly higher energy charge - 22 - for all other LV users. Also, part of the low-voltage capacity costs should be rolled over onto the energy rate to make the low voltage tariffs more socially and politically palatable. 6.7 Prior to these refinements. though, the level of tariffs needs to be fixed, thus deciding which scenario is most relevant for the purpose of rate setting. The point that can be made in favor of the low-growth-scenario is that predicted gross consumption is close to EDM's most recent demand forecast and, more importantly, that both the tariff assumptions underlying this scenario and the revenue requirements it generates are in line with EDM's financial targets. On the other hand, betting on the low-growth scenario is tantamount to assuming that Mozambique fails to make palpable economic progress. With real GDP growing at 3% a year (which is what the scenario assumes), there would be no scope for increases in per-capita income. It would only justify increases in electricity tariffs. 6.8 This study takes the more optimistic view and suggests that EDM build a revised tariff system on the medium-growth scenario. The resulting rates are summarized in Table 7. Table 7: Revised Tariff Regime Fixed Charge Energy Charge (USS/kW/month) (UScents/kWh) HV 15.30 1.57 MV 18.00 1.62 LV - Residential 7.50 6.25 - Non-Residential 14.00 4.40 Social Tariff" 1.50 6.25 1/ Applies to small-volume residential users with a maximum demand not exceeding I kW. 6.9 HV and MV-rates are set in accordance with LRAIC and no longer offer choice options. The proposed refinements at the LV-level include: * a two-part schedule for residential users, with almost 70% of the capacity costs subsumed under the energy charge; * a two-part schedule for non-residential users, with about 42% of the capacity costs subsumed under the energy rate; * a social tariff that applies to small-volume residential users with a maximum demand of I kW or less and subject to the installation of a load limiter (or an inexpensive meter, e.g. Chinese mechanical meters). - 23 - ANNEX 1: KEY DATA Table 1: Gross Electricity Consumption (MWh), 1985-95 Gross Cons. System Gross Gross Cons. Central and Peak Year Generation Imports Acquisit. Consump. South North (MW) 1985 264,400 229,000 51,900 545,300 na na 118 1986 216,300 304,700 38,100 559,100 na na na 1987 268,800 330,200 25,700 624,700 na na na 1988 263,100 340.700 40,500 644,300 452,500 191,800 na 1989 342,200 307,100 92,500 741,800 457,00 284,80n na 1990 322,200 321,800 94.200 738.200 455,100 283,100 144 1991 325.700 373,300 98.700 797,700 501.900 295,800 na 1992 273,800 436,200 95,000 805.000 505,600 299,600 150 1993 223,600 510,900 118,100 852,600 538,200 314,400 na 1994 200,000 559,000 140,000 899,000 574,584 324,416 na 19951 170,936 607,291 169,657 947,874 615,291 332,593 1612 1/ EDM forecast. 2/ Estimate based on first quarter. Source: EDM Table 2: Billed Electricity By Tariff Category (MWh), 1988-94 Year General Residential Medium - High Total Voltage 1988 78,900 180,800 264,100 526,000 1989 90,100 203,000 284,500 591,100 1990 93,100 224,400 276,900 595,300 1991 126,000 257,700 311,000 694,700 1992 133,230 252,222 291,738 678,702 1993 133,207 285,439 265,626 692,617 1994 151,713 317,766 258,125 727,604 Source: EDM - 24- Table 3a: Monthly Gross Electricity Consumption (MWh), 1989-95 1989 1990 1991 1992 1993 1994 1995 January 59122 54104 6467 62440 68618 73698 79200 February 52264 55100 63283 70438 60937 71369 73157 March 63446 62179 68593 69871 71431 78659 82636 April 59757 50127 70606 66752 70295 72397 May 63112 66035 74116 66042 72927 74889 June 61719 62919 58104 67942 68692 71501 July 66153 67132 76955 65156 72013 73359 August 57700 67682 58627 66814 71239 76237 September 61040 64294 65274 62975 71048 74910 October 58995 68231 67624 67284 75009 75980 November 58739 62004 64887 63544 74143 76270 December 61357 57807 65622 67651 76227 78938 Source: EDM Table 3b: Monthly Gross Electricity Consumption (MWh), Moving 12 Month Average, January 1990 - March 1995 1990 1991 1992 1993 1994 1995 January 60283.7 61467.8 66530.2 66409.1 71048.2 74850.6 February 59865.5 62348.4 66344.3 66923.9 71471.6 75309.1 March 60101.8 63030.3 66940.5 66132.2 72340.9 75458.1 April 59996.2 63564.8 67047.0 66262.2 72943.3 May 59193.8 65271.4 66725.8 66557.4 73118.4 June 59437.3 65944.8 66053.0 67131.2 73281.9 July 59537.3 65543.6 66872.8 67193.7 73516.0 August 59618.9 66362.2 65889.6 67765.1 73628.2 September 60450.7 65607.6 66571.8 68133.8 74044.7 October 60721.9 65689.2 66380.3 68806.6 74366.5 November 61491.6 65638.7 66351.9 69450.3 74447.4 December 61763.7 65878.9 66240.0 70333.6 74624.7 - 25 - Table 4: Annual Peak by Operational Area (MW) Area 1985 1990 1992 1993 Pemba 1.3 1.9 1.9 2.6 Lichinga 0.8 1.2 0.9 1.0 Cuamba na 0.4 0.4 0.5 Nacala 2.6 6.0 4.6 4.9 Nampula 4.2 5.5 6.0 6.3 Angoche 0.6 0.6 0.7 0.7 Gurue 0.1 na na na Mocuba 0.6 0.7 0.6 0.8 Quelimane 2.9 3.5 na 4.4 Tete 4.1 4.7 4.3 3.0 Chimoio 15.6 17.0 17.8 18.0 Beira 12.6 14.9 15.4 15.6 Inhambane 0.9 1.3 1.3 1.4 Lionde 1.9 3.2 3.5 4.9 Xai Xai 1.7 2.2 2.4 2.9 Maputo 66.2 86.0 92.0 90.5 Source: EDM Table 5: EDM Sales Forecast 1995 System (MWh) % of Gross Supply Thermal Generation 40,135 - Station Use 1,876 = Therm.Energy Sent Out 38,259 4.0 + Imports 607,291 64.2 + Acquistions 130,801 13.8 + Hydro 169,657 17.9 Gross Supply 945,998 100.0 - Transmission Losses 78,197 8.3 - Distribution Losses 48,046 5.1 - Internal Consumption 4,952 0.5 = Net Supply 814,803 86.1 - Non-technical Losses 149,187 15.8 - Public Lighting 13,792 1.5 = Sales 651,824 68.9 Source: EDM - 26 - Table 6: Electricity Sales First Quarter 1995 (MWh) South Central North System Therm.Generation 2,057 620 5,907 8,584 -Station Use 51 361 569 981 Thermal Energy 2,006 259 5,338 7,603 Sent Out + Imports 150,399 150,393 + Acquistytions 35,037 35,037 Hydro 42,161 42,161 = Gross Supply 152.399 42,420 40,375 235,194 - Transm. Losses 5,724 2,758 7,719 16,201 - Distrib.Losses 7,335. 2,658 3,648 13,641 - Intern.Consum. 846 321 504 1,671 = Net Supply 138.494 36.683 28,504 303,681 - Non-techn.Losses 32.891 5.750 2,457 41,098 - Public Lighting 1,890 626 756 3,272 =Sales 103.713 30,307 25,291 159,311 Table 7: EDM Sales Forecast By Consumer Group (MWh), 1995 Year % Actual Ist Quarter % Domestic 243,034 37.3 61,360 38.4 Geral 85,813 13.2 23,347 14.6 Low Voltage 38.345 5.9 10.318 6.5 - Short 922 0.1 139 0.1 - Medium 35,200 5.4 9,410 5.9 - Long 2,223 0.3 769 0.5 Medium Voltage 264,548 40.6 62,485 39.1 - Short 9,209 1.4 1,274 0.8 - Medium 118,051 18.1 30,333 19.0 - Long 137,299 21.1 30,878 19.3 High Voltage 20,089 3.1 2,272 1.4 - Short - - Medium - - Long 20,089 3.1 2,272 1.4 Total 651,824 100.0 159,782 100.0 Source: EDM - 27 - Table 8: Index of Real GDP and Electricity Tariffs, 1988-94" Year GDP CPI AT RAT 1988 77.89 11.00 23.75 215.90 1989 81.59 15.95 36.55 229.10 1990 83.11 22.33 63.9 286.10 1991 84.91 29.81 90.05 302.00 1992 81.31 43.12 177.20 410.90 1993 94.88 61.5 235.30 382.60 1994 100.00 100.00 347.00 347.00 1/ GDP = index of real GDP (1994=100); CPI = consumer price index (1994=100); AT = average tariff revenues (MT/kWh); RAT = inflation-adjusted average tariff revenues Source: EDM and DNE. Table 9: Quarterly Data on Gross Electricity Consumption and Tariff Revenues (1/1992-1/1995) Quarter GC 110 CPI AT RAT 1/92 202749 100.0 100.0 132 132.0 2192 200736 106.8 112.1 140 124.9 3/92 194945 113.3 111.4 145 130.2 4/92 198479 106.8 126.2 209 165.6 1/93 200986 81.0 145.0 216 149.0 2/93 211914 78.5 154.3 199 129.0 3/93 214300 88.3 160.7 233 145.0 4/93 225379 88.2 179.8 261 145.2 1/94 223726 57.8 224.2 315 140.5 2/94 218787 64.5 243.7 320 131.3 3/94 224506 89.6 272.4 355 130.3 4/94 231183 84.3 303.7 351 115.6 1/95 234993 46.2 344.5 460 133.5 GC = gross electricity consumption (MWh) 110 = index of industrial output CPI = consumer price index AT = average tariff revenues (MT/kWh) RAT = real average tariff revenues Source: EDM and DNE. -28- Table 10: Key Data on Consumer Groups" Number of Connections Specific Consumption2 Year General Residential Medium and General Residential High Voltage 1990 17,560 97,620 822 5,131 2,079 1990 17,997 100,028 882 5,183 2,223 1991 19,615 107,206 882 6,574 2,404 1992 19,656 111,891 865 6,676 2,254 1993 20,509 115,070 746 6,495 2,481 1994 20,464 118,957 934 7.413 2,671 1/ Classified by tariff. 2/ kWh per connection. Source: EDM. - 29 - Table 11: Available Generating Capacity Station Type Installed Cap. Available Cap. Retirement (MW) (MW) North Pemba Diesel 6.4 5.5 > 2010 Lichinga Diesel 1.3 0.6 2010 Nacala Diesel 21.5 3.7 > 2010 Quelimane Diesel 7.1 6.0 > 2010 Nampula Diesel 6.4 1.41 > 2010 Mocuba Diesel 0.8 0.7 2005 Angoche Diesel 1.6 1.0 > 2010 Lichinga Hydro 0.7 0.6 > 2010 Cuamba Hydro 1.1 1.0 > 2010 Central Mavuzi Hydro 52.0 36.0 > 2010 Chicamba Hydro 38.4 34.0' > 2010 Beira Gas (Jet) 12.0 12.0 > 2010 South 1.15MW by 1998, Inhambane Diesel 3.9 1.6 0.45MW>2010 Lionde Diesel 4.0 2.7 2000 Xai-Xai Diesel 2.7 1.7 1998 Maputo Coal 57.5 20.0 2000 Gas' 78.5 64.0 41MW by 2005, 23MW > 2010 Corurnana Hydro 14.5 14.0 ' > 2010 1/ Upon completion of rehabilitation works, about 5.5 MW will be available by 1998. 2/ When Chicamba and Mavuzi are run simultaneausly, the joint firm capacity is assumed to be 50 MW at 3400 hours a year (170 Gwh/year). 3/ The firm capacity is assumed to be 12 MW at 2500 hours a year (30 GWh/year). Source: EDM and mission estimates. - 30 - ANNEX 2: EDM'S TARIFF SYSTEM The structure of the current tariff system, which was introduced in January 1992, looks rather complex. Tariffs are uniform across the country. but differ by voltage level (kV), maximum demand (kVA or the kW-equivalent), and customer class. At each voltage level, customers with a contracted capacity of more than 19.8 kVA are offered three optional two-part tariffs. Medium- and high-voltage users are entitled to choose between billing periods lasting from 7 a.m. to 9 p.m. (system peak) and from 9 p.m. to 7 a.m. (off peak). There is a special low-tension schedule composed of increasing block charges for contracted capacity up to 19.8 kVA, with a uniform energy rate for residential and non-residential users. And small-volume consumers ( 30 kWh/month) are offered a social tariff ("tarifa social") with a 75% discount on the capacity charge. As per the new law enacted in 1991, EDM is permitted to adjust the level of average tariffs for inflatio'(CPI) and exchange rate depreciation (ED). Indexation of nominal tariffs (T) is defined on a moithly basis by 'the following formula: T CPI ED 0.3 +0.7 - T CPI ED In practice, however, EDM's leeway to raise tariffs has been limited by the need to seek political approval. During the last three years, nominal tariffs were regularly frozen between four to six months before EDM was allowed to catch up with inflation. As a consequence, changes in real tariffs showed a marked cyclical pattern with no drift. That is, by early 1995 inflation- 2 adjusted tariffs were at the same level that had prevailed in early 1992 (see Table 9, Annex 1). Voltage levels are defined as: Low Voltage (LV): LV 1 kV; Medium Voltage (MV): I kV < MV 45 kV; High Voltage (HV): HV > 45 kV. Low voltage users with a maximim demand up to 39.6 kVA are charged for contracted capacity. Customers with a maximum load in excess of 39.6 kVA are subject to a monthly capacity charge. The capacity (in kW) billed is determined as K = CC - 0.8*[CC-ML], where CC denotes the contracted capacity and ML is the recorded maximum load with a duration of at least 15 minutes. Decreto do Conselho de Ministros No. 32/91. The structure of the tariff system is based on a study prepared by Coopers & Lybrand. At the outset the nominal rates charged by EDM were exactly those proposed by the study. 2 The nominal tariffs referred to in this section were valid until July, 1995. The latest increase, which was on August 1, 1995, is not reflected in the numerical examples presented below. -31 - Table 1: Optional Tariffs by Voltage Level" Short Utilization Medium Utilization Long Utilization HIGH VOLTAGE MT/kW/month 30,331 36,941 43,700 MT/kWh 384 259 199 Optimal utilization region3 X 52.88 52.88 X 112.65 X 112.65 (hours/month) Design utilization rate4' 150 250 400 (hours/month) MEDIUM VOLTAGE MT/kW/month 30.331 36,941 43,700 MT/kWh 4108 279 206 Optimal utilization region3 X 51.24 51.24 X 92.59 X 92.59 (hoursimonth) Design utilization rate 125 200 350 (hours/month) LOW VOLTAGE2 MT/kW/month 30,331 36,941 43,700 MT/kWh 429 299 224 Optimal utilization region3' X 50.85 50.85 X 90.12 X 90.12 (hours/month) Design utilization rate4' 100 150 300 (hours/month) 1/ As of May 1, 1995. 2/ With a contracted capacity of more than 19.8 kVA. 3/ The optimal utilization region is implied by the rate structure and, thus, changes when nominal tariff adjustments alter the rate structure. In the past, though, these changes were marginal. 4/ Assumed utilization rate when the tariff system was designed. Source: EDM and mission estimates. As is shown in Table 1, low-, medium-, and high voltage consumers are given the option to choose among different two-part tariffs. The idea underlying the schedule is that consumers supplied at a given voltage level differ in terms of demand (i.e., in terms of the rate at which contracted capacity is utilized). Its effect is that consumers have an incentive to opt for a longer- utilization tariff if their load factor is such that the gains from a lower energy charge more than offset the higher costs of capacity (compared to a shorter-utilization tariff, and vice versa).3 Put differently, consumers are offered the opportunity (or, as some may feel, are saddled with the task) to self-select a two-part tariff that minimizes costs relative to their expected load profile. Another feature of the scheme is that for a given customer class (defined in terms of contracted As a result, consumers with a higher load factor will enjoy a lower average tariff (provided they select the appropriate tariff option). - 32 - capacity utilization) the rates for energy increase as the voltage level decreases. Capacity charges, however, are insensitive to the voltage level. That is, higher downstream costs (network losses, etc.) are reflected by energy charges only. Basically, the rationale for properly designed self-selecting two-part tariffs is that with different consumer classes the resulting choices tend to increase welfare, at least as long as the tariff system on the whole involves deadweight losses. EDM's two-part tariffs, however, are designed in a way that aggravates rather than reduces distortions. This is because the capacity fees are fixed at essentially arbitrary levels, while the rates for energy are adjusted so as to recover notional total costs. Consumer class-specific total costs are inferred from system average capacity and energy costs, subject to a predetermined utlization rate.5 The outcome of the ad hoc assumptions underlying EDM's tariff system is that compared with estimated average costs (see Table 2) customers are overcharged for energy and u4eTP1harged for capacity. As a consequence, at each voltage level customers would pay less than estimated average costs if their actual utilization rate (load factor) fell short of the maximum design utilization rate. To put this impact into perspective. consider the forecasted sales for 1995 (see Table 7, Annex 1) and let us assume that the updated average cost estimates of Cooper & Lybrand are correct (which they are not) and that the consumers' choice of tariffs will be optimal relative to their utilization rates. Under the latter assumption. forecasted short- and medium-utilization demand should be far below the maximum tariff design level. In addition, some of the high- voltage-long-utilization consumption may not reach this level.7 Therefore, at least 94.2% of the low voltage supply (>19.8 kVA) and 48.1% of the medium voltage supply would be billed below average cost, and it is also possible that some of the long-utilization customers would pay less than is required to cover the cost of supplying them with electricity. It is even conceivable that the net- revenues from customers whose utilization rate exceeds the maximum design level, would not cover the losses generated by the shorter-utilization customers. Note that deadweight losses occur if prices depart from marginal costs. To illustrate the point, consider the two-part tariff that applies to short-utilization-high-voltage consumers: EDM's capacity and energy costs are estimated 66,631 MT/kW and 142 MT/kWh. which is an update of the original estimates submitted by Cooper & Lybrand in 1991 (19,770 MT/kW and 42.2 MT/kWh). Under the assumption that the monthly utilization rate is 150 hours, total costs amount to 87,931 MT (= 66,631 + 150*142). Setting the capacity charge at 30.331 MT/kW (9,000 MT/kW according to Cooper & Lybrand in 1991) implies that the energy charge should be set at 384 MT/kWh in order to recover total costs (30,331 + 150*384 = 87,931). 6 For instance, the two-part tariff charged to high voltage customers does not recoup estimated average costs if the actual utilization rate is less than 400 hours a month. Note that all high-voltage consumption is of the long- utlization type. Nonetheless, some of the high-voltage users who consume more than the minimum amount justifying the choice of the long-utilization tariff (112.7 hours) may have a utilization rate that falls short of the maximum tariff design level (400 hours). Of course, the same argument applies to low- and medium voltage customers falling into the long-utilization category. - 33 - Table 2: Notional Average Costs of Electricity Supply, May 1995u Capacity Energy (MT/kW/month) (MT/kWh) High Voltage 66,631 142 Medium Voltage 61,831 156 Low Voltage (>19.8 kVA) 47,851 168 Low Voltage 33,644 168 (519.8 kVA) 1/ Updated Cooper & Lybrand estimates. not to be confused with actual average costs. Source: Mission estimates. It should also be noted that the utilization rates at which customers have an incentive to switch to a longer-utilization tariff correspond to very low load factors, particularly in the case of the short-utilization tariff. So it does not come as a surprise that there are no short- and medium utilization customers at the high voltage level, and that short-utilization consumption at the low- and medium voltage level accounts for only 3% of total demand affected by optional tariffs. Small low-voltage consumers ( 19.8 kVA) are faced with two energy rates. As of May 1995, households pay 337 MT/kWh (tarifa domestica), while non-residential customers are charged 541 MT/kWh (tarifa geral). These rates are determined in the same way as the optional tariffs offered to larger users. In the case of -csidential users, the estimated capacity costs amount to 33,644 MT/kW, while energy costs are 168 MT/kWh.8 Based on a standard utlization rate of 100 hours, the monthly capacity charge corresponding to 337 MT/kWh is about 16,744 MT/kW or MT 13,. 95/kVA. According to the Cooper & Lybrand tariff study of 1991, the capacity payments of small low-voltage consumers should be linear in the amount of contracted capacity. Customers who consume 30 kWh/month or less, however, should be eligible for a 75% discount on the capacity charge rated at 1.1 kVA. Also, the study recommended the installation of load limiters so that contracted capacity becomes a binding constraint. EDM implemented the proposed social capacity fee, yet decided to introduce discriminatory capacity charges for low-voltage consumers subsumed under the "domestic" and "general" tariff. Moreover, since load limiters have not been installed to date, EDM estimates maximum demand by using the consumption level as a proxy. As is shown in Table 3, the unit capacity charge rises under the domestic and general tariff from 13,395 MT/kVA for a rated capacity of 1.1 kVA (corresponding to a consumption level of 165 kWh/month) to 28,147 MT/kVA for a maximum load of 19.8 kVA (or 2,970 kWh/month). Clearly, this kind of nonlinearity is hard to justify on the basis of costs. In fact, there is no reason why unit capacity costs should increase with the level of load served. So the These are updates of the Cooper & Lybrand estimates of 1991. -34- discriminatory capacity fees seem to be a means of generating additional revenues. Even in this context, though, block rates would only make sense if the price elasticity of demand for an additional (block of) kWh declines as consumption increases, which is unlikely. (By the same token. quantity discounts would be warranted if the price elasticity rises with the level of consumption). After all, increasing block rates are an unnecessary complication. In order to avoid losses. EDM could simply base the rates for capacity and energy on estimated average costs. Table 3: Low Voltage Capacity Paymentsu Implied % Share of Monthly Flat Rate for Capacity Consumers Max. Load Consumption Capacity Charge affected by (kVA) (kWh) (MT/month) (MT/kVA) the scheme Tarifa Social 1.1 0 - 30 3,737 3,397 36.3 Tarifa Geral and 1.1 31 - 165 14,735 13,395 24.2 Tarifa Domestica 2.2 166- 330 29,612 13,460 18.1 3.3 331 -495 56,206 17,032 12.1 6.6 496-990 127,219 19,276 4.5 9.9 991 - 1.485 212,606 21,475 4.2 13.2 1,486 - 1.980 312,800 23,697 0.3 16.5 1,981 - 2.475 427,369 25,901 0.2 19.8 2,476 - 2.970 557,319 28,147 0.1 1/ As of May 1, 1995. The rates for energy are 337 MT/kWh (domestic and social tariff) and 541 MT/kWh (general tariff). In sum. EDM's tariff regime has serious flaws. Basically, the structure of (system) costs used as a reference for computing tariffs at different voltage level has a downward bias with respect to capacity and an upward bias with respect to energy. These built-in distortions are aggravated by fact that EDM, following the advice of Cooper & Lybrand, is offering optional two- part tariffs that provide economically meaningless choices. The main effect of this schedule is that it complicates tariff selection and billing. Moreover, it establishes voltage-dependent threshold utilization rates (consumption levels) at which average tariff revenues match estimated average costs, such that consumers with a shorter (longer) utilization rate pay less (more) than it costs to serve them. Another distortionary feature of the current tariff system is that small low-voltage consumer are subjected to discriminatory capacity charges that have no economic justification. It should also be mentioned that the cost estimates used to design the system are shaky, to say the least. 9 Note that if EDM had followed the Cooper & Lybrand proposal of charging a uniform capacity fee (13,395 MT/kVA/month), the revenues generated by customers with a monthly utilization rate of less than 100 hours would not cover the estimated costs of serving these customers. So the increasing block rates may have been designed to make up for these losses. - 35 - ANNEX 3: MINIMIZING THE COSTS OF ELECTRICITY IMPORTS FROM ESKOM According to the most recent power supply agreement between Eskom and EDM signed in March 1995, electricity imports from South Africa (which are used to supply the southern system) are subject to the following rates:'0 Contracted Capacity: 8.44 US$/kW/month Firm Energy: 0.5078 UScents/kWh Emergency Energy: 3.2439 UScents/kWh (3.194 UScents/kWh until April 1995) The above rates are valid until December 31, 1995. On January I of each new year the charges are automatically adjusted in direct proportion to the annual rate of change in the US producer price index recorded during the previous year. The agreement will be in force until July 31, 2000, or three years after the resumption of supply from Cahora Bassa to Eskom. whichever is earlier. EDM's capacity requests (expected maximum demand) must be forwarded on a monthly basis not later than three working days before expiry of the previous contract. Contracts have a minimum duration of one calendar month. Capacity payments depend on contracted maximum demand rather than on the actual load profile. Instantaneous (hourly) load below or equal to the contracted capacity level is served at the firm energy rate. The emergency energy rate applies to (hourly) loads served in excess of the contracted capacity. Clearly, at the margin, electricity from South Africa is cheaper than thermal power generated with EDM plant.'2 So it would not even be economic to substitute thermal power for emergency energy from Eskom. On the other hand, it can be assumed that the marginal costs of hydro power from Corumana are below the rates Eskom charges for firm and emergency energy. As a consequence. EDM's strategy should be to operate the Corumana hydro plant and serve all the to In addition, EDM has the option to choose the tariff schedule delineated in the former agreement of May 1984 (amended on May 7, 1993). Since the beginning of this year, however, EDM has opted for the terms of the new agreement. Whenever the 275 kV line from Komatipoort to Maputo is out of service, EDM is entitled to a rebate amounting to the hourly equivalent of the monthly payments for contracted capacity times the number of hours the line is not operated. For instance. if the contracted monthly capacity is 85,000 kW at 8.44 USSikW and the line does not operate for 4 hours, the rebate is about USS 3,985 (= 8.44*85,000*4/720). On the other hand, if the 110 kV line from Komatipoort is used in lieu of the 275 kV line, the capacity payment is equal to the maximum load served during the operation of the 110 kV line, times the number of hours of the disruption, times the hourly equivalent of the monthly capacity charge. Thus, in the event that power has to be wheeled through the 110 kV line and the actual peak load falls short of the contracted maximum demand, EDM pays less than under normal conditions of supply. 12 Fuel costs are about 12.0 UScents/kWh for a gas turbine run with jet juel fuel, 7.2 UScents/kWh for a diesel plant, and 3.8 UScents/kWh for coal-thermal generation, which in either case exceeds the cost of emergency energy. - 36 - load not covered by hydro power with imports.13 In what follows we assume that the loads (including maximum demand) served by imports are adjusted for the share of indigenous hydro power. Moreover, under the new tariff agreement, EDM has an incentive to contract less capacity than would be needed to meet expected maximum demand. In fact, the optimal amount of capacity contracted from Eskom is reached when the expected savings from an additional kW of subscribed capacity (i.e.. the extra costs of emergency energy that EDM needs to import when the actual load exceeds contracted maximum demand) would be just offset by the capacity charge. The above condition can be easily derived by using the following notation: P = capacity charge (S/kW/month), cl = firm energy rate (cents/kWh), c- = emergency ener,v rate (cents/kWh). F(g) = normalized load duration function. g(F) inverse normalized load duration function. i.e. g=F , H = duration of month (hours), F = predicted load (MW), L = actual load (MW), which is stochastic, p = index of peak load, K = contracted capacity (kW), g, = g(F) for F=K. The load duration curve is normalized by setting period's length equal to unity (i.e, H=1). Since future loads are stochastic, the inverse g(F), which assumes values between zero and one, can be interpreted as the (cumulative) distribution function of L. In this connection, the probability that the actual load L exceeds the predicted load F is g(h) = P (L > F] For a given load curve, the choice of K determines the expected amount of firm energy imports El: (1) E, = Hff g(F) dF. By the same token, electricity demand that EDM expects to meet through emergency energy imports, E, is (2)E2 HFJ g (F) dF. 1 In particular. the water should be stored in the off-peak period and dispatched in the peak period, which is exactly what EDM intends to do: Corumana is scheduled to operate between 8 a.m and 10 p.m on weekdays and between 7 p.m and 9 p.m during the weekend. In addition, the plant is supposed to be run between 11 am and 12 a.m on Saturday and between 8 am and 9 a.m on Sunday. - 37 - Total expected costs of electricity imports amount to (3)C= PK+c;Ei+c2E2. The necessary condition for a cost minimum is 0 C = P + g, [c - c]H= 0, aK which can be rewritten as (4) C g C 1 Note that the right-hand-side of (4) can be interpreted as P{L>K}. Hence, condition (4) states that capacity should be contracted to the point where the probability that the contracted capacity falls short of the actual peak is equal to the ratio of the hourly capacity charge to the premium paid for emergency energy. Under the currently prevailing rates, the critical ratio is about 0.42 (=313/744)." The corresponding optimal level of contracted capacity, however, depends on the monthly load duration curve F(g) which is uncertain. It follows that EDM's key problem is to forecast F(gl). Histograms of hourly loads would be helpful in this connection. Unfortunately, the available data are incomplete. EDM has records of chronological loads for selected weeks. but there are no series of hourly loads for an entire month, let alone a year. Nevertheless, the available evidence suggests that the capacity contracted by EDM from January to April 1995 tended to exceed the level that was justified on economic grounds.'5 Yet EDM has been on the right track in reducing the level of purchased capacity as is shown in Table I below. Table 1: Electricity Imports Southern System 1995" January February March April Firm Energy 49,823 45,807 51,829 47,501 Emergency Energy 393 1,207 1,173 522 Total Energy 50,216 47,014 53,002 48,023 Contracted Capacity 87 83 83 82 Peak 99.6 102.7 98.5 97.3 Average Costs (UScents/kWh) 1.98 2.06 1.88 1.98 I1 In MWh (Energy) and MW (capacity, peak). Source: EDM ' In February the ratio would be 0.46. is Needless to say, this is an ex-post statement that neglects the impact that EDM's desire to take or to avoid risks has on the decision to buy capacity. - 38 - Based on hourly load figures for the first seven days of March 1995, we have approximated the load duration curve of March 95 through a 5th-degree polynomial. Normalization was done by setting both the peak load and the maximum number of hours equal to unity. The resulting load duration curve is referred to as f(x) rather than F(g). The estimate is f(x)= 0.978 - 0.976853x - 4.40412x'- 12.387x-+ 13.4813x - 5.01158x5 Integrating the above function over the interval [0,1] and multiplying the integral by H = 743 hours and LP = 98.5 MW gives Lp H f'f(x)dx t 52,83OMWh. which is close to the actual imports of 53.002 MWh (see Table 1). Since gjH 313 hours, the optimal capacity would have been 77 MW, compared to 83 MW contracted by EDM. The corresponding total costs (in US$) of electricity imports from Eskom work out at C(77)994,600 and C(83)=1,001.831. Thus. if EDM had contracted 77 MW rather than 83 MW, it would have saved about USD$ 7,231. Again, this calculation is possible with the benefit of hindsight. It can also be assumed that the potential for cost savings was even larger in January 1995 when EDM contracted 87 MW while the (southern) peak turned out to be 99.6 MW.16 Our estimates lead to the conclusion that EDM's decision to reduce the level of contracted capacity down to 82 MW in April 1995 was a directionally correct policy. 16 This conclusion rests on the assumption that in January the shape of the load duration curve was similar to that of March. In fact, weekly time series of hourly loads for the period March-June suggest that the normalized load duration curves lie close together in the neighborhood of g, (see Annex 4). We have therefore used an average normalized load duration f(x) to calculate the potential savings in January and February. The results indicate that if EDM had correctly predicted the monthly peaks it would have contracted 77 MW in January and 76 MW in February (on the basis of the average normalized load duration function) and, as a result, saved USS 26,000 in January and US$ 7,100 in February. -39- ANNEX 4: LOAD ANALYSIS SOUTHERN SYSTEM Daily load curves of the southern system show marked seasonal changes. Between April and October the daily peak is around 7 p.m. The morning peak occurs at noon and is significantly below its evening counterpart. Typically, the load drops between 9 a.m and 12 a.m (on weekdays) and in the afternoon (see Plot 1). By contrast, from November to March (rainy season) the daily peak tends to be at 12 a.m on a workday and at 7 p.m on weekends. The load does not fall before noon, but picks up at 3 p.m (on weekdays) and at 7 p.m. Occasionally, the evening peak may prove as high as the morning peak (see Plot 2). Based on weekly data provided by EDM, we have approximated normalized monthly load duration curves using a 5-th degree polynomial.'7 The estimates presented below suggest that the curves have a fairly similar shape. particularly for the period April-June (see also Plot 3). The average load factor is about 0.68. During the first half of 1995 the highest load was recorded on February 9 at noon (102.7 MW). Table 1: Coefficients of Normalized Load Duration Curves, March- June 1995" Month Peak (MW) Constant x x2 x3 x4 X March 96.0 0.9962 -0.94758 4.08436 -11.7165 12.9537 -4.87317 April 97.4 0.9888 -1.75006 9.3187 -24.5924 26.9083 -10.5178 May 93.5 1.0 -1.60139 6.45321 -12.9041 10.1341 -2.61081 June 92.3 0.9998 -1.62171 7.61268 -17.5073 16.0175 -5.05001 Sample 94.8 0.99646 -1.479 6.87174 -16.7261 16.5885 - 5.80825 Average 1/ 5th degree polynomial; estimates based on weekly time series of hourly loads. Source: EDM and mission estimates. Central and Northern System Unfortunately, there are no data available from the northern system. EDM, however,, provided data on hourly loads of the central system for selected days in 1994 and 1995. The figures suggest that the central region load follows a pattern similar to that observed in the southern system: On weekdays, there is a seasonal shift in the daily peak from 12 a.m (wet season) to 7 p.m (dry season). The hourly load declines after noon, rebounds at 4 p.m and reaches the evening peak at 7 p.m. During the weekend and on holidays the daily peak occurs at 7 p.m. 17 The normalized load curve is obtained by setting the peak load and the length of the period under consideration equal to unity. - 40 - According to rough estimates, in February 1995 the load factor of the central system was in the vicinity of 0.62, compared to 0.68 in the southern system. EDM records suggest that the load factor of the northern system is higher than in the central region, but below that of the southern system. Plot 1: Hourly Load Curve Southern System (MW), May 15-21 1995 -41- Plot 2: Hourly Load Curve Southern System, March 1-7 1995 - 42 - Plot 3: Normalized Monthly Load Duration Curves Southern System, March-June 1995 -43- ANNEX 5: ECONOMETRIC ANALYSIS OF ELECTRICITY DEMAND This annex presents and evaluates parameter estimates that can be used to forecast electricity demand. The estimates are obtained from causal and non-causal regression models and incorporate data provided by EDM and DNE as well as extraneous (a priori) information. Abbreviations The following abbreviations will be used: CI = condition index DF = Dickey-Fuller test df = degrees of freedom DW = Durbin-Watson statistics err = residuals F = F statistic GLS = generalized least squares IV = instrumental variables Log = natural logarithm N = number of observations OLS = ordinary least squares R2 = coefficient of determination Rbar2 = adjusted R2 Reg = OLS-regression SER = standard error of regression (=sigma) t = t-statistic 1. Causal Models Causal models specify and estimate the linkage between dependent and independent variables. In the present context, the focus is on electricity consumption as a function of GDP and average tariffs. Functional Form The general form of the multiple linear regression model used in this annex is (1) y = 0;+Ix+.;'-+ Pk-1Xi.k-1,+s,t = 1,2,...,n. -44- There are n observations for the dependent variable y and the k-i independent variables x,. The error terms are denoted by 6. The constant (intercept) Po and the (slope) coefficients Pi need to be estimated. Estimates are denoted by b. Classical statistical analysis assumes that the true, albeit unknown, values of the coefficients are fixed. From a Bayesian viewpoint, however, the coefficients are treated as random variables. Equation (1) can be written in matrix form as (I')y = X3 +6. where X is a n*k matrix, with the first column consisting of elements set equal to unity, while P is k-dimensional. The log-linear version of (1') is (2) Log [y] = Log [X]P + e. Equation (2) has the advantage that its coefficients can be interpreted as elasticities, i.e., Pi = (&y/axi)/(y/xi). The basic equation estimated below accounts for a lagged dependent variable and has the form (3) Log[y,] = Log[P J] + P I Log[x,.;}+... P .; Log[x, k -if1 + a Log[y,_J] + F-,,t=12. Equation (3) is commonly referred to as the partial adjustment model8 and distinguishes between short-run elasticities. 3= (Jv/axi)/(y/x), and long-run elasticities. Pi/(1-cc). with a<l. OLS and IV The available time series used to estimate Equation (3) cover the period 1988-1994. Since EDM's records of past electricity sales are flawed, annual gross consumption, which has been measured with greater accuracy, serves as a proxy for electricity demand.'9 Annual average tariffs " Let y be the equilibrium level of y relative to X. which is defined for t=t,.,= ...=tn. Adjustment of y, to its equilibrium level y is incomplete. The time path of y, starting from the initial value ye, is y, = y* (y/ )Y* 19 Gross electricity consumption includes technical and non-technical network losses and station use in thermal generation. -45- are computed as the ratio of billings to net supply (including non-technical losses). Figures on real GDP are obtained by using the consumer price index as deflator. The variables and data under consideration are: C = constant (= intercept) y = Log of gross electricity consumption (MWh) y(-1) = one year lag of y Y = Log of real GDP (1994=100) P = Log of real average tariffs (MT/kWh) Y(-1) = one year lag of Y X = {C, Y. P, y(-1)}, n*k-matrix Xl = {C, Y, P, Y(-1)}, n*k-matrix X = 1 4.3553"5.3748 13.345 1 4.4018 5.4343 13.376 1 4.4202 5.6565 13.517 1 4.4416 5.7106 13.512 1 4.3983 6.0182 13.589 1 4.5526 5.9469 13.599 1 4.6052 5.8493 13.656 Estimating Equation (3) by OLS gives: Reg[y, X,]; Dependent variable is y; R2= 0.903749; RBar2= 0.807498; SER = 0.0479503; N= 7; df= 3;dw = 2.85687 with 0 missing obs. coef. st. err. t C 8.460 7.100 1.192 Y 0.676 0.446 1.517 P 0.188 0.267 0.704 y(-1) 0.076 0.744 0.102 err(-1) -1.246 0.530 -2.353 Variance-decomposition proportions CI: 1375.1 159.2 65.90 1.0 C 0.990 0.009 0.002 0.000 Y 0.623 0.375 0.002 0.000 P 0.869 0.022 0.109 0.000 y(-1) 0.999 0.001 0.000 0.000 -46 - Given the small and spurious data base, it comes not much as a surprise that the OLS-estimates look flimsy. The coefficients are insignificant, and that pertaining to price has the "wrong" sign. The low significance levels together with the strong dependencies among the data at a scaled condition index of 1375 suggest the presence of collinearity. Serial correlation, however, seems not to be a problem.: Note also that the estimates are subject to a small-sample bias due to the presence of a lagged dependent variable. Moreover, it is likely that the explanatory variables have been measured with errors, thus leading to inconsistent estimates.21 Consistent estimates can be obtained by the IV-method. Using Y(-1) as an instrument for y(-1), the results are: IV[y, X, X1); Dependent variable is y; R2= 0.873814; RBar2 = 0.747629: SER = 0.0551756 N= 7; df= 3; dw = 3.31176 with 0 missing obs: Instruments {C, Y, P. Y(-1)}. coef. st. err. t C 1.526 10.256 0.149 Y 0.321 0.603 0.533 P -0.057 0.377 -0.151 y(-1)y 0.809 1.078 0.750 err(-1) -0.932 0.514 -1.814 Variance-decomposition proportions CI: 1666.5 9.1 3.6 1.0 C 1.000 0.000 0.000 0.000 Y 1.000 0.000 0.000 0.000 P 0.999 0.001 0.000 0.000 y(-1) 1.000 0.000 0.000 0.000 Based on the IV-estimates, the elasticity with respect to price has a correct sign. Collinearity, however, continues to be a problem, and the coefficients lack significance. 20 Note that with a lagged dependent variable the dw-statistics is biased towards the null of no serial correlation. A better test is to regress the residuals on the explanatory variables and the (one period) lag of the residuals and check on the significance of the coefficient associated with err(-1): If the lagged residuals are insignificant, the null of no (first-order) serial correlation can be accepted. 21 While unbiasedness is a finite-sample property (i.e., does not depend on the sample size), consistency means, simply speaking, asymptotic unbiasedness. -47 - Mixed Estimation To correct for collineanity and to improve the small sample properties of the estimate. we can resort to extraneous information about the coefficients. The information considered are coefficient estimates from Mauritius. The estimates are given by the vector c, while the corresponding diagonal elements of the variance-covariance matrix are denoted by S: c=f{2.838, 0.431. -0.098, 0.669}; R=f{ { 1,0,0,0}, (0,1.0,0}, f 0.0.1,0}, { 0.0,0.1}}; S=DiagonalMatrix[0.108775, 0.00625779. 0.00150073, 0.00296514]; Using the vector c as a stochastic restriction. Mixed Estimation gives the following GLS-results: Mixed Estimation(y, X, c, R. S]; Dependent variable is v: N = 11; df = 7; dw= 1.67584 with 0 missing obs. coef. st. err. t C 2.855 0.302 9.441 Y 0.442 0.071 6.242 P -0.083 0.035 -2.383 y(-1) 0.682 0.032 21.600 Bayesian Estimation Since Mixed Estimation is a special case of Bayesian statistics. we can harness the prior information from Mauritius to conduct a full-fledged Bayesian analysis. To this end it is assumed that the (-coefficients and the residual variance c2 are stochastic with a joint normal gamma distribution. Based on the information from MauritiuS22, the prior distributions have the means priorb = {2.838, 0.431, -0.098, 0.669}; priorsigma = 0.0248961. In addition, given the variance matrix priorsigma*A' associated with the observations from Mauritius, the implied confidence in the estimates is reflected by 2 It would be desirable to have prior information from a country that is more similar to Mozambique. Unfortunately, this kind of data is not available. But Mauritius is clearly a more appropriate comparator country than, say, Denmark. -48- A =0.0056982 0.030873 -0.76786 -0.042737 0.030873 0.099047 0.45046 -0.14769 -0.767860 0.45046 0.41301 -0.57833 0.042737 -0.14769 -0.57833 0.20903 Combining the above information with the data from Mozambique yields the following posterior estimates:23 b_ = 2.81619, b, = 0.459614. b, = -0.0972486, b3= cc = 0.685094; postsigma = 0.0286687. Hence. the exponent of the intercept is equal to 16.73 1 and the elasticities have the following vaiues: Bayesian Elasticity Estimates Variable Short-Run Long-Run GDP 0.460 1.460 Price -0.097 -0.309 Based on the posterior distributions of the coefficients, we can calculate priority density regions for a given probability content. The table presented below tells, for instance, that there is a 95% probability that the short-run GDP-elasticity lies in the Bayesian interval (0.367, 0.553). Highest Priority Density Regions Probability Content: 0.80 0.95 b, (0.400, 0.519) (0.367, 0.5530) b2 (-0.098, -0.096) (-0.099, -0.095) b3 (0.660, 0.710) (0.0646. 0.724) 2 Note that estimating the coefficients without (consistent) prior information about the intercept does not change the results. -49 - 2. Time Series Analysis Time series analysis is an example of non-causal modeling. It tries to delineate the behavior of a variable on the basis of its stochastic properties observed in the past. In this section, univariate time series analysis is applied to data on monthly gross electricity consumption for the period January 1989- March 1995 (see Annex 1, Table 3a). The vector of 75 observations is denoted by MC={MC , ., MC75}. One would expect that the stochastic process underlying the variable is a random walk with a trend, at least in the medium-to-longer term. The upward trend shown by the available data, however, is not particularly significant (Rbar2=0.63).24 Therefore, we base the test for stationarity on the assumption that the sample MC is generated by a random walk with almost no drift, i.e. y, = k+ oy,_,+ E,,wvithk ;: 0. Testing the null that cc=1 (unit root test) gives a DF-statistic of 74*(0.097-1)=- 66.822. while the critical 5% value is about -13.5. Hence. the hypothesis that MC is nonstationary can be be rejected. Likewise, the F-statistic for the null that ax=1 and k=0 is 8.1596, compared to a critical 5% value of about 4.75 (according to the DF-distribution). Given that the series follows a stationary stochastic process, the remaining problem is to find a model that captures the dynamics underlying the data. The dynamics can be described by the frequency spectrum shown below.25 The plot suggests that the low frequency components are by far the most important determinants of the sample variance (j=2, j=3, and j=8). As a consequence, we simply assume that the data fit an autoregressive process given by (4)y, =C t y,.2 +a 2Y,.3 +CC 3y,., + = 1, 2,... Estimating equation (4) by OLS gives Reg[MC. {lag[MC,2], lag[MC,3], lag[MC,8]}]; Dependent variable is MC; R2 = 0.641494; RBar2 = 0.63029; SER = 4037.1; N = 67; df= 64; coef. st. err. t MC(-2) 0.50806 0.093 5.468 MC(-3) 0.28555 0.098 2.923 MC(-8) 0.21942 0.094 2.322 24 Another unusual feature of the data is the absence of annual seasonality. 25 The spectrum has been derived from the Fourier transform of the data. -50- The above coefficient estimates will be used to conduct a short-term forecast of monthly gross -26 electricity consumption. Frequency Spectrum of Monthly Gross Electricity Consumption .6 Note that the sum of the estimated coefficient values is slightly greater than unity. So there is a small drift or, what comes to the same thing, almost no drift (as was assumed at the outset). In the future, however, the drift is likely to become more important. At any rate, the short-term forecasting model should be refined and updated once additional data become available. -51 - ANNEX 6: FORECAST OF ELECTRICITY CONSUMPTION 1. Short-Term Forecast Based on the autoregressive model estimated in Annex 6 (Equation 4), monthly gross consumption has been forecasted for the period April 1995 - December 1996. The results are shown below. Table 1: Forecast Monthly Gross Electricity Consumption (MWh)u Year 1995 1996 January 79200 81531 February 73157 81279 March 82636 81539 April 76504 82156 May 79304 82189 June 79129 82413 July 78865 83130 August 80160 82867 September 80034 83650 October 79290 83665 November 81676 84045 December 79916 84412 Total Annual 949871 992876 1/ Actual consumption for the first quarter of 1995. According to the predictions of monthly consumption, annual gross consumption can be expected to reach 950,000 MWh in 1995 and 993,000 MWh in 1996. EDM's prediction for 1995 is 948,000 MWh. It should also be mentioned that the lead time of the 1996 forecasts exceed the length of the largest lag considered in the model (8 months). Therefore, the projections made for 1996 tend to be less accurate than the figures predicted for 1995. - 52 - 2. Long-Term Forecast A major difficulty with long-term forecasts based on causal models described in Annex 6 is that even when the parameters of the model were known with certainty (which they are not), forecasting is conditional on predictions of the future behavior of the explanatory (independent) variables. Moreover, since the explanatory variables are subject to policy interventions, structural changes and other factors not incorporated in the model, forecasting rests on assumptions about the policy environment and its impact on the coefficients of the model. The Bayesian coefficient estimates that will be used to forecast gross electricity consumption on the basis of Equation 3. Annex 6. are: Log of intercept: 2.811619 Short-run GDP elasticity: 0.4596 Short-run price elasticity: -0.0973 Coefficient pertaining to the lagged dependent variable: 0.6851 The corresponding long-run elasticities are 1.460 for GDP and -0.309 for price. The large differential between long-run and short-run elasticities is attributable to the low speed of adjustment (i.e., the comparatively high value of the coefficient associated with the lagged dependent variable). For instance, suppose that current electricity consumption is 10% below the equilibrium level. Given GDP and price. then our estimate of the adjustment coefficient implies that it will take 6.2 years to come as close as 1% to the equilibrium level. If the estimate were 0.60 (rather than 0.6851), the same result would be achieved after 4.6 years. So in the context of our estimates. the "long-run" is, asymptotically speaking, far into the future. Since the future development of the explanatory variables (average tariff revenues in constant prices of 1994; index of real GDP. with 1994=100) as well as the policy environment are uncertain, we consider two tariff scenarios and three GDP scenarios. The forecast horizon covers the period 1995-2010. The first tariff scenario assumes that average tariffs rise from 347 MT/kWh in 1994 to 575 MT/kWh in 1998 (in constant prices of 1994). Based on the average annual rate of exchange that prevailed in 1994 (6054 MT/US$), this is equivalent to increasing the tariffs from 5.7 UScents/kWh in 1994 to 9.5 UScents/kWh in 1998. The latter figure reflects the tariff target level that EDM thinks is necessary for its financial viability. On the other hand, we also consider a lower target level 7.5 UScents/kWh (454 MT/kWh). Under both tariff scenarios, only modest increases (in real terms) are expected for the period after 1998. 27 The results can be obtained by solving the formula describing the time path of the dependent variable for the index of time (see Annex 5. footnote referring to Equation (3)). - 53 - Regarding the prospects for GDP-growth. we consider three scenarios, ranging from "low growth" to "high growth". Alternatively, the scenarios may be classified in terms of the economic progress made by Mozambique during the transition from peace to economic recovery. The least optimistic scenario can be.squeezed into the assumption that real GDP on average rises at 3% a year. Medium growth translates into average annual growth rates of 4.7% for 1995-2000, 4.2% for 2000-2010, and 4.5% for 1995-2010. The implied growth rates of the most optimistic GDP-scenario are 6.7% for 1995-2000, 5.5% for 2000-2010, and 5.9% for 1995-2010.28 The different assumptions concerning tariffs and GDP-growth are summarized in Table 2. Table 2: Summary of Long-Term Scenarios' - rTaiff , GDP Year High Low Low Medium High 1995 412 412 102.00 103.80 105.50 1996 454 424 104.55 108.47 111.83 1997 515 454 107.69 113.35 118.76 1998 575 454 111.03 118.45 126.48 1999 575 454 114.91 124.38 135.97 2000 575 454 118.93 130.59 146.17 2001 580 459 123.33 137.39 155.67 2002 580 459 127.90 144.94 165.01 2003 580 459 131.99 151.75 174.91 2004 585 464 136.21 158.13 184.53 2005 585 464 140.30 164.45 193.76 2006 585 464 144.51 171.03 203.44 2007 590 469 148.84 177.87 213.62 2008 590 469 153.31 184.99 224.30 2009 595 474 157.14 191.46 235.51 2010 596 474 161.07 198.16 246.11 1/ Tariff: MT/kWh in constant prices of 1994: GDP: index of GDP (1994= 100); the average annual rate of exchange used for 1994 is 6054 MT/USS. Combining the different tariff assumptions and GDP projections, we obtain six electricity consumption scenarios, running the gamut from "low growth cum high tariffs" to "high growth cum low tariffs". The resulting forecasts are shown in Table 3. Since the predictions cover a fairly broad region of possible developments, no consideration is given to confidence intervals. 2n Growth rates are least squares estimates. - 54 - Table 3: Forecasts of Gross Electricity Consumption (GWh)u Year Fla Flb F2a F2b F3a F3b 1995 935.03 935.03 942.58 942.58 949.64 949.64 1996 962.38 968.80 984.21 990.77 1003.22 1009.92 1997 982.90 999.57 1021.89 1039.22 1057.80 1075.74 1998 1000.52 1035.64 1058.56 1095.71 1117.08 1156.29 1999 1028.89 1077.99 1109.06 1161.98 1198.'1 1256.01 2000 1065.50 1125.64 1170.97 1237.07 1300.75 1374.18 2001 1108.77 1177.77 1243.01 1320.37 1414.76 1502.80 2002 1158.64 1235.36 1327.14 1415.02 1539.26 1641.18 2003 1211.49 1295.03 1417.64 1515.39 1675.08 1790.58 2004 1266.23 1355.62 1510.27 1616.89 1817.69 1946.01 2005 1323.02 1417.91 1605.85 1721.04 1965.94 2106.95 2006 1382.04 1482.24 1705.29 1828.92 2121.44 2275.24 2007 1442.25 1547.25 1807.76 1939.36 2283.85 2450.11 2008 1505.34 1615.24 1915.74 2055.60 2456.71 2636.06 2009 1566.54 1680.78 2023.50 2171.06 2639.00 2831.43 2010 1628.02 1746.93 2133.94 2289.79 2827.83 3034.36 ARG': 1995-2010 4.0 4.5 5.9 6.4 7.9 8.4 1995-2000 2.6 3.8 4.2 5.4 6.4 7.6 1/ Fla = low GDP, high tariff, Fib =low GDP, low tariff, F2a = medium GDP, high tanff: F2b = medium GDP, low tariff; F3a = high GDP, high tariff: F3b = high GDP, low tariff. 2/ Average annual rate of growth (%); least squares estimates. Clearly, the assumptions about GDP-growth have the strongest impact on predicted electricity consumption. Switching from a lower- to a higher-growth-scenario accelerates electricity consumption by about 2 percentage points. By contrast, the impact that the choice between the two tariff scenarios has on predicted consumption growth (for a given given time path of GDP) is equivalent to 0.5 percentage points. Depending on the scenario. the average annual rates of growth of gross consumption vary between 4.0% (low GDP, high tariffs) and 8.4% (high growth, low tariffs). For comparison, during the last three years gross consumption grew at about 5.5%. What the forecasts have in common is that beyond the year 2000 electricity demand rises markedly faster than in the period 1995-2000, mainly because the assumed increases in price are highest during the initial three years. Another reason for the more rapid growth in later years is the delayed adjustment to changes in price and, more importantly, GDP. It is also worth noting that the long-term predictions made under the "medium- growth-cum-low-tariffs" scenario for 1995 and 1996 are closest to the short-term consumption estimates based on time-series analysis. A comparison between the above scenario forecasts and previous predictions carried out by Norconsult (1993) and EDM (1995) shows the following (see Table 4): - 55 - If adjusted for technical losses, the Norconsult forecasts for 2010 lie between our "low-growth-cum-high-tariffs" and "high-growth-cum-high-tariffs" predictions. On the other hand, for the year 2000, Norconsult's medium- and high-case figures are significantly above those of the highest forecast (high growth, low tariffs) generated from our scenarios. Thus, the growth rates implied by the Norconsult predictions for the period 2000-2010 are consistently less optimistic than those of our forecasts. A possible explanation of this discrepancy is that our forecasts are based on a model in which the elasticities of demand with respect to GDP and price are high in the long-run and low in the short-run, while Norconsult may have used an entirely different model. The demand forecast that EDM' sfinancial department prepared in May 1995 covers the period 1995-2005 and is part of a study on cash-flow problems besetting the utility. The predictions, which reflect ad-hoc assumptions, are in close vicinity of our "low-growth-cum-high- tariffs" scenario. Table 4: Previous Demand Forecasts: Norconsult and EDMu 1995 2000 2005 2010 Norconsult - Low 1,320 1,626 - Medium 1,769 2,419 - High 1,851 2,708 EDM 904 1,108 1,318 1/ Gross electricity consumption (GWh); original estimates refer to sales and have been adjusted for technical losses. No attempt has been made in this study to model and estimate a peak load function which could be used to predict maximum demand conditional on gross electricity consumption and other variables deemed relevant. Instead, system peak or, what comes to the same thing, the system load factor is assumed to be linear in gross consumption. Resorting to this simplification, however, poses the problem of determining a standard (or representative) load factor for the lead time of the forecast. What adds to this difficulty is that there is no information on the coincidence of the peaks recorded in the three subsystems of Mozambique's power sector. In the near future, however, the subsystems will be interconnected either directly between the northern and central region, or indirectly, via South Africa, between the northern and southern region. As a consequence, systemwide planning will gain importance. According to rough estimates, the combined system load factor rose from 0.53 in 1985 to 0.58 in 1990 and reached 0.67 in 1994. The sharp increase in the last four years mainly reflects improvement in the reliability of supply, notably after the peace accord of 1992, which more than offset the adverse impact that the decline in industrial demand and the momentum of residential consumption and other small-volume use of electricity may have had on the load factor. Assuming that peaceful conditions persist and that economic recovery will take the lead in determining the changes in the system's load profile, we expect the load factor to remain in the - 56 - neighborhood of 0.66. The resulting annual peaks corresponding to the scenario-dependent projections of gross electricity consumption are presented in Table 5. Table 5: Forecasts of System Peak (MW)u Year Fla Flb F2a F2b F3a F3b 1995 61.7 161.7 163.0 163.0 164.3 164.3 1996 166.5 167.6 170.2 171.4 173.5 174.7 1997 170.0 172.9 176.7 179.7 183.0 186.1 1998 173.1 179.1 183.1 189.5 193.2 200.0 1999 178.0 186.5 191.8 201.0 207.3 217.2 2000 184.3 194.7 202.5 214.0 225.0 237.7 2001 191.8 203.7 215.0 228.4 244.7 259.9 2002 200.4 213.7 229.5 244.7 266.2 283.9 2003 209.5 224.0 245.2 262.1 289.7 309.7 2004 219.0 234.5 261.2 279.7 314.4 336.6 2005 228.8 245.2 277.8 297.7 340.0 364.4 2006 239.0 256.4 295.0 316.3 366.9 393.5 2007 249.5 267.6 312.7 335.4 395.0 423.8 2008 260.4 279.4 331.4 355.5 424.9 455.9 2009 271.0 290.7 350.0 375.5 456.4 489.7 2010 281.6 302.2 369.1 396.0 489.1 524.8 1/ Fla = low GDP. high tariff; Fib = low GDP, low tariff. F2a = medium GDP, high tariff, F2b = medium GDP, low tariff; F3a= high GDP, high tariff; F3b = high GDP. low tariff. There are no foolproof criteria for selecting among the different scenarios underlying the long-term forecasts presented above. Therefore, the focus will be on a forecast interval within which we expect electricity consumption to develop (disregarding the possibility of a long-lasting stagnation or significant temporary decline). The lower band is clearly the forecast based on the "low-growth-cum-high-tariffs" scenario. For the higher band, we consider the "medium-growth-cum-low-tariffs" scenario as the best bet. Regarding the distribution of gross electricity consumption between the southern region and the rest of the country, we assume that the share accounted for by the southern system increases from 64% in 1994 to 65% in 1995, thereafter gradually declines to 62.5% by 2000 and remains at this level until 2010. Besides, all systems will be directly or indirectly interconnected by 1997. - 57 - ANNEX 7: LONG-RUN AVERAGE INCREMENTAL COSTS OF POWER SUPPLY In computing the long-run average incremental costs (LRAIC) of power supplied to the domestic market, we make a number of assumptions explained below. Power from HCB and Import Opportunities A key assumption is that Hydroelectrica de Cohora Bassa (HCB) starts supplying Eskom in March 1997. Henceforth. the southern system may retrieve up to 190 MW (=200/1.05) from the Komatipoort busbar. Given the share of northern-central demand covered by HCP-power, Lb, the balance available for the southern system is L, = (200-Lb)/1.05. Currently, the monthly payments for HCB-power injected into the northern system are: F, = 0.8*L*h*P if E<0.8*L*h, or F2 F, + (E - 0.8*L*h)P/3 if E>0.8*L*h. where L = maximum demand (MW), h = length of month (hours), P = 5 Rand/MW, E = metered energy supply (MWh). The above formulas are part of an agreement that HCB and EDM signed in May 1983. F, can be interpreted as a capacity charge. F, includes a penalty for energy supplied in excess of the level determined by a load factor of 0.8. The old agreement calls for slightly different terms in the event that HCB wheels power to the southern system through Komatipoort, which has not been the case since the early 1890s. Here, the formula for the payments by EDM to HCB reads F3 = 1.05(L*h*Ps + (E - L*h)Ps/3) F3 is composed of a capacity fee and a surcharge (discount) for energy that exceeds (falls short of) the ceiling given by the contracted capacity. The factor 1.05 accounts for transmission losses. Under the old contract29 Ps was fixed at 7.5 Rand/MW. In 1989, Eskom agreed to pay HCB 20 Rand/MW if and when the link to Apollo is put back into service. So the 20 Rand/MW would, ceteris paribus, become a benchmark for imminent renegotiations between HCB, Eskom and EDM. In addition, EDM would be required to pay to Eskom a standby charge of 500 Rand/MW plus a monthly transmission fee of Rand 30,000 (according to the old contract). Also, 2 The contract was assumed to remain valid for 30-35 years, i.e. until 2010-2015. - 58 - shoula Hi wheel power to the soutnern system, these supplies would substitute for the current imports from Eskom, at least as long as the power recalled from HCB is sufficient to meet growing demand in Mozambique's south. In any case, the above formulas, and more so the rates, are apt to change as soon as HCB is in position to transmit power from Cahora Bassa to Eskom and EDM's southern customers. EDM at present pays 0.4 UScents/kV-h under the 1983 contract, and Eskom would pay about 1.5 UScents/kWh if the 20 Rand/MW agreement of 1989 were to apply. With -uch low rates, HCB will be unable to repay its accumulated debt. So once the line from Cahora Bassa to Apollo is operational, HCB will seize the opportunity and press for higher rates. Another uncertainty is whether and on what conditions EDM will be permitted to increase its share of HCB-power, assuming that Eskom would give up some of its claims in the first place. In this connection it is worth -rentioning that even though the current import agreement with Eskom would expire in March. 2060 (3 years after the resumption of supply from HCB), there is the option to keep the import agreement in force beyond the year 2000 (Article 9). Therefore. it can be assumed that if Eskom were to release HCB-capacity to EDM. EDM would have to pay at least the rates charged under the current import agreement (which is about 2 UScentsikWh on average) or whatever rate Eskom agrees to pay HCB. Apart from the rehabilitation of the 533 kV line from Cahora Bassa to Apollo, we expect that a new 275 kV line between Zombodze (Swaziland) and Matola will be completed by the year 2000. With this additional transmission capacity of 150 MW, the southern system would be in a position to meet up to 350 MW through imports (including power from HCB). Regarding the connection betweeen the northern and central system, it is supposed that EDM completes and operates a temporary 110 kV line (10 MW capacity) by 1996, thus facilitating the supply of HCB-power to the central region. Also, we assume that the 220 kV line between Matambo and Chibata/Xigadora will be ready by the year 2001 (for details, see Table 5 below). Power Supply from Existing Generating Facilities The firm hydro capacity in the central and northern system is assumed to be 50 MW at 3500 hours a year (175 GWh/year). The Corumana plant is rated at 12 MW for 2500 hours a year (30 Gwh/year), while Cahora Bassa has a capability equivalent to 6750 hours a year. The availability of existing thermal generating capacity is assumed to follow the path outlined in Table 11, Annex 1. The resulting scope for power generation and power imports (including acquisitions from HCB and subject to transmission line constraints mentioned above) is summarized in Table I below. -59- Table 1: Potential for Power Imports and Power Generation from Existing Plant (MW), 1995-2010 Southern System Central + Northern System Year Hydro Thermal Imports Total Hydro Thermal Imports" Total 1995 12 90 200 302 50 30 200 280 1996 12 90 200 302 50 30 200 280 1997 12 90 200 302 50 30 200 280 1998 12 87 200 299 50 35 200 284 1999 12 87 200 299 50 35 200 284 2000 12 67 350 429 50 35 200 284 2001 12 67 350 429 50 35 200 284 2002 12 67 350 429 50 35 200 284 2003 12 67 350 429 50 35 200 284 2004 12 67 350 429 50 35 200 284 2005 12 26 350 388 50 34 200 283 2006 12 26 350 388 50 34 200 283 2007 12 26 350 388 50 34 200 283 2008 12 26 350 388 50 34 200 283 2009 12 26 350 388 50 34 200 283 2010 12 26 350 388 50 34 200 283 1/ Not including the 400 kV line to Zimabawe (for HCB-exports) which is likely to be operational by 1998. Source: EDM and mission estimates. Future Demand Table 2 presents a regional breakdown of the forecasts of gross electricity consumption pertaining to the low-GDP-cum-high-tariffs scenario and the medium-GDP-cum-low- tariffs scenario, respectively. The corresponding predictions of peak load are shown in Table 3 (for details, see Annex 6). - 60 - Table 2: Forecasts of Gross Electricity Consumption by Region(GWh)" F2a F2b Year SS CN ST SS CN ST 1995 607.8 327.2 935.0 612.7 329.9 942.6 1996 620.7 341.7 962.4 639.1 351.7 990.8 1997 629.0 353.9 982.9 665.1 374.1 1039.2 1998 635.3 365.2 1000.5 695.8 399.9 1095.7 1999 648.2 380.7 1028.9 732.1 429.9 1162.0 2000 665.9 399.6 1065.5 773.2 463.9 1237.1 2001 693.0 415.8 1108.8 825.2 495.2 1320.4 2002 724.1 434.5 1158.6 884.4 530.6 1415.0 2003 757.2 454.3 1211.5 947.1 568.3 1515.4 2004 791.3 474.9 1266.2 1010.6 606.3 1616.9 2005 826.9 496.1 1323.0 1075.6 645.4 1721.0 2006 863.8 518.2 1382.0 1143.1 685.8 1828.9 2007 901.4 540.9 1442.3 1212.1 727.3 1939.4 2008 940.8 564.5 1505.3 1284.7 770.9 2055.6 2009 979.1 587.4 1566.5 1356.9 814.2 2171.1 2010 1017.5 610.5 1628.0 1431.1 858.7 2289.8 1/ Fla = low GDP. high tanff: F2b = medium GDP, low tariff: SS = southern system: CN = central and northern system: ST = system total. Table 3: Forecasts of Peak Load by Region (MW)" Fla F2b Year SS NC ST SS NC ST 1995 105.1 56.6 161.7 106.0 57.0 163.0 1996 107.4 59.1 166.5 110.5 60.9 171.4 1997 108.8 61.2 170.0 115.0 64.7 179.7 1998 109.9 63.2 173.1 120.3 69.2 189.5 1999 112.1 65.9 178.0 126.6 74.4 201.0 2000 115.2 69.1 184.3 133.7 80.3 214.0 2001 119.9 71.9 191.8 142.7 85.7 228.4 2002 125.3 75.1 200.4 153.0 91.7 244.7 2003 131.0 78.5 209.5 163.8 98.29 262.1 2004 136.9 82.1 219.0 174.8 104.9 279.7 2005 143.0 85.8 228.8 186.0 111.6 297.7 2006 149.4 89.6 239.0 197.7 118.6 316.3 2007 155.9 93.6 249.5 209.6 125.8 335.4 2008 162.7 97.7 260.4 222.2 133.3 355.5 2009 169.3 101.7 271.0 234.7 140.8 375.5 2010 176.0 105.6 281.6 247.5 148.5 396.0 1/ Fla = low GDP, high tariff; F2b = medium GDP, low tariff; SS = southern system; CN = central and northern system; ST = system total. Investments in Power Generation A comparison of the demand forecasts with EDM's supply options reveals the following: Under the current agreement, power from HCB, together with EDM's existing -61- generating capacity, is sufficient to meet demand both in the northern/central region and in the southern system until 2006 (low growth, high tariffs) or 2001 (medium growth, low tariffs), assuming that the thermal plants are by and large operated on a standby basis (see Table 4). So no matter what scenario is used, EDM has to seek additional sources of supply. Basically, it could resort to imports from Eskom, count on the option to increase its ICB-share, or expand its own generating capacity. In what follows, we assume that either of the first two options will be available, and that the respective costs of power supply would be lower than those associated with additional generating plant (e.g. the proposed Alto Malema hydro scheme). Table 4: Projected Supply Deficits Year Low Growth cum High Tariffs Medium Growth cum Low Tariffs Capacity (MW) Energy (GWh) Capacity (MW) Energy (GWh) 2002 - - 10 58 2003 - 25 145 2004 - - 43 250 2005 - - 65 327 2006 - - 80 420 2007 10 58 100 530 2008 25 145 124 648 2009 35 203 144 755 2010 45 260 164 875 Source: Mission estimates. Another problem is that the risk of line outages is comparatively high in the north- eastern region, which is supplied from Cahora Bassa over a distance of more than 1000 km. Also, the planned grid extensions to Pemba and Lichinga will increase this risk. While the option to reinforce the grid through the construction of parallel lines can be dismmissed as prohibitively expensive, it may be reasonable to hedge against the risk of transmission failures through the provision of backup generating facilities in the main load centers. In fact, EDM, following the advice of NORCONSULT, plans to install a 25 MW gas turbine in Nacala as an additional backup. Whether such a project is economic, however, depends on the probability of transmission failures - 62 - on the one hand, and the customers' willingness to pay for improvements in the reliability of supply on the other.30 Unfortunately, we do not know what premium the different customer groups of the northern region are prepared to pay for uninterrupted power. Nor do we have estimates of the probability of line outages at different locations. It can be assumed, though, that the reliability of the line will improve, which will increase the premium that customers would have to pay for further reducing the risk of transmission failures. Also, a sizeable group of consumers with an exceptionally high value of lost load is more likely to emerge under the medium-growth-cum-low- tariffs scenario than under the low-growth-cum-high-tariffs scenario. Therefore, we include the proposed stand-by gas turbine (25 MW) only in the investment program associated with the medium growth demand forecast. Under both scenarios. we also account for planned investments in thermal plant overhauls and the rehabilitation of the Chicamba dam. As has been mentioned above, no convincing case can be made for the construction of a 80 MW hydro plant at Alto Malema (with estimated costs of US$ 110 million). The resulting investment program is shown in Table 5. GI = thermal rehabilitation G3 = Alto Malema Hydro. 80 MW G4 = gas turbine Nacala G5 = Chicamba dam rehabilitation 30 Suppose that line outages are exponentially distributed with parameter l/P, where P is the expected duration of outages per unit of time. Moreover, let the long-run willingness to pay for a reliability level I-P be W(P). Then it is economic to provide a backup generating plant if F W(P) > + c, hP(1 -Q) where F = fixed costs of the backup ($/kW/year) c = operating costs of the backup ($/kWh) h = annual load factor (hours per year) Q = outage rate of the backup (per unit of time). For instance, based on a gas turbine with specific investment costs of 650 $kW (10% interest, lifetime: 25 years, fixed O+M: 13 $/year), we have F = 85 S/kW, c = 0.12 S/kWh. Let Q = 0.15 and h = 0.66. Then for a backup to be economic. the willingness to pay must be at least 0.29 S/kWh for P= 0.1, 0.47 S/kWh for P= 0.05, and 0.81 $/kWh for P = 0.025. - 63 - Table 5: Investment Plan Generation (million USS of 1995) 96 97 98 99 00 01 02 03 04 Total GI 1.0 1.0 1.0 - - - - - - 3.0 G4 - - - - - - 16.0 - - 16.0 G5 - - - 1.5 1.5 - - - - 3.0 1/ Onl- the medium-growth-cum-low-tariffs scenario. Source .' and mission estimates. Inves. in Transmission Facilities For calculating the LRAIC of domestic power supply, we have revised EDM's tenta estment program as follows. The overhaul of the 220 kV line from Songo to Nampula (TI) is spread over a perioc *ears. EDM's cost estimate of US$ 20 million has been adjusted downwards to US$14 millio: kewise, the costs of overhauling the line from Nampula to Nacala (T3) has been reestir -it US$ 7 million. An additional US$ 2 million have been allocated to the HCB-EDM interce ion projects (T7). The rehabilitation of the second Mavuzi-Nhamatanda line (T4) starts in 199 1 er than in 2001. First measures to improve voltage control in Beira (T6) are scheduled for 19- The planned grid extension to Cuambo and Lichinga is postponed, starting in 2001. Simila ie lines to Montepuez and Pemba (T9) are installed between 2001 and 2004 rather than in the d 1996-1998. The extension projects T12, T13 and T15 are commissioned not before 2006. ..n though the timing of the planned investments in the XaiXai-Inhambane line (TI0) is questic le on economic grounds, it is assumed that the required funds are committed and the projecl i be implemented on EDM's schedule (1996-1998). The planned interconnection between Swaziland and Mozambique is considered to be ess. . for the future supply of the southern system. A less convincing case can be made for the pr ed interconnection between Orange Grove (Zimbabwe) and Chibata. We assume, howev. :at the line will contribute to the reliability of supply in Mozambique, with the works startin. 2006. On the other hand, typical export projects such as the 400 kV line from Songo to Harare nbabwe) and the planned 220 kV line to Blantyre (Malawi) are not included in the investv program used for computing LRAIC. A list of all transmission projects under consideration is presented below. Table 6 shows . sequence of investments deemed relevant for LRAIC. TI = 22', kV, rehab Songo-Nampula, split into Stage I and Stage 2 T3 = ' kV, replacement Nampula-Nacala T4 = I kV, rehab of 2nd line Mavuzi-Nhamatanda T5 = 245 kV, Swaziland (Zombodze)-Mozambique (Matola) - 64- T6 = reactive power devices Beira T7 = 110 kV + 220 kV, interconnection HCB-EDM (at Chibata) T8 = 110 kV, Alto Molocue-Gurue + substation Gurue T9 = 110 kV, Nampula-Ancuabe-Montepues-Pemba T10 = 110 kV, XaiXai-Inhambane T 11 = 110 kV, Gurue-Cuamba-Lichinga T12 = 110 kV, Corumana-Ximavane T13 = 110 kV, Massingir-Chokwe T14 = 220 kV, Mozambique (Matambo)-Malawi (Blantyre) T15 = 110 kV, Caia-Luabo-Marromeu Table 6: Investment Plan Transmmision System (US$ million of 1995) Year 96 97 98 99 00 01 02 03 4 Total TI 2.0 1.0 1.0 2.5 2.5 2.0 1.0 1.0 .0 14.0 T3 3.0 1.0 1.0 1.0 1.0 - - - - 7.0 T7 0.5 - 3.0 3.0 3.0 3.0 - - - 12.5 T9 - - - - - 8.0 5.2 5.2 1.6 20.0 TIO 6.0 5.0 5.0 4.0 - - - - - 20.0 T8 - - - 2.0 2.0 - - - - 4.0 T4 1.0 1.0 1.0 - - - - - - 3.0 T6 - 0.25 0.25 0.25 0.25 0.25 0.25 0.25 0.25 2.0 TII - - - - - 8.0 5.2 5.2 1.6 20.0 T5/ 2.2 3.5 4.6 1.1 - - - - 11.4 T16u -.------ -7.5 T12 - - - - - - - - 4.0 T13 - - - - - - - - - 4.0 T15 - - - - - - - - - 6.0 1/ EDM's share. Source: EDM and mission estimates. Investments in Distribution Facilities The investments that EDM plans to allocate to primary and secondary distribution have been revised as follows: Rehabilitation of the Nampula and Beira distribution systems (D1, D2) is brought forward, starting in 1996. The estimated costs for Quelimane (D3) have been trimmed to US$ 4 - 65 - million. Overhauling the Maputo distribution network (D5) starts in 1999 at reestimated costs of US$ 3 million. The Maputo substation extensions (D8) are postponed by three years, with the costs adjusted upwards to US$ 5 million. The investments in the distribution systems of Gurue, Cuamba and Lichinge (D9) are deferred, stretching from 1999 to 2005. Rehabilitation of electrification schemes in the Chimoio and Chokwe areas (D10) is commissioned earlier (1996), with the costs cut to US$ 6 million. Likewise, rehabilitation of southern electrification schemes starts in 1996, while costs are cut to US$ 18 million and spread over nine years. Table 7 gives an overview of the investment program referring to the distribution projects listed below. - 66 - Dl = rehab and extension of Nampula distribution system D2 = rehab and extension of Beira distribution system D3 = rehab and extension of Quelimane distribution system D4 = rehab and extension of Nacala distribution system D5 = rehab and extension of Maputo distribution system D6 = power supply to Buzi D7 = rehab XaiXai, Mampula. Nacala substations D8 = extension Maputo substations SE4, SE5, SE6 D9 = primary and secondary distribution Gurue, Cuamba. Lichinge D10 = rehab electrification Chimoio and Chokwe DII = rehab electrification south D12 = Maputo substation extension D13 = overhaul of Matola distribution system D14 = overhaul distribution system XaiXai and Chokwe D15 = overhaul of Angoche distribution system D16 = conversion from to diesel to electric pumping in Limpopo D17 = new substation Matola D18 = new substation in Tete and Matundo El = urban household energy E2 = emergency program Nacala Table 7: Investment Plan Distribution (US$ million of 1995) 96 97 98 99 00 01 02 03 04 0.5 Total Dl 2.0 3.0 1.0 1.0 1.0 - - - - - 8.0 D2 - 3.0 2.0 2.0 1.0 - - - - - 8.0 D3 1.0 0.5 0.5 1.0 1.0 - - - - - 4.0 D4 1.3 1.6 1.5 1.5 0.3 - - - - - 6.2 D5 - - - 1.0 0.5 0.5 0.5 0.5 - - 3.0 D6 - - 1.1 0.8 0.2 - - - - 2.1 D7 5.2 1.4 0.6 - - - - - - 7.2 D8 - - - 1.0 1.0 1.0 0.5 0.5 0.5 0.5 5.0 D9 . - - - - - 2.0 2.0 2.0 - 6.0 DIO 1.0 1.0 1.0 1.5 1.5 - - - - - 6.0 DI 2.0 2.0 2.0 2.5 2.5 2.0 2.0 1.0 1.0 1.0 18.0 D12 1.6 2.5 3.4 0.8 - - - - . - 8.3 D13 - - - - - 2.0 2.0 1.0 1.0 1.0 7.0 D14 - . . - - 1.5 1.5 1.0 0.5 0.5 5.0 D15 . . - - - 1.0 0.5 0.5 - - 2.0 D16 - - - - - - 2.0 2.0 - - 4.0 D17 - - - - - 2.0 2.0 1.0 1.0 1.0 7.0 D18 - - - - - - 3.0 2.0 1.0 1.0 7.0 El 1.0 - - - - - - - - 1.0 E2 0.6 - - - - - - 0.6 Source: EDM and mission estimates - 67 - Other Investments Additional investments are required for rehabilitating and upgrading the control facilities in the Beira corridor (Cl), installing a remote center system in Maputo (C2), and improving EDM's telecommunication system (C3). Another item (C4) covers the planned expenditures for expanding EDM's vehicle fleet and construction a new headquarter in Maputo. C1 = rehab control system central region C2 = remote control center Maputo C3 = Telecommunication Phase II C4 = Cohara Bassa study C5 = vehicles, equipment. EDM headquarters building Table 8: Investment Plan: Miscellenous (million USS of 1995) 1996 1997 1998 1999 2000 2001 Total Cl 0.8 1.0 0.3 . - - 2.1 C2 - - - 1.5 1.5 - 3.0 C3 1.1 1.6 0.3 - - - 3.0 C5 0.3 1.3 1.5 1.5 1.0 0.3 5.9 Source: EDM and mission estimates. LRAIC Usually, long-run average incremental costs (LRAIC) are defined as (1)LRAIC - , q n".;(X, - X.)q-' where It = investment costs in period t, q = discount factor (I+i), with i as the discount rate, t =index of time. -68- For the sake of convenience, we assume that the investment program leads to a stationary state, i.e., from period t-n to t-co demand remains at the level X,. In this case, the denominator of Equation (1) can be expressed as " _ X,q' (2) , where LX, denotes incremental demand (GWh) in period t. i.e., LXt = Xt-Xt-1, t=1.2,...,n. Likewise, if the useful life of the additional facilities is uniformly equal to T years, the numerator of (1) becomes (3) F11Vq ___ S1- Av, stands for the incremental variable costs in period t. In view of the above transformations. Equation (1) can be rewritten as (')LRAIC= Z7=, V, q'+ i("Il,q')/ (1-q) Ziff=/ , X,q-' In calculating LRAIC on the basis of Equation (1'), the following additional assumptions are made (apart from those related to the investment program): (a) Average fuel costs of thermal power generation amount to be 0.08 US$/kWh. (b) With the resumption of supply from Cahora Bassa to Eskom. HCB pushes for higher rates. Renegotiations with Eskom and EDM lead to a rate structure which is close to what EDM currently pays for power imports from Eskom (see Annex 3). The same rates would apply to additional supplies from Cahora Bassa (> 200 MW), since any increase in EDM's share would come at the expense of Eskom's entitlements and, therefore, is tantamount to importing from the RSA. (c) Annual expenditures for operating and maintaining (O+M) the grid are equivalent to 0.75% (1.25%) of the investments in transmission (distribution) facilities. The useful life of these facilities is 30 years. ' Specific fuel consumption is 0.3 1/kWh for diesel, 0.4 1/kWh for jet fuel (to run a gas turbine), and 1.1 kg/kWh for coal. Economic costs of fuel use (including lubes) are 0.24 USS/1 for diesel, 0,30 USS/1 for jet fuel, and 28 USS/t for coal. Average fuel costs are based on a diesel-jet fuel mix of 85:15. - 69 - (d) The useful life of the gas turbine proposed for Nacala is 25 years, considering that it will be operated on a standby basis. (e) The rate of discount is 10% a year. (f) The projections of the incremental payments for EDM's labor force and the incremental maintenance costs associated with generating plant are shown in Table 9. The table also presents the scenario-independent changes in thermal generation. Table 9: Selected Incremental Costs and Incremental Thermal Generation Year ISAL" IMG' IOMG" 1996 0.0 0.0 -2 1997 0.1 0.0 0 1998 0.1 0.1 1 1999 0.1 0.0 1 2000 0.3 0.0 0 2001 0.4 -0.1 1 2002 0.2 -0.1 -16 2003 0.1 0.0 1 2004 0.1 0.0 0 2005 0.2 0.0 0 2006 0.2 -0.1 0 2007 0.2 0.0 2 2008 0.2 0.0 0 2009 0.2 0.0 0 2010 0.2 0.0 0 1/ Incremental salaries (US$ million), total salaries 1995: US$ 7.6 million. 2/ Incremental maintenance in generation (US$ million); total costs in 1995: US$ 2 million. 3/ Incremental thermal generation (GWh), total thermal generation in 1995: 38 GWh. Source: EDM and mission estimates. - 70 - Table 10: LRAIC of Power Supply" Low Growth, High Tariffs Medium Growth. Low Tariffs Generation:' UScents/kWh 2.795 2.803 Transmission: US$/kW/year 164.93 82.81 UScents/kWh 2.853 1.432 Distribution: US$/kW/year 150.68 75.66 UScents/kWh 2.606 1.308 Total T+D: US$/kW/vear 315.61 158.47 UScentsikWh 5.459 2.740 O+M: Transmissission Distribution 0.202 0.101 (UScents/kWh) 0.307 0.154 Miscellaneous: UScents/kWh 0.367 0.184 I/ In constant prices of 1995. 2/ Based on investment costs and expenditures for fuel, maintenance, etc. Source: Mission estimates. As is shown in Table 10, total LRAIC vary significantly in direct proportion to the projected increase in demand. Under the low-growth-cum-high-tariffs scenario, total incremental costs net of technical losses amount to 9.13 UScents/kWh, which is consistent with the underlying tariff assumption (i.e.. EDM's target level of 9.5 UScents/kWh). With medium GDP-growth and a targeted tariff level of 7.5 UScents/kWh, the corresponding costs work out at 5.98 UScents/kWh. At the generation end, the scenario-dependent incremental costs lie close together (about 2.8 UScents/kWh). This is because the higher (investment) costs incurred under the medium growth forecast are distributed over larger volumes of consumption and, also, because the scenarios do not affect the principal sources of supply. By the same token, there is a considerable gap between the scenario-dependent transmission- and distribution costs (5.46 vs. 2.74 UScents/kWh) since the respective investment programs are insensitive to the rate at which demand is projected to 32 grow. 32 It should be noted that we have already trimmed the T+D investment program submitted by EDM. Without these revisions, the long-run incremental costs would be even higher. On the other hand, postponing some investments would reduce the cost. For instance, a two years delay of the entire T+D investment program would certeris paribus reduce the incremental T+D costs to 4.47 UScents/kWh for the low growth scenario, and to 2.23 UScents/kWh for the medium growth scenario. - 71 - Assuming that the investments aimed at reinforcing and extending the grid are useful and have been fashioned in a prudent way, it can be concluded that in EDM's system the T+D functions account for an exceptionally large share of costs. While this reflects the low level of consumption relative to the size of the T+D assets (apart from the low level of generation costs), it also implies that the development of future demand is a key variable in determining the trend for long-run incremental costs of power supply. - 72 - ANNEX 8: BASIC FEATURES OF A REVISED TARIFF SYSTEM The main concern of a revised tariff regime should be that it more closely signals to consumers the structure and level of costs they impose on the system. Moreover, tariffs should be transparent. non-distortionary, easy to implement, and tailored to the system's load profile. With the present mix of supply, load pattern and consumer characteristics, there is no need for a sophisticated and highly responsive tariff system (e.g., time-of-use pricing). In particular, changes in the dispatch of the available sources of supply (in response to shifts in the system's load duration curve) are not apt to have a marked impact on marginal costs. Also, the responsibility of the different consumer groups for the system peak is fairly similar, and unless a boost in industrial consumption changes the picture. opportunities for load management will be limited. Thus. EDM should opt for a simple two-part tariff which distinguishes between capacity and energy costs and discriminates across voltage levels in accordance with average incremental costs. Based on the LARIC-estimates presented in Annex 7, the proposed structure of tariffs can be determined as follows: At the generation end, we assume that the peaker plant is a gas turbine. The annual capacity costs, including fixed expenditures for maintenance and repair, amount to 90 US$/kW At a load factor of 0.66, this is equivalent to 1.6 UScents/kWh.34 Since total incremental costs of bulk supply (generation plus imports and/or acquisitions) work out at 2.8 UScents/kWh, the balance of 1.2 UScents/kWh is attributable to variable costs which should be recovered through an energy rate. In addition, we suggest to adjust the energy rate so as to make up for the miscelleneous investment costs. Overall technical losses are assumed to devour 14% of bulk supply: 6% at the high voltage (HV) level, 2.5% at the medium voltage (MV) level, and 5.5% at the low voltage (LV) level. Incremental transmission costs are charged at the HV-level, while primary and secondary distribution costs are apportioned among LV and MV consumption on a 2:1 basis. Transmission-related O+M expenditures are factored into the HV rates, and those for the distribution system are charged at the LV-level. 3 Specific investment costs: 650 USS per installed kW: useful life: 20 years; interest: 10% a year; annual maintenance: 14 USS/kW. " Note that the load factor of 0.66 is an (essentially arbitrary) assumption made to convert capacity costs into a per-kWh rate. It should not be confused with the load factor of a plant used to serve peak load, which depends on the length of the peak period and would only be relevant in the case of strict peak-load pricing. - 73 - It does not seem expedient to differentiate tariffs by region or location. Once the northern and central system are inteconnected and the southern region has access to power from HCB, the average cost of bulk supplies become fairly uniform across the subsystems. Also, average network losses in the southern and northern-central region tend to converge. Clearly, there are significant locational differences in transmission losses. For instance, along the northern grid losses increase from 1% at Tete to about 6% in Mucuba. and to 10% in Nacala. Yet there is no compelling economic reason why these losses should not be treated as a public bad. A stronger case could be made for regionally discriminatory tariffs on account of the geographic distribution of EDM's planned investments. About two-thirds of the proposed expenditures are accounted for by projects aimed at strengthening and expanding the network in the northern and central part of the country. So why should customers connected to the southern system pay for these investments? Government would justify the implied cross-subsidies on political grounds. While this is a valid point, it does not relieve EDM of the need to select solutions that meet political imperatives at least cost, which is what we tried to accomplish by streamlining EDM's investment program (see Annex 7). Table I presents in skeletal form the structure of a tariff sytem based on the above assumptions and considerations. The level of tariffs varies with the rate at which consumption is expected to increase. Also, tariffs do not account for non-technical losses. Table 1: Structure of a Revised Tariff Systemu Low Growth, High Tariffs Medium Growth. Low Tariffs Cumulative Capacity Energy Capacity Energy Technical (USS/kW/mont) (UScents/kWh) (US$/kW/month) (UScents/kWh) Losses (%) Generation2" 7.5 1.57 7.5 1.38 Transmission 13.75 0.20 6.9 0.10 Losses 1.36 0.11 0.92 0.09 Total HV 6.0 22.61 1.88 15.32 1.57 Primary Distrib. 4.17 2.10 Losses 1.00 0.06 0.61 0.05 Total MV 8.5 27.78 1.94 18.03 1.62 Secondary Distrib. 8.39 0.31 4.2 0.16 Losses 3.14 0.17 1.85 0.13 Total LV 14.0 39.14 2.42 24.08 1.91 1/ In prices of 1995. 2/ Energy rate is adjusted so as to recover miscelleneous investment costs. Source: Mission estimates. - 74 - Table 2 gives an idea of how the proposed two-part tariffs would translate into (notional) average rates depending on voltage-specific load factors. It also shows the corresponding system average tariffs vary between 6.95 UScents/kWh (medium growth) and 10.43 UScents/kWh (low growth). Table 2: Notional Average Tariffs (UScents/kWh) LF" Low Growth Medium Growth HV 0.66 6.65 4.79 MV 0.58 8.50 5.88 LV 0.54 1235 8.02 System Average7 10.43 6.95 1/ Assumed load factor. 2/ Demand is assumed to be composed of 8% HV, 38% MV and 54% LV. Source: Mission estimates. CATALOGUERS/FILE CONFIDENTIAL Report No: 15150 mOZ Type: SR