42043 v2 THE WORLD BANK ANGOLA Assessing the Impacts of Phasing Out Fuel and Utility Price Subsidies VOLUME II - ANNEXES December 2005 TABLE OF CONTENTS APPENDIX I: ANGOLAN FUEL SUPPLIES AND DEMANDS, 2000-2004........................................ 5 APPENDIX II: ECONOMETRIC ANALYSES..................................................................................... 14 A. THE WORLD PRICE OF CRUDE OIL AS DETERMINANT OF REFINED OIL PRODUCT PRICES AND ESTIMATED DIESEL, GASOLINE AND KEROSENE SUBSIDY EXPENDITURES................................................ 15 B. THE WORLD PRICE OF CRUDE OIL, NOMINAL EXCHANGE RATE AND CPI AS DETERMINANT OF SELECTED GOVERNMENTAL REVENUES AND EXPENDITURES.................................................................... 19 C. DETERMINANTS OF CPI, BBPI AND MBPI: NOMINAL EXCHANGE RATE AND FUEL PRICES ....... 22 D. LUANDA'S ESTIMATED DEMAND FUNCTIONS FOR FUELS AND UTILITIES ................................... 25 E. ESTIMATED WATER AND GAS DEMAND FUNCTIONS FOR LUANDA'S MUNICIPALITIES ............... 29 F. ESTIMATED WATER AND GAS DEMAND FUNCTIONS FOR QUINTILE GROUPS.............................. 33 G. COMPARATIVE DEMAND ANALYSIS ............................................................................................ 35 H. BUDGET SHARES AS FUNCTIONS OF EXPENDITURE ..................................................................... 82 APPENDIX III: WELFARE ANALYSIS ............................................................................................... 85 APPENDIX IV: INCEPTION REPORT, MARCH 2005 ...................................................................... 87 I. BACKGROUND............................................................................................................................. 89 II. ORIGINAL SCOPE......................................................................................................................... 91 III. FRUITS FROM STAKEHOLDER CONSULTATIONS........................................................................... 92 IV. REVISED SCOPE......................................................................................................................... 103 V. METHODOLOGY......................................................................................................................... 104 VI. TIMELINE FOR THE REMAINDER OF THE PROJECT...................................................................... 105 APPENDIX V: RETAIL PRICES OF BASKETS OF GOODS AND SERVICES IN LUANDA.... 106 2 Acknowledgments This is the second volume of a report that has been prepared by a team led by Francisco Carneiro, Sr. Country Economist of the World Bank for Angola. The main author of the report was Prof. Emilson Silva, Tulane University, who carried out the main analysis under the supervision of the task manager. Camilla Rossaak and Paulo Loureiro provided valuable technical support at the various stages of this project. Much of the quality of this product is fruit of their efforts. The team wishes also to thank the financial support provided by the British Embassy at Luanda and the British Ambassadors, past and present, Mr. John Thompson and Mr. Ralph Publicover, respectively, for kindly sharing their time and thoughts about this important issue. The team kindly acknowledges the support from the Ministry of Finance through Vice-Minister Mr. Eduardo Severim de Morais, Mr. Manuel Neto da Costa, Director of Studies and International Relations, and Ms. Francisca Fortes, Director of Prices and Competition. Many thanks go also to the following governmental officials who welcomed us in their offices and provided us with data and other valuable information: Mr. Pedro Luis Fonseca, Director, Ministry of Planning, Mr. António da G. Lopes Teixeira, General Associate Director, Angola's Roads Institute (Instituto de Estradas de Angola (INEA)), Mr. Rui Augusto Tito, Vice Minister, Ministry of Energy and Water Affairs. Mr. Paulo Matos, Director, Ministry of Energy and Water Affairs, Mr. Bonifácio Manuel, Director, Ministry of Agriculture and Rural Development, Ms. Luzia B. da Costa, Head, Department of Planning and Statistics, Ministry of Petroleum, Mr. Mavinga B. David, Head, Department of Investments and Projects, Ministry of Petroleum, Ms. Efigénia da Purificação S. S. Martins, Head, Department of Public Enterprises, Ministry of Finance, Ms. Joana Cordeiro dos Santos, National Director of Accounting, Ministry of Finance, Mr. Simão Neto, Director, Office of Informatics, Ministry of Finance, Mr. Joaquim Flávio de Sousa Couto, General Director, National Statistics Institute (Instituto Nacional de Estatística (INE)), Mr. Domingos Bernardo, Technical Associate, Office of the President of the Administration Council, Water Public Enterprise (Empresa Pública de Águas (EPAL)) and Mr. José Ambriz, Administrator, Office of the President of the Administration Council, Water Public Enterprise (Empresa Pública de Águas (EPAL)). The civil society organizations, donors and professors who met with us during our consultations contributed to our learning about the topic and offered some new insights. The team really appreciates their donations of time and expertise to this venture. The non-governmental and civil society organizations consulted included FAS, FMEA, CARE ­ Angola, YME, CEEA, DW, SAL, ADRA, COIEPA, FONGA, MEDAIR, AIA, Angola 2000. The donors consulted included British Embassy, UNDP, US Embassy, South African Embassy, WHO, Swedish Embassy, German Embassy, Norwegian Embassy, Italian Embassy, Spanish Embassy, UNFPA, European Commission, UN Human Rights. The team also met with Mr. Paulo Filipe, professors Laurinda Hoygaard, University Agostinho Neto, and Justino Pinto de Andrade and Manuel Alves da Rocha, Catholic University of Angola. 3 The team is also highly indebted to Mr. Alexandre Rands, Mr. Alfredo Soares, and their co-workers at Datametrica, who helped polishing the questionnaire and subsequently coordinating and implementing all the data collection activities in Luanda as well as preparing the data set for posterior analysis. Datametrica data collection efforts were substantially augmented by their Angolan partners, Mr. Bernardo Vieira and his team at the Catholic University of Angola, and the numerous dedicated Angolan surveyors who worked diligently to deliver timely and high quality survey information. This work would not have been possible without the cooperation and assistance of many experts and staff at both World Bank headquarters in Washington and the local office in Luanda. Special thanks go to Louise Fox who was very generous in providing the team with data, advice and comments in the early stages of this study, both in Washington and Luanda. A great number of people at headquarters helped the team in the conceptualization of the study and on the drafting of the questionnaire. They also provided valuable feedback on earlier drafts of this report. The group of dedicated staff who helped us during the many stages of this work includes Ms. Masami Kojima, Mr. Stefano Paternostro and Ms. Sarah Keener. The team would like to thank the World Bank staff based in Luanda, in special Laurence Clarke, Olivier Lambert, Ana Maria Carvalho, Margarida Baessa Mendes, Domingas Pegado, Carla Balça and José Lima da Silva for not only helping us in any way they could, but also for making us feel at home in Luanda. 4 APPENDIX I: ANGOLAN FUEL SUPPLIES AND DEMANDS, 2000-2004 1. This section provides a brief overview of how quantities supplied and demanded of various fuels evolved in Angola during 2000-2004. The text refers to some of the important findings noted here. 2. Table A1 below informs us about domestic production of fuels from 2000 to 2004. It describes the sources of domestic production ­ products originated from the Refinery or from Cabinda Gulf Oil Company (CABGOC). During the period considered, the Refinery produced all fuel items. CABGOC produced gas, gasoline (except in 2000) and diesel, though at lower quantities than the Refinery. Within the five-year period examined, the production growth rates were: (i) gas, 4.79%; (ii) gasoline, 225.73%; (iii) kerosene, 62.70%; (iv) diesel, 81.88%; (v) light fuel, -55.93%; (vi) heavy fuel, 19.16%; and (vii) asphalt, 32.40%. In 2000, diesel and gasoline accounted for 66.05% and 15.48% of the total production at the refinery, respectively. In 2004, diesel and gasoline contribution rates were 65.44% and 15.36%, respectively. Kerosene production at the refinery became relatively more important within the five-year period. Of total production, kerosene production accounted for 4.78% in 2000 and 7.08% in 2004. In 2004, the three most important fuel items produced at the refinery, in terms of production magnitude, were diesel, gasoline and kerosene, in descending order of importance. Table A1: Domestic Production of Fuels (2000-2004) Refinery CABGOC Product 2000 2001 2002 2003 2004 2000 2001 2002 2003 2004 LPG* 34254 30270 32566 33645 28663 23085 29102 13028 18277 31423 Gasoline** 149026 165902 181816 180694 162584 0 51522 52718 150593 322833 Kerosene** 46060 57879 57095 76669 74941 0 0 0 0 0 Diesel** 636003 616000 564302 575012 692798 42525 205476 258510 464478 541288 L. Fuel* 23647 13352 18049 17647 10421 0 0 0 0 0 H. Fuel* 65132 69112 59775 76927 77611 0 0 0 0 0 Asphalt* 8762 8051 7893 5987 11601 0 0 0 0 0 Total 962884 960566 921496 966581 1058619 65610 286100 324256 633348 895544 Source: Sonangol * Volume expressed in tons; **volume expressed in cubic meters. 3. Table A2 shows the quantities of fuels available in Angola during 2000-2004 according to whether they were domestically produced or imported. Annual domestic production levels resulted from summing annual fuel quantities produced at the refinery and CABGOC ­ quantities presented in Table A1. We find that overall domestic production increased by 90% from 2000 to 2004. Diesel and gasoline production were the main driving forces behind this expansion ­ while diesel production grew by 81.88%, gasoline production expanded by 225.73%. Together, these two products accounted for 5 87.99% of total domestic fuel production in 2004. Only gas, kerosene and diesel were imported during the period. Total imports were small relative to total domestic production. Of total available fuel supply, given by the sum of total fuel quantity produced domestically and total fuel imported, total imported fuel quantity accounted for 7.64% in 2000, 8.63% in 2001, 8.45% in 2002, 9.65% in 2003 and 7.70% in 2004. Table A2: Domestic Production and Importation of Fuels (2000-2004) Domestic Production Imports Product 2000 2001 2002 2003 2004 2000 2001 2002 2003 2004 1241 LPG* 57339 59372 45594 51922 60086 0 19202 29335 42571 47831 Gasoline* * 149026 217424 234534 331287 485417 0 0 0 0 0 Kerosene 1092 ** 46060 57879 57095 76669 74941 1 22377 16393 21559 20998 103949 123408 6171 10674 Diesel** 678528 821476 822812 0 6 9 76166 69229 0 94209 L. Fuel* 23647 13352 18049 17647 10421 0 0 0 0 0 H. Fuel* 65132 69112 59775 76927 77611 0 0 0 0 0 Asphalt* 8762 8051 7893 5987 11601 0 0 0 0 0 102849 124666 124575 159992 195416 8505 11774 11495 17087 16303 Total 4 6 2 9 3 0 5 7 0 8 Source: Sonangol * Volume expressed in tons; **volume expressed in cubic meters. 4. In Table A3, we gather information about total available fuel supplies in Angola during 2000-2004. For the sake of comparison, we also included the domestic production levels in the table. From 2000 to 2004, total available fuel supply in Angola grew by 90.13%, resembling the growth in domestic fuel production. This is not surprising given that domestic production accounted for most of the available supplies during the period of study, as can be inferred from Tables A2 and A3. Table A3: Available Fuel Supplies (2000-2004) Domestic Production Available Fuel Supplies Product 2000 2001 2002 2003 2004 2000 2001 2002 2003 2004 LPG* 57339 59372 45594 51922 60086 69749 78574 74929 94493 107917 Gasoline** 149026 217424 234534 331287 485417 149026 217424 234534 331287 485417 Kerosene** 46060 57879 57095 76669 74941 56981 80256 73488 98228 95939 Diesel** 678528 821476 822812 1039490 1234086 740247 897642 892041 1146230 1328295 L. Fuel* 23647 13352 18049 17647 10421 23647 13352 18049 17647 10421 H. Fuel* 65132 69112 59775 76927 77611 65132 69112 59775 76927 77611 Asphalt* 8762 8051 7893 5987 11601 8762 8051 7893 5987 11601 Total 1028494 1246666 1245752 1599929 1954163 1113544 1364411 1360709 1770799 2117201 Source: Sonangol * Volume expressed in tons; **volume expressed in cubic meters. 5. Turning our attention to the demand side, we now examine the patterns of fuel consumption in Angola. Table A4 below describes the evolutions of main fuels' consumption levels in Angola and Angola's North region over 2000-2004. The North 6 region is composed of the following provinces: Luanda, Zaire, Uige, Cuanza Norte and Bengo. Luanda is undoubtedly the most important province and should account for most of the North's fuel consumption, according to Sonangol's officials. However, they were not able to provide us information about Luanda's actual participation rate. 6. Considering consumption levels of all fuels, we see that Angola's total consumption level grew steadily every year and peaked in 2004. The annual total consumption growth rates were 13.87% in 2001, 5.95% in 2002, 26.30% in 2003 and 11.76% in 2004. The North was Angola's largest consumer region over the entire period. Its contribution rates were 75.89% in 2000, 66.90% in 2001, 70.10% in 2002, 67.16% in 2003 and 64.65% in 2004. Table A4: Fuel Consumption Levels 2000 2001 2002 2003 2004 Product North Angola North Angola North Angola North Angola North Angola LPG* 41642 51815 52260 67188 56179 77443 62853 84906 57128 85407 Gasoline** 103734 141667 105663 150060 125062 180152 163586 244287 233099 348212 Kerosene** 39333 57695 47808 72020 51103 79807 55138 101440 50181 93431 Diesel** 376816 512009 405225 588487 412705 612467 505583 781357 524329 849734 L. Fuel* 6070 9515 7610 9829 6671 8928 6500 7731 5698 6503 H. Fuel* 70822 70822 75444 75453 62035 62035 74079 74145 59536 59536 Asphalt* 8767 9306 7461 8105 7484 8076 4986 5594 9017 9502 Total 647184 852829 649733 971142 721239 1028907 872726 1299460 938988 1452325 Source: Sonangol * Volume expressed in tons; **volume expressed in cubic meters. 7. In Angola, as well as in the North, diesel was the most consumed fuel. Considering first various fuels' consumption patterns for the country as a whole, one can easily verify that diesel's consumption participation rates were 60.04% in 2000, 60.60% in 2001, 59.53% in 2002, 60.13% in 2003 and 58.51% in 2004. Diesel's consumption level peaked in 2004 at 849734 m3, growing by 65.96% from 2000 to 2004. Over most of the period considered, the second, third and fourth most important fuels in terms of consumption magnitude were gasoline, kerosene and gas, respectively. Consumption levels for gasoline, gas and kerosene expanded by 145.80%, 64.83% and 61.40%, respectively, from 2000 to 2004. The annual consumption growth rate for gasoline increased over all years considered. They were 5.92% in 2001, 20.05% in 2002, 35.60% in 2003 and 42.54% in 2004. The annual consumption growth rates for kerosene were 24.83% in 2001, 10.81% in 2002, 27.11% in 2003 and -7.90% in 2004. Annual gas consumption growth rates were 29.67% in 2001, 15.26% in 2002, 9.64% in 2003 and 0.59% in 2004. 8. In the North, diesel consumption grew by 39.15% from 2000 to 2004. Its yearly participation rates were 58.22% in 2000, 62.37% in 2001, 57.22% in 2002, 57.93% in 2003 and 55.84% in 2004. Over most of the period considered, gasoline, gas and kerosene were respectively the second, third and fourth most important fuels in terms of consumption magnitude. Gas was relatively more important than kerosene as a consumption item in the North than in the nation as a whole. Consumption levels for 7 gasoline, gas and kerosene expanded by 124.71%, 37.19% and 27.58%, respectively from 2000 to 2004. These expansion rates were smaller than the ones observed for the country as a whole. It is also worth noting that gasoline consumption grew by a larger amount than diesel consumption in the North, similarly to what we observed in Angola. In fact, the relative rate of expansion was larger in the North than in Angola. 9. Having just concluded that fuel consumption growth rates in the North were more moderate than for the country as a whole, we now consider consumption patterns across regions. In addition to the North region, there are five other regions: Northeast, Center East, South, Cabinda and West. The Northeast is composed of four provinces, Malanje, Lunda Norte, Lunda Sul and Moxico. The Center East contains three provinces, Huambo, Bié and Cuando Cubango. Huila, Namibe and Cunene are the provinces that compose the South region. The Cabinda region contains only one province, with the same name. The West region consists of Benguela and Cuanza Sul provinces. Diesel 900000 800000 700000 600000 NORTHEAST CENTER EAST 500000 SOUTH m3 CABINDA 400000 WEST NORTH 300000 ANGOLA 200000 100000 0 2000 2001 2002 2003 2004 NORTHEAST 4006 1936 3680 14456 24214 CENTER EAST 4019 3470 13797 34970 41449 SOUTH 39244 43375 54632 58820 63689 CABINDA 29634 31868 32456 71701 84676 WEST 58290 102612 95196 95828 111378 NORTH 376816 405225 412705 505583 524329 ANGOLA 512009 588487 612467 781357 849734 Figure A1 10. Figures A1 through A4 will help us better understand regional consumption disparities for the four main fuel items; namely, diesel, gasoline, kerosene and gas. As Figure A1 makes it clear, in 2004, the three largest diesel consumers were the regions North, West and Cabinda, in descending order of importance. The North consumed almost five times as much as the West, and more than all regions considered together. From 2000 to 2004, the regional consumption growth rates were: (i) Northeast, 504.39%; (ii) Center East, 931.44%; (iii) South, 62.29%; (iv) Cabinda, 185.74%; (v) West, 91.08%; 8 and (vi) North, 39.15%. Annual consumption growth rates in the Northeast and Center East regions were large in 2002, 2003 and 2004, but their consumption contribution rates remained small in 2004. Gasoline 400000 350000 300000 NORTHEAST 250000 CENTER EAST CABINDA m3 200000 SOUTH WEST 150000 NORTH ANGOLA 100000 50000 0 2000 2001 2002 2003 2004 NORTHEAST 466 332 658 4233 9767 CENTER EAST 441 400 2630 7438 13797 CABINDA 17883 18216 21560 27664 27404 SOUTH 10026 13161 17025 23278 30601 WEST 9117 12287 13216 18088 33544 NORTH 103734 105663 125062 163586 233099 ANGOLA 141667 150060 180152 244287 348212 Figure A2 11. Figure A2 illustrates regional consumption patterns for gasoline. The North's dominating position is again very impressive. For 2004, we observe the following regional ranking for gasoline consumption, in descending order of importance: North, West, South, Cabinda, Center East and Northeast. The North consumed almost seven times as much as the West, and more than all regions considered together. From 2000 to 2004, the regional consumption growth rates were: (i) Northeast, 1997.64%; (ii) Center East, 3026.49%; (iii) Cabinda, 53.24%; (iv) South, 205.21%; (v) West, 267.94%; and (vi) North, 124.71%. As with diesel consumption, gasoline consumption in the Northeast and Center East regions skyrocketed within the five-year period. 12. Although the North was the leading consumer of kerosene throughout the period examined, regional consumption disparities were not as pronounced as in the cases of diesel and gasoline ­ see Figure A3 below. The regional ranking for kerosene consumption in 2004 in descending order of importance was: North, Cabinda, South, Center East, Northeast and West. Consumption in the North was over twice as large as in Cabinda, and more than in all other regions considered together. From 2000 to 2004, the regional consumption growth rates for kerosene were: (i) West, -53.95%; (ii) Northeast, 1073.61%; (iii) Center East, 212.67%; (iv) South, 299.10%; (v) Cabinda, 72.90%; and 9 (vi) North, 27.58%. Thus, the Northeast region accounted for much of the dynamism observed in kerosene consumption during the period of study. Kerosene 120000 100000 80000 WEST NORTHEAST CENTER EAST m3 60000 SOUTH CABINDA NORTH 40000 ANGOLA 20000 0 2000 2001 2002 2003 2004 WEST 1108 390 508 559 510 NORTHEAST 81 154 281 1623 952 CENTER EAST 597 263 522 1739 1865 SOUTH 1098 2660 2196 4714 4382 CABINDA 13228 16078 17187 25964 22871 NORTH 39333 47808 51103 55138 50181 ANGOLA 57695 72020 79807 101440 93431 Figure A3 10 13. Regional gas consumption patterns are illustrated in Figure A4. LPG 90000 80000 70000 60000 NORTHEAST CENTER EAST 50000 CABINDA N SOUTH TO 40000 WEST NORTH 30000 ANGOLA 20000 10000 0 2000 2001 2002 2003 2004 NORTHEAST 465 312 317 501 556 CENTER EAST 270 298 623 1851 2241 CABINDA 1810 2272 2442 2984 6254 SOUTH 3828 5565 6659 7814 8430 WEST 3800 6481 11222 8902 10799 NORTH 41642 52260 56179 62853 57128 ANGOLA 51815 67188 77443 84906 85407 Figure A4 14. The North dominating position is again very clear. In 2004, consumption in the North was over five times as large as in the West, the second largest consumer. The Northeast region was the lowest consumer. The gas consumption growth rate in this region within the five-year period was very small, 19.44%, in comparison to its spectacular consumption growth rates for the other three fuel items. From 2000 to 2004, the regional gas consumption growth rates were: (i) Northeast, 19.44%; (ii) Center East, 730.74%; (iii) Cabinda, 245.48%; (iv) South, 120.21%; (v) West, 184.19%; and (vi) North, 37.19%. 11 Light Fuel 12000 10000 8000 CABINDA CENTER EAST NORTHEAST N 6000 SOUTH TO WEST NORTH 4000 ANGOLA 2000 0 2000 2001 2002 2003 2004 CABINDA 0 0 0 0 0 CENTER EAST 0 0 0 0 0 NORTHEAST 0 6 0 0 0 SOUTH 2337 1823 1749 673 295 WEST 1108 390 508 559 510 NORTH 6070 7610 6671 6450 5698 ANGOLA 9515 9829 8928 7681 6503 Figure A5 15. The graph and table in Figure A5 above inform us about the evolution of the demand for light fuel in the various regions and the country as a whole. Quantities demanded declined over the period for all relevant regions and hence Angola. Similarly to the patterns observed for the other fuels already examined, the North region is dominant, with consumption shares higher than 63% over the entire period considered. Its consumption share rose from 66.80% in 2000 to 86.08% in 2004. In 2004, the North region consumed over eleven times as much as the West region, the second largest consumer in the nation. The consumption levels in Cabinda, Center East and Northeast regions were either nil or negligible. Consumption levels in the West and South regions fell by large amounts from 2000 to 2004. Relative to consumption levels in 2000, the consumption levels in 2004 for the West and South regions were 53.95% and 87.36% smaller, respectively. 16. For no other fuel, the North region's consumption level was as dominant as in the case of heavy fuel ­ see Figure A6. It essentially commanded the entire nation's demand! Angola's consumption of heavy fuel reached a peak in 2001, with a consumption level of 75453 tons. During the five-year period, consumption fell by 15.94%. The consumption level in 2004 represented 78.90% of the peak. 12 Heavy Fuel 80000 70000 60000 CABINDA 50000 WEST CENTER EAST N 40000 SOUTH TO NORTHEAST 30000 NORTH ANGOLA 20000 10000 0 2000 2001 2002 2003 2004 CABINDA 0 0 0 0 0 WEST 0 9 0 66 0 CENTER EAST 0 0 0 0 0 SOUTH 0 0 0 0 0 NORTHEAST 0 0 0 0 0 NORTH 70822 75444 62035 74079 59536 ANGOLA 70822 75453 62035 74145 59536 Figure A6 17. Figure A7 describes asphalt's consumption evolution during the period. Asphalt 10000 9000 8000 7000 NORTHEAST 6000 CENTER EAST CABINDA N 5000 SOUTH TO WEST 4000 NORTH ANGOLA 3000 2000 1000 0 2000 2001 2002 2003 2004 NORTHEAST 0 3 0 0 0 CENTER EAST 18 0 0 1 0 CABINDA 204 173 174 0 7 SOUTH 12 82 112 37 48 WEST 306 387 307 569 430 NORTH 8767 7461 7484 4986 9017 ANGOLA 9306 8105 8076 5594 9502 13 Figure A7 18. Asphalt's consumption level reached its peak in 2004, with the North region accounting for 94.90% of the country's consumption of 9502 tons. The increment rate in consumption relative to 2003 was significant, around 69.89%. The North's region dominant position prevailed throughout the entire period examined, with consumption shares larger than 92%. It is also interesting to note the sharp drop in Cabinda's consumption. While in 2000 the asphalt consumption level in this region was 204 tons, in 2004 it was only 7 tons! 19. It is now desirable to juxtapose quantities supplied and demanded within the five- year period. These pieces of information, which originate from Tables A3 and A4, are collected in Table A5 below. Close inspection of the information presented reveals that the quantities supplied generally exceed the quantities demanded. It follows that Angola typically increased its reserves of fuels during the period examined. Noteworthy exceptions to the rule occurred for kerosene, high fuel and asphalt. Kerosene's quantities demanded exceeded its quantities supplied in 2000, 2002 and 2003. The quantities demanded of high fuel and asphalt exceeded their respective quantities supplied in 2000, 2001 and 2002. Table A5: Fuels Supplied and Demanded in Angola (2000-2004) 2000 2001 2002 2003 2004 Product Supply Demand Supply Demand Supply Demand Supply Demand Supply Demand LPG* 69749 51815 78574 67188 74929 77443 94493 84906 107917 85407 Gasoline** 149026 141667 217424 150060 234534 180152 331287 244287 485417 348212 Kerosene** 56981 57695 80256 72020 73488 79807 98228 101440 95939 93431 Diesel** 740247 512009 897642 588487 892041 612467 1146230 781357 1328295 849734 L. Fuel* 23647 9515 13352 9829 18049 8928 17647 7731 10421 6503 H. Fuel* 65132 70822 69112 75453 59775 62035 76927 74145 77611 59536 Asphalt* 8762 9306 8051 8105 7893 8076 5987 5594 11601 9502 Total 1113544 852829 1364411 971142 1360709 1028907 1770799 1299460 2117201 1452325 Source: Sonangol. * Volume expressed in tons; **volume expressed in cubic meters. APPENDIX II: ECONOMETRIC ANALYSES 1. All statistical findings mentioned in the text originate from this section. There are four types of econometric analyses: (i) one whose goal is to demonstrate that the world price of crude oil is a key determinant of refined oil prices, governmental revenues and selected governmental expenditures; (ii) one which shows that fuel price hikes are important causes of inflation, even though most of the inflation is apparently explained by depreciation of the nominal exchange rate; (iii) one which tries to capture the direct effects of fuel and utility price changes on individuals demands of fuels and utilities; and (iv) one which investigates how fuel and utility budget shares vary with income. There are eight subsections, ranging from A to H. 14 A. The World Price of Crude Oil as Determinant of Refined Oil Product Prices and Estimated Diesel, Gasoline and Kerosene Subsidy Expenditures 2. In this subsection, we investigate how important the world price of crude oil is as a determinant of refined oil product (diesel, gasoline and kerosene) prices and of estimated diesel, gasoline and kerosene subsidy expenditures. 3. We start by examining how closely related refinery prices for diesel, gasoline and kerosene were to world prices for these fuels in selected months for which refinery prices were made available to us. Table A6 displays refinery and world prices for these items in April 2000, September 2000, June 2003, January 2004 and October 2004. The information available in the table reveals that for gasoline, kerosene and diesel refinery prices were closely and positively related to world prices. First, notice that while refinery gasoline and kerosene prices were consistently higher than gasoline and kerosene world prices, the refinery diesel price was consistently lower than diesel's world price. According to officials at the Ministry of Finance, this fact is easy to explain. The refinery was originally built to produce diesel. Hence, its relative efficiency in producing diesel is hardly surprising. Second, gasoline and kerosene refinery prices exceeded their world prices respectively by: (i) 9.60% and 8.80% in April 2000; (ii) 23.24% and 12.71% in September 2000; (iii) 21.93% and 35.65% in June 2003; (iv) 21.48% and 19.84% in January 2004; and (v) 21.13% and 7.40% in October 2004. Third, diesel's world price exceeded its refinery price by: (i) 38.30% in April 2000; (ii) 33.45% in September 2000; (iii) 16.57% in June 2003; (iv) 23.51% in January 2004; and (v) 40.17% in October 2004. Thus, for the five months studied here, the relationships between world and refinery prices appear to be stronger for gasoline than for kerosene and diesel, but even the weaker relationships seem to indicate that there may be a common factor which attaches refinery price adjustments to world fuel price changes. We conjecture that the common denominator is the world price of crude oil. Table A6: Gasoline, Kerosene and Diesel Refinery and World Prices (Kz) April 2000 September 2000 June 2003 January 2004 October 2004 Refinery World Refinery World Refinery World Refinery World Refinery World Product Price Price Price Price Price Price Price Price Price Price Gasoline (Lt) 1.37 1.25 4.03 3.27 20.52 16.83 25.51 21.00 38.69 31.94 Kerosene (Lt) 1.36 1.25 3.99 3.54 20.32 14.98 25.25 21.07 38.31 35.67 Diesel (Lt) 0.94 1.30 2.75 3.67 14.00 16.32 17.40 21.49 26.39 36.99 Sources: Ministry of Finance, Sonangol and Energy Information Administration, Department of Energy, United States Government (daily FOB fuel prices). 4. It seems plausible and logical to hypothesize that diesel, gasoline and kerosene world prices are positively related to the world price of crude oil. These hypotheses are tested below, utilizing monthly world prices for crude oil, diesel, gasoline and kerosene from January 2000 to December 2004, computed from daily FOB fuel prices. The data set was downloaded from the official website of the USA's Energy Information Administration, Department of Energy. 15 5. The results of three regressions are presented in Table A7. In each regression, the sole independent variable is the (log) price of crude oil. All estimators presented in the table are statistically significant, since their implied t statistics are all greater than 1.68 in absolute value. We adopt this convention throughout this report. The regressions suggest that a 10% increase in the price of a barrel of crude oil yields increases of 10.37% in the price of a liter of diesel, 9.46% in the price of a liter of gasoline and 10.29% in the price of a liter of kerosene. The relationships between world oil price and the world prices of three of its most important refined products, therefore, appear to be very close to 1 to 1. 16 Table A7: Relationships Between World Prices: Diesel, Gasoline and Kerosene vis-à-vis Crude Oil Dependent Variables Log Diesel Log Gasoline Log Kerosene Price Price Price Independent Coefficient Coefficient Coefficient Variable (S. E.) (S. E.) (S. E.) - 9.606* - 8.926* - 9.590* Constant (0.781) (0.854) (0.780) 1.037* 0.946* 1.029* Log Oil Price (0.099) (0.109) (0.099) Observations 60 60 60 F 108.96 75.93 107.59 Adjusted R2 0.65 0.56 0.64 * This symbol is used throughout the paper to inform the reader about statistically significant estimators. 6. Had we had access to refinery monthly prices for diesel, gasoline and kerosene during the entire five-year period, we would have been able to run similar regressions for these prices and the world price of crude oil. We would then expect to find qualitatively identical results, indicating that refinery fuel prices are closely and positively related to the world price of crude oil. Indeed, this finding appears to be an immediate implication of results for estimated relationships between estimates of subsidy amounts spent on the provision of diesel, gasoline and kerosene and the world price of crude oil examined below. 7. Based on average differential rates between monthly refinery and world price observations for diesel, gasoline and kerosene during eight months within 2000-2004, we constructed hypothetical refinery prices for these three products for the entire period departing from readily observable world FOB fuel prices. From such hypothetical monthly figures we were able to construct hypothetical Sonangol total prices for diesel, gasoline and kerosene, employing the fuel pricing structure explained above. Given the readily available actual prices for these products during the period in question, we could then build the series of hypothetical monthly unit subsidies. Given also the available annual fuel consumption data, previously discussed, we computed monthly averages related to consumption of diesel, gasoline and kerosene in Angola during the five-year period. Finally, by multiplying average monthly consumption by unit subsidy for each fuel considered we obtained hypothetical monthly subsidy amounts for each item. Table A8 below presents the hypothetical annual subsidy amounts which resulted from this exercise. It also reports annual subsidy amounts derived from applying Sonangol's fuel pricing formulas to world prices of diesel, gasoline and kerosene. Since world FOB fuel prices should be viewed as proxies for the real opportunity (marginal) costs faced by the refinery in selling its refined products in the domestic market instead of exporting them, we decided to also estimate monthly subsidy amounts departing from such prices. 17 Table A8: Estimates of Subsidy Amounts (US$ Millions) From World Fuel Prices From Hypothetical Refinery Fuel Prices Year Diesel Gasoline Kerosene Total Diesel Gasoline Kerosene Total 2000 292.25 52.02 19.15 363.42 387.99 80.53 22.91 468.52 2001 170.41 36.33 11.75 218.49 232.65 54.01 14.36 286.66 2002 178.12 50.20 12.12 240.44 246.83 73.40 15.23 320.23 2003 361.75 129.23 30.23 521.21 459.37 164.74 34.55 624.11 2004 314.88 204.18 124.81 643.87 402.61 262.22 142.91 664.83 Sources: Energy Information Administration, Department of Energy, United States Government (daily FOB fuel prices) and Sonangol (annual fuel consumption levels). 8. Both types of estimates reflect some facts about fuel consumption patterns in Angola that have already been noted. First, in terms of consumption magnitude, diesel was the most important fuel in the period in question. Second, the annual quantities of gasoline consumed were larger than those of kerosene. Third, gasoline consumption grew at a faster pace than consumptions of diesel and kerosene during the period. 9. The first set of estimates, calculated from world fuel prices, indicate that in 2004 government expenses associated with the subsidy scheme for the provision of diesel, gasoline and kerosene were US$ 314.88 million, US$ 204.18 million and US$ 124.81 million, respectively. Hence, of the estimated total bill faced by the government to subsidize provision of the three fuel products in 2004, diesel, gasoline and kerosene accounted for 48.90%, 31.71% and 19.38%, respectively. In 2000, the contribution rates were 80.42% for diesel, 14.31% for gasoline and 5.27% for kerosene. This result suggests that gasoline and kerosene subsidization increased in importance relative to diesel subsidization throughout the period. Indeed, according to our estimates, between 2000 and 2004, government expenses associated with subsidization of gasoline and kerosene grew by 292.50% and 551.75%, respectively, while they increased by 7.74% only in the case of diesel. We also see that the financing of subsidies for these three fuel items may have nearly doubled within five years. According to our estimates, the government expense grew by 77.17% between 2000 and 2004. Qualitatively identical results are found from inspection of the estimates calculated from hypothetical refinery fuel prices. 10. The results of eight regressions, which establish the statistical relationships between estimates of subsidy amounts and the world price of crude oil, are presented in Table A9. All estimators are statistically significant. Consider first regressions (1) through (4). The dependent variables are (logs of) subsidy amounts constructed from world prices of diesel, gasoline and kerosene. The regressions inform us that a 10% rise in the world price of crude oil leads to increases in the amounts expended by the government to subsidize: (i) diesel by 20.88%; (ii) gasoline by 34.21%; (iii) kerosene by 46.45%; and (iv) all three fuels together by 27.26%. The dependent variables in equations (5) through (8) are (logs of) subsidy amounts derived from hypothetical refinery prices. These equations suggest that a 10% rise in the world price of crude oil yields increases in the amounts expended by the government to subsidize: (i) diesel by 17.55%; (ii) gasoline by 28.95%; (iii) kerosene by 45.34%; and (iv) all three fuels together by 22.83%. Interesting to note that the goodness of fit for equations (5), (7) and (8), as measured by 18 the adjusted R2 statistic, are better than for their counterparts, equations (1), (3) and (4), respectively. Table A9: Relationships Between Subsidy Expenditures and World Price of Crude Oil Dependent Variables From World Fuel Prices From Hypothetical Refinery Prices Log Log Log Log Log Log Log Log Diesel Gasoline Kerosene Total Diesel Gasoline Kerosene Total Subsidy Subsidy Subsidy Subsidy Subsidy Subsidy Subsidy Subsidy (1) (2) (3) (4) (5) (6) (7) (8) Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variable (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) - 6.818* - 18.425* - 29.278* - 11.497* - 3.889* 13.955* - 28.262* - 7.680* Constant (2.506) (4.627) (6.465) (3.497) (2.013) (3.935) (5.404) (2.656) Log Crude 2.088* 3.421* 4.645* 2.726* 1.755* 2.895* 4.534* 2.283* Oil Price (0.319) (0.586) (0.821) (0.445) (0.256) (0.500) (0.687) (0.338) Observations 60 56 57 60 60 58 60 60 F 42.94 36.89 32.01 37.60 47.02 33.50 43.55 45.72 Adjusted R2 0.42 0.37 0.36 0.38 0.44 0.36 0.42 0.43 B. The World Price of Crude Oil, Nominal Exchange Rate and CPI as Determinant of Selected Governmental Revenues and Expenditures 11. In this subsection, we show that key governmental revenue and expenditure figures are positively related to the world price of crude oil, depreciation of the nominal exchange rate and to the inflation rate. To begin with, consider the models presented in Table A10. The crude oil price is a statistically significant estimator in all models. Explanatory powers range from 0.47 (for models (1) and (6)) to 0.62 (for model (4)). The results indicate that while a $1 increase in the price of a barrel of crude oil leads to an increase of approximately $1 in revenues, it generates much larger increases in subsidy expenditure and transfer payments, approximately $2.6 and $2.5, respectively. 19 Table A10: Governmental Revenue and Expenditure Regressions World Price of Crude Oil Estimates Dependent Variables Log Log Log Log Log Log Oil Tax Total Tax Total Subsidy Transfer Total Revenue Revenue Revenue Expenditure Payments Expenditure (1) (2) (3) (4) (5) (6) Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variable (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) 11.827* 11.805* 11.835* - 2.164 - 1.810 11.439* Constant (1.182) (1.088) (1.095) (2.261) (2.575) (1.637) Log Crude Oil 0.988* 1.022* 1.020* 2.562* 2.471* 1.057* Price (0.151) (0.139) (0.139) (0.288) (0.328) (0.209) Observations 48 48 48 48 48 48 F 43.06 54.43 53.46 79.18 56.77 25.68 Adjusted R2 0.47 0.53 0.53 0.62 0.54 0.47 12. The models in Table A11 build on the basic models presented in Table A10 by including the log of the nominal exchange rate as an additional independent variable. Except for model (4), the adjusted R2 statistics are higher for the expanded models in Table A11, implying that for all these models the inclusion of the log of the nominal exchange rate as an independent variable increases their explanatory powers. 20 Table A11: Governmental Revenue and Expenditure Regressions World Price of Crude Oil and Nominal Exchange Rate Estimates Dependent Variables Log Log Log Log Log Log Oil Tax Total Tax Total Subsidy Transfer Total Revenue Revenue Revenue Expenditure Payments Expenditure (1) (2) (3) (4) (5) (6) Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variable (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) 13.635* 14.069* 14.158* 1.502 0.843 14.743* Constant (1.240) (1.036) (1.033) (2.780) (2.429) (1.581) Log Crude Oil 0.665* 0.617* 0.604* 1.878* 2.024* 0.466* Price (0.175) (0.146) (0.146) (0.392) (0.342) (0.223) Log Nominal 0.188* 0.235* 0.241* 0.344* 0.312* 0.343* Exchange Rate (0.062) (0.051) (0.051) (0.138) (0.121) (0.079) Observations 48 48 48 48 48 48 F 30.04 49.33 49.97 34.68 47.82 27.40 Adjusted R2 0.55 0.67 0.68 0.59 0.67 0.53 13. It is also worth noting that the values of the estimates for the log of the world price of crude oil are lower in the models of Table A11 than for their counterparts in Table A10. Table A12: Governmental Revenue and Expenditure Regressions World Price of Crude Oil and Consumer Price Index Estimates Dependent Variables Log Log Log Log Log Log Oil Tax Total Tax Total Subsidy Transfer Total Revenue Revenue Revenue Expenditure Payments Expenditure (1) (2) (3) (4) (5) (6) Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variable (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) 14.000* 14.394* 14.473* 1.643 0.998 14.883* Constant (1.198) (0.981) (0.979) (2.789) (2.434) (1.578) Log Crude Oil 0.545* 0.494* 0.482* 1.766* 1.917* 0.354 Price (0.179) (0.146) (0.146) (0.416) (0.363) (0.235) Log Consumer 0.182* 0.217* 0.221* 0.289* 0.265* 0.288* Price Index (0.049) (0.040) (0.040) (0.114) (0.099) (0.064) Observations 48 48 48 48 48 48 F 34.44 58.52 59.10 34.96 48.38 28.15 Adjusted R2 0.59 0.71 0.71 0.59 0.67 0.54 14. The models displayed in Table A12 build on the basic models presented in Table A10 by including the log of the consumer price index as an explanatory variable. Again, we notice that the explanatory powers of the expanded models are higher than of their 21 counterparts, except for model (4). The values of the estimates for the log of the world price of crude oil are also smaller than the corresponding values in the basic models. C. Determinants of CPI, BBPI and MBPI: Nominal Exchange Rate and Fuel Prices 15. We now show that the nominal exchange rate and the weighted fuel price are important determinants of the three price indexes considered in this report. In Table A13, models (1), (3) and (5) are the regressions with the nominal exchange rate as the sole independent variable. Models (2), (4) and (6) are straightforward expansions of these simple models, since they are obtained by including a dummy to represent the shift in policy regime in the original models. Table A13: Nominal Exchange Rate as Determinants of CPI, BBI and MBI Dependent Variables Log CPI Log BBI Log MBI Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) Independent Variables (1) (2) (3) (4) (5) (6) 2.51* 2.72* - 0.34* 0.41* 0.21* 0.83* Constant (0.08) (0.07) (0.18) (0.14) (0.16) (0.15) 1.20* 1.12* 1.34* 1.15* 1.20* 1.04* Log Nominal Exchange Rate (0.02) (0.02) (0.04) (0.03) (0.04) (0.38) - 0.24* - 0.14* - 0.11* Dummy (Sep. 03 on = 1) (0.04) (0.02) (0.02) Number of Observations 60 60 33 33 33 33 F 3319.70 2765.48 1047.02 1635.17 968.05 1036.66 Adjusted R2 0.98 0.99 0.97 0.99 0.97 0.98 16. Consider model (1). The results suggest that the nominal exchange rate is an important determinant of the CPI: it is statistically significant and the model explains 98% of the variance in the CPI.1 We find that a 10% monthly depreciation rate generates a 12% monthly inflation rate. When we modify model (1) by including a dummy to account for the policy shift, the explanatory power of the model improves, reaching 99%. Model (2) informs us that a 10% monthly depreciation rate yields an 11.2% monthly inflation rate, a lower elasticity rate than in the first model. Since the dummy variable is statistically significant and its coefficient is positive, we conclude that monthly changes in the nominal exchange rate are more important in explaining monthly inflation rates in 1The results of the Dickey-Fuller test to verify the presence of unit root in the CPI and nominal exchange rate time series indicate that the variables co-integrate, being non-stationary in levels and stationary in first differences. The error terms are also stationary. These results enable us to find a long-run relationship between the variables in question, and not encountering a spurious regression. In addition, the Granger causality test informs us that we can reject the hypothesis that the nominal exchange rate does not cause the CPI, but cannot reject the hypothesis that the CPI does not cause the nominal exchange rate. Hence, it appears that the causation relation runs just in one direction; namely, the nominal exchange rate causes CPI. These tests are available from the author upon request. 22 the new regime than in the previous one. Although the large hikes in administered prices appear to have caused higher than average monthly inflation rates in May and November 2004, as we mentioned earlier, this result suggests that for the other months during the new regime the variance in the CPI was almost completely explained by the variance in the nominal exchange rate, and such relationship was stronger than the one observed in the period preceding the policy shift. 17. Close inspection of the other models reveals similar findings. First, the nominal exchange rate is an important determinant of each the BBPI and the MBPI. Second, the explanatory powers of the models with the policy shift dummy are higher. Third, the nominal exchange rate coefficients in regressions (4) and (6) are smaller than in regressions (3) and (5). Fourth, the dummy variable is statistically significant in each case and its coefficients are positive. We can thus conclude that monthly changes in the nominal exchange rate are more important in determining both baskets' monthly inflation rates in the Hard Kwanza regime than in the previous one. 18. The next regressions try to capture the direct effects that fuel price changes may have had on the three price indexes. As in the models examined above, we will also study the impacts, if any, brought about the structural change in policy regime. Given that all fuel prices rose simultaneously, we should expect that they are highly correlated with each other. This is indeed true, as the correlation matrix, Table A14, below shows. Table A14: Correlation Matrix for Prices of Fuels LPG Gasoline Kerosene Diesel L. Fuel H. Fuel Asphalt LPG 1.00 Gasoline 0.96 1.00 Kerosene 0.92 0.94 1.00 Diesel 0.75 0.87 0.89 1.00 L. Fuel 0.87 0.87 0.91 0.79 1.00 H. Fuel 0.57 0.60 0.76 0.60 0.81 1.00 Asphalt 0.80 0.83 0.90 0.81 0.99 0.86 1.00 19. Because fuel prices demonstrated a rising tendency when measured in kwanza terms but a declining one when measured in dollars, we run separate regressions with fuel prices expressed in kwanza and dollar terms for the same independent variables. As most explanatory variables are highly correlated with each other, we run separate regressions with (log) fuel prices as single independent variables and then later run separate regressions with the (log) weighted fuel price, expressed both in kwanza and dollar terms as single independent variables. 20. In Table A15 below, we see that the F-test indicates that each model is statistically significant. We also observe that the t-tests for the independent variables reveal that all of them, except kerosene and diesel prices measured in dollar terms, are statistically significant. In all regressions, the sign of the dummy variable is positive. This suggests that fuel prices have been relatively more important in explaining the behavior of the CPI in the new regime than in the previous one. The CPI is negatively affected by the weighted fuel price index when this variable is measured in dollar terms, but 23 positively affected when the independent variable is measured in kwanza terms. We claim that these disparate results follow from the fact that the measure in dollar terms is capturing the fall in the price index relative to the prices of other inputs, goods and services observed in the period prior to the Hard Kwanza one, while the other measure is capturing the relatively large changes in fuel prices in the new regime. The regressions on the various fuel prices reveal the same kinds of behavior. Table A15: Fuel Prices as Determinants of CPI Log CPI Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) 4.83* 4.83* 5.53* 4.50* 5.79* 5.09* 4.86* 4.25* 0.06 5.17* Constant (0.62) (0.13) (0.32) (0.20) (0.49) (0.12) (0.31) (0.21) (0.44) (0.13) Dummy 1.39* 0.48* 1.38* 0.69* 1.48* 0.69* 1.13* 0.69* 1.54* 0.60* (Sep 03 on = 1) (0.22) (0.14) (0.22) (0.16) (0.23) (0.14) (0.20) (0.15) (0.23) (0.15) Log Weighted - 0.76* Fuel Price (US$) (0.30) - - - - - - - - - Log Weighted 1.06* Fuel Price (Kz) - (0.08) - - - - - - - - Log Gasoline - 0.63* Price (US$) - - (0.22) - - - - - - - Log Gasoline 0.97* Price (Kz) - - - (0.09) - - - - - - Log Kerosene - 0.31 Price (US$) - - - - (0.25) - - - - - Log Kerosene 0.93* Price (Kz) - - - - - (0.08) - - - - Log LPG - 1.06* Price (US$) - - - - - - (0.21) - - - Log LPG 1.16* Price (Kz) - - - - - - - (0.11) - - Log Diesel - 0.17 Price (US$) - - - - - - - - (0.23) - Log Diesel 0.86* Price (Kz) - - - - - - - - - (0.08) Number of Observations 60 60 60 60 60 60 60 60 60 60 F 28.92 184.27 30.46 119.76 24.50 158.87 46.51 123.64 23.66 138.73 Adjusted R2 0.49 0.86 0.50 0.80 0.44 0.84 0.61 0.81 0.43 0.82 21. Models (2), (4), (6), (8) and (10) have higher explanatory powers than their counterparts. These models tell us that a 10% increase: (i) in the weighted fuel price yields a 10.6% inflation rate; (ii) in gasoline price leads to a 9.7% inflation rate; (iii) in kerosene price generates a 9.3% inflation rate; (iv) in LPG price causes a 11.6% inflation rate; and (v) in diesel price results in a 8.6% inflation rate. Hence, apparently, an increase in LPG price has the largest impact on the inflation rate. 24 22. We now consider our main econometric specifications, which possess the nominal exchange rate, the weighted fuel price and the policy shift dummy as independent variables. Table A16 informs us that all models are statistically significant, and all independent variables are statistically significant in each of the three models examined. We also observe similar patterns across the regressions. The coefficients for the dummy variable are all positive, implying that both the nominal exchange rate and the weighted fuel price are relatively more important in explaining the variances of the price indexes in the Hard Kwanza regime than in the previous one. We find that a 10% depreciation rate yields a 10.1% inflation rate measured by the CPI, an 11.4% inflation rate measured by the BBPI and a 10.2% inflation rate measured by the MBPI. Finally, the effects of a 10% increase in the weighted fuel price on the inflation rates measured by CPI, BBPI and MBPI are 1.4%, 0.5% and 0.8%, respectively. Table A16: Determinants of CPI, BBI and MBI: Nominal Exchange Rate and Weighted Fuel Price Dependent Variables Log CPI Log BBPI Log MBPI Coefficient Coefficient Coefficient Independent Variables (Std. Error) (Std. Error) (Std. Error) (1) (2) (3) 2.88* 0.35* 0.73* Constant (0.07) (0.13) (0.12) 1.01* 1.14* 1.02* Log Nominal Exchange Rate (0.03) (0.03) (0.03) 0.14* 0.05* 0.08* Log Weighted Fuel Price (Kz) (0.04) (0.02) (0.02) 0.23* 0.12* 0.09* Dummy (Sep. 03 on = 1) (0.03) (0.02) (0.02) Number of Observations 60 33 33 F 2322.02 1274.32 1129.04 Adjusted R2 0.99 0.99 0.99 D. Luanda's Estimated Demand Functions for Fuels and Utilities 23. Table A17 shows us estimated demand functions for water, gas, gasoline, diesel, coal, kerosene and electricity. The most representative demand functions, regarding the number of observations, are those for gas, water, electricity and gasoline in that order, with 983, 741, 599 and 185 observations, respectively. At the other end of the spectrum, we find the demand functions for kerosene, diesel and coal, with the demand for kerosene being the least representative. 24. All demand functions meet our expectations regarding own price and income effects. Household total expenditure on goods and services consumed is our measure of household income. The estimated water demand function tells us that a 10% increase in expenditure leads to an expansion of 4.61% in water quantity demanded while a 10% 25 increase in price leads to a 13.88% reduction in water quantity demanded. As for gas, the estimation is that increases of 10% in expenditure and price generate an expansion of 2.81% and a reduction of 2.30%, respectively, in gas quantity demanded. The estimations therefore suggest that the demand for water is more elastic, both in terms of income and price, than the demand for gas. 25. The other estimated fuel demand functions also offer interesting insights about behavioral responses in Luanda. While a 10% increase in expenditure yields a 4.59% rise in gasoline quantity demanded, it leads to a 6.31% increase in diesel quantity demanded. A 10% increase in the price of each product generates reductions of 6.01% in gasoline quantity demanded and of 15.34% in diesel quantity demanded. Hence, the estimations suggest that the demand for diesel is more responsive to changes in both income and price than the demand for gasoline. In addition, the estimated coal and kerosene demand functions inform us that these products' prices are not statistically significant estimators. A 10% increase in expenditure yields a 4.39% expansion in coal quantity demanded, but has a nil estimated impact on kerosene. 26 Table A17: Estimated Demand Functions Dependent Variables Log Log Log Log Log Log Log Water LPG Gasoline Diesel Coal Kerosene Electricity Quantity Quantity Quantity Quantity Quantity Quantity Quantity Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variables (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Constant 2.102* - 1.959* 2.038* 2.922 - 6.505* 1.068 1.244* (0.183) (0.219) (0.609) (2.043) (0.742) (1.746) (0.210) Log Expenditure 0.461* 0.281* 0.459* 0.631* 0.439* 0.028 0.435* (0.026) (0.013) (0.056) (0.172) (0.068) (0.173) (0.028) Log Water Price - 1.388* - - - - - - (0.035) Log LPG Price - - 0.230* - - - - - (0.031) Log Gasoline Price - - - 0.601* - - - - (0.234) Log Diesel Price - - - - 1.534* - - - (0.396) Log Coal Price - - - - 0.195 - - (0.149) Log Kerosene Price - - - - - - 0.135 - (0.399) Observations 741 983 185 51 135 28 599 F 1021.00 252.75 37.09 15.81 24.75 0.06 236.54 Adjusted R2 0.73 0.34 0.28 0.37 0.26 - 0.07 0.28 * Remember that this indicates that the estimator is statistically significant, since its implied t statistic is greater than 1.68 in absolute value. 26. Table A17 also informs us how the estimated electricity quantity demanded varies as expenditure changes. A 10% increase in expenditure leads to a 4.35% expansion in electricity quantity demanded. The survey did not attempt to elicit information about electricity prices. Hence, the estimated electricity demand function does not include the electricity price as an estimator. However, the estimated electricity demand functions described in Table A18 include prices of water and fuels. In carrying out these estimations, the rationale is twofold. First, it is important to find out whether the other products, water and fuels, are viewed as substitutes or complements to electricity. If viewed as substitutes, the signs of the price estimates should be positive. If viewed as complements, the signs should be negative. Second, it is also imperative to know which model provides the best fit for the behavior described by the data. 27. The first four models in Table A18 capture the estimated impacts that the prices of water and each of the fuel products, except kerosene, have on the electricity quantity demanded. The water and gas price estimators are statistically significant, but those for gasoline and diesel prices are not. Model (5) captures the estimated effects of water and gas prices, when considered together, on the quantity of electricity demanded. According to the adjusted R2 statistic, this is the model that best fits the data. The estimates suggest 27 that the quantity of electricity demanded expands by: (i) 4.62% when expenditure increases by 10%; (ii) 1.66% when the price of water rises by 10%; and (iii) 2.28% when the price of gas increases by 10%. The estimated positive price effects indicate that water and gas are viewed as substitutes to electricity. Table A18: Expanded Electricity Demand Functions Dependent Variables Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 1.022* - 0.640 1.160 1.658* - 0.642 - 0.015 (0.251) (0.473) (1.118) (0.617) (0.548) (1.789) Log Expenditure 0.458* 0.426* 0.497* 0.327* 0.462* 0.513* (0.034) (0.028) (0.070) (0.057) (0.034) (0.092) Log Water Price 0.190* - - - 0.166* 0.249* (0.048) (0.047) (0.137) Log LPG Price - 0.275* - - 0.228* 0.125 (0.063) (0.073) (0.185) Log Gasoline Price - - - 0.217 - - - 0.128 (0.513) (0.564) Log Coal Price - - - 0.093 - - (0.116) Observations 408 588 111 98 386 77 F 97.67 131.21 25.04 18.21 72.26 10.50 Adjusted R2 0.32 0.31 0.30 0.26 0.35 0.33 28. Remember that, in addition to electricity, water, gas and gasoline are the products whose estimated demand functions are the most representative. Using the same reasoning that motivated us to study the estimations in Table A18, we also consider the costs and benefits of expanding the estimated demand functions for water, gas and gasoline by including prices of other products as estimators. Table A19 collects these estimations. According to the adjusted R2 statistics, the best models that fit the data are models (1), (4) and (6). This is the case not only when we restrict our comparisons to the estimations present in Table A19, but also when we include the estimations for water, gas and gasoline of Table A17 in the groups of estimations used for comparisons. Model (1) of Table A19 informs us that gas price should be included as an estimator of the water quantity demanded even though it is not statistically significant. Likewise, model (4) tells us that water and gasoline prices should be included as estimators of gas quantity demanded even though they are not statistically significant. A similar pattern is also observed in model (6), in which water and gas prices are included as estimators of gasoline quantity demanded but they are not statistically significant. 29. A severe penalty, or cost, that is incurred by selecting models (1), (4) and (6) of Table A19 rather than their counterparts, however, is the substantial shrinkage in the 28 numbers of observations used in the estimations. This phenomenon is best captured by the F statistics ­ they are much smaller in Table A19 than in Table A17 for the estimated water, gas and gasoline demand functions. Hence, the estimations gain in terms of accuracy but lose power in terms of representing behavior of the overall population. Although the gains are very small in the estimations of water and gasoline demands, it is large in the estimation of gas demand. The costs, however, appear to outweigh the gains in all cases considered. Given this and also the fact that in models (1), (4) and (6) only own price estimators are statistically significant, henceforward we will restrict our attention to the estimated demand functions for water, gas, gasoline and diesel presented in Table A17 and the estimated electricity demand function (5) of Table A18. Table A19: Generalized Demand Functions for Selected Variables Dependent Variables Log Log Log Log Log Log Water Water LPG LPG Gasoline Gasoline Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 1.567* - 0.612 - 1.934* - 2.566* 0.314 0.841 (0.439) (1.191) (0.245) (0.635) (1.040) (1.102) Log 0.456* 0.371* 0.350* 0.429* 0.440* 0.446* Expenditure (0.027) (0.073) (0.015) (0.037) (0.058) (0.065) Log Water - 1.420* - 1.319* 0.048* - 0.031 - 0.042 Price (0.037) (0.094) (0.021) (0.051) (0.089) Log LPG 0.082 0.358* - 0.306* - 0.352* 0.270* 0.212 Price (0.061) (0.150) (0.034) (0.081) (0.127) (0.142) Log Gasoline - 0.403 - 0.128 - 0.595* - 0.701* Price (0.250) (0.136) (0.232) (0.234) Observations 657 129 675 135 171 132 F 625.85 56.49 182.44 34.63 25.33 16.32 Adjusted R2 0.74 0.63 0.45 0.50 0.30 0.32 E. Estimated Water and Gas Demand Functions for Luanda's Municipalities 30. Having considered Luanda's estimated demand functions for fuels and utilities, we now turn our attention to Luanda's municipalities. When we considered the estimated demand functions for the great Luanda area we had to restrict the scope of our analysis because of sample size limitations. For the municipalities, the scope of the analysis will be even more restrictive than before. We are forced to study the estimations for water and gas demand functions only, since they are the ones that provide us with consistently statistically significant estimators. 31. To facilitate discussion, the nine municipalities are divided into three groups of three each by alphabetical order: (i) Cacuaco, Cazenga and Ingombota; (ii) Kilamba, 29 Maianga and Rangel; and (iii) Samba, Sambizanga and Viana. Table A20 describes the estimated water and gas demand functions for the first group. The estimated demand functions for Cacuaco inform us that by increasing expenditure by 10% water and gas demands should expand by 4.40% and 4.01%, respectively. An increase of 10% in water price should lead to a reduction of 16.27% in water demand and an increase of 10% in gas price should reduce gas demand by 5.95%. In Cazenga, an increase of 10% in expenditure should expand water and gas demands by 4.02% and 2.75%, respectively. Water demand should fall by 16.29% if water price increases by 10%. The gas price estimator is not statistically significant. As for Ingombota, an increase of 10% in expenditure should yield increases of 4.31% and 2.51% in water and gas demands, respectively, while 10% increases in water and gas prices should lead to reductions of 10.96% and 2.70% in water and gas demands, respectively. 30 Table A20: Estimated Water and Gas Demand Functions for Cacuaco, Cazenga and Ingombota Municipalities Cacuaco Cazenga Ingombota Dependent Variables Dependent Variables Dependent Variables Log Log Log Log Log Log Independent Water LPG Water LPG Water LPG Variables Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) Constant 2.236* - 0.179 2.793* - 2.712* 1.993* - 1.436* (0.479) (0.717) (0.750) (1.077) (0.508) (0.494) Log Expenditure 0.440* 0.401* 0.402* 0.275* 0.431* 0.251* (0.072) (0.064) (0.111) (0.058) (0.068) (0.030) Log Water Price - 1.627* - - 1.629* - - 1.096* - (0.069) (0.135) (0.082) Log LPG Price - - 0.595* - - 0.144 - - 0.270* (0.103) (0.157) (0.068) Observations 86 82 54 70 140 220 F 344.87 28.39 100.25 11.46 108.49 38.64 Adjusted R2 0.89 0.40 0.79 0.23 0.61 0.26 Table A21: Estimated Water and Gas Demand Functions for Kilamba, Maianga and Rangel Municipalities Kilamba Maianga Rangel Dependent Variables Dependent Variables Dependent Variables Log Log Log Log Log Log Water LPG Water LPG Water LPG Quantity Quantity Quantity Quantity Quantity Quantity Independent Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Variables (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) Constant 1.597* - 2.178* 1.608* - 3.756* 2.800* - 1.646* (0.416) (0.811) (0.436) (0.333) (0.807) (0.969) Log Expenditure 0.549* 0.267* 0.506* 0.313* 0.387* 0.262* (0.059) (0.031) (0.062) (0.019) (0.102) (0.055) Log Water Price - 1.314* - - 1.303* - - 1.596* - (0.093) (0.095) (0.190) Log LPG Price - - 0.155 - - 0.011 - - 0.249* (0.124) (0.044) (0.126) Observations 90 131 129 202 35 60 F 152.25 40.45 124.43 133.97 52.84 13.21 Adjusted R2 0.77 0.38 0.66 0.57 0.75 0.29 32. The estimated water demand functions for Kilamba, Maianga and Rangel are price elastic, with the larger elasticity being observed in Rangel ­ an increase of 10% in water price leads to a 15.96% reduction in water demand ­ see Table A21. Only in 31 Rangel the gas price estimator is statistically significant. A 10% increase in gas price reduces gas demand by 2.62%. Table A22: Estimated Water and Gas Demand Functions for Samba, Sambizanga and Viana Municipalities Samba Sambizanga Viana Dependent Variables Dependent Variables Dependent Variables Log Log Log Log Log Log Independent Water LPG Water LPG Water LPG Variables Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) (Std. Err.) Constant 1.146 - 4.192* 2.563* - 1.572* 2.348* - 2.029* (0.929) (1.084) (0.462) (0.942) (0.463) (0.629) Log Expenditure 0.555* 0.341* 0.422* 0.367* 0.489* 0.284* (0.117) (0.056) (0.062) (0.051) (0.067) (0.034) Log Water Price - 1.391* - - 1.575* - - 1.630* - (0.149) (0.093) (0.073) Log LPG Price - 0.007 - - 0.376* - - 0.239* (0.134) (0.129) (0.091) Observations 46 59 65 94 96 106 F 58.56 18.98 181.59 26.70 356.58 36.45 Adjusted R2 0.72 0.37 0.85 0.35 0.88 0.40 33. The patterns described above are also observed in Samba, Sambizanga and Viana in what respects the elastic response of water quantities demanded to changes in water price and in Sambizanga and Viana concerning the small responses of gas quantities demanded to changes in gas price ­ see Table A22. Viana has the highest water demand price elasticity and the lowest gas demand price elasticity in the group ­ a 10% increase in water price reduces water quantity demanded by 16.30% and a 10% increase in gas price lowers gas demand by 2.39%. 34. In sum, our findings suggest that the responses to changes in income and water and gas prices of average residents of Luanda's municipalities are as follows: (i) in Cacuaco, an expansion of 10% in (a) income yields and increase of 1. 4.40% in water consumption; 2. 4.01% in gas consumption; (b) water price leads to a decrease of 16.27% in water consumption; (c) gas price leads to a decrease of 5.95% in gas consumption; (ii) in Cazenga, an expansion of 10% in (a) income yields an increase of 1. 4.02% in water consumption; 2. 2.75% in gas consumption; (b) water price leads to a decrease of 16.29% in water consumption; 32 (iii) in Ingombota, an expansion of 10% in (a) income yields an increase of 1. 4.31% in water consumption; 2. 2.51% in gas consumption; (b) water price leads to a decrease of 10.96% in water consumption; (c) gas price leads to a decrease of 2.70% in gas consumption; (iv) in Kilamba, an expansion of 10% in (a) income yields an increase of 1. 5.49% in water consumption; 2. 2.67% in gas consumption; (b) water price leads to a decrease of 13.14% in water consumption; (v) in Maianga, an expansion of 10% in (a) income yields an increase of 1. 5.06% in water consumption; 2. 3.13% in gas consumption; (b) water price leads to a decrease of 13.03% in water consumption; (vi) in Rangel, an expansion of 10% in (a) income yields an increase of 1. 3.87% in water consumption; 2. 2.62% in gas consumption; (b) water price leads to a decrease of 15.96% in water consumption; (c) gas price leads to a decrease of 2.49% in gas consumption; (vii) in Samba, an expansion of 10% in (a) income yields an increase of 1. 5.55% in water consumption; 2. 3.41% in gas consumption; (b) water price leads to a decrease of 13.91% in water consumption; (viii) in Sambizanga, an expansion of 10% in (a) income yields an increase of 1. 4.22% in water consumption; 2. 3.67% in gas consumption; (b) water price leads to a decrease of 15.75% in water consumption; (c) gas price leads to a decrease of 3.76% in gas consumption; (ix) in Viana, an expansion of 10% in (a) income yields an increase of 1. 4.89% in water consumption; 2. 2.84% in gas consumption; (b) water price leads to a decrease of 16.30% in water consumption (c) gas price leads to a decrease of 2.39% in gas consumption. F. Estimated Water and Gas Demand Functions for Quintile Groups 35. The poor segments of Luanda's population may behave quite differently than the rich segments. Hence, the removal of fuel and utility price subsidies may have different impacts on these groups. This section sheds some light on the likely impacts on demands of five population groups in Luanda, ranging from the poorest 20% to the richest 20% of 33 the population, brought about changes in income and prices of water and gas. As in the previous section, our analysis focuses on estimation of water and gas demand functions because of data limitations. Table A23: Estimated Water and LPG Demand Functions per Quintile First Quintile Second Quintile Third Quintile Fourth Quintile Fifth Quintile Dependent Dependent Dependent Dependent Dependent Variables Variables Variables Variables Variables Log Log Log Log Log Log Log Log Log Log Water LPG Water LPG Water LPG Water LPG Water LPG Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Constant 2.898* - 1.758* 0.992 0.605 2.830 - 2.331 2.808 - 0.985 1.190 - 4.706* (0.990) (0.359) (1.850) (1.035) (2.758) (1.585) (2.820) (1.462) (0.737) (0.763) Log Expen- 0.321* 0.004 0.631* - 0.107 0.358 0.415* 0.375 0.199 0.560* 0.466* diture (0.180) (0.029) (0.289) (0.154) (0.394) (0.222) (0.371) (0.182) (0.080) (0.050) Log - 1.337* - - 1.358* - - 1.348* - - 1.376* - - 1.524* - Water Price (0.083) (0.067) (0.073) (0.090) (0.084) Log - - 0.030 - - 0.244* - - 0.312* - - 0.287* - - 0.061 LPG Price (0.051) (0.061) (0.063) (0.066) (0.081) Observations 122 180 164 207 160 215 147 204 148 224 F 132.74 0.21 217 8.48 169.41 13.24 115.74 10.04 190.23 46.14 Adjusted R2 0.69 - 0.01 0.73 0.07 0.68 0.10 0.61 0.08 0.72 0.29 36. Table A23 displays the estimated water and gas demand functions by quintile. For the first quintile, water quantity demanded rises by 3.21% if expenditure increases by 10% and it falls by 13.37% if water price increases by 10%. The estimation of gas demand for this group is poor, as indicated by both F and adjusted R2 statistics. Neither expenditure nor gas price is a statistically significant estimator of gas demand. For the second quintile, water quantity demanded expands by 6.31% if expenditure increases by 10% but it falls by 13.58% if water price rises by 10%. While expenditure is not a statistically significant estimator of gas demand, gas price it is. A 10% increase in gas price reduces gas quantity demanded by 2.44%. 37. For the third and fourth quintiles, expenditure is not a statistically significant estimator of water demand. The expenditure estimator also fails to be statistically significant in the gas demand estimation for the fourth quintile. A 10% increase in expenditure yields an expansion of 4.15% in gas quantity demanded for the third quintile. As for price effects, we see that a 10% increase in water price reduces water quantity demanded by 13.48% for the third quintile and by 13.76% for the fourth quintile. In addition, a 10% increase in gas price lowers gas quantity demanded by 3.12% and 2.87% for the third and fourth quintiles, respectively. For the fifth quintile, a 10% increase in expenditure results in expansions of 5.60% and 4.66% in water and gas quantities 34 demanded, respectively, and a 10% increase in water price reduces water quantity demanded by 15.24%. The gas price estimator is not statistically significant. 38. Close inspection of the statistics presented in Table A23 reveals an apparent pattern regarding the water demand price elasticity. The water demand function becomes more elastic as we move from the first quintile to the second and then from the third to the fifth, with the only exception to the rule being observed when we move from the second to the third quintile. Even though the coefficient values of the water price estimator are almost identical for the first four quintiles, the coefficient value of the same estimator for the fifth quintile is substantially higher than its counterpart for the fourth quintile. G. Comparative Demand Analysis 39. In subsection D, the estimated demand functions for water, electricity, gas, gasoline and diesel were studied in detail. The reader was informed about price and income elasticities of demand. The analysis, however, was silent about demand comparisons across different groups of society. In subsections E and F, we were able to make comparisons. For each water and gas demands, we compared statistically significant price and income estimates across municipalities in subsection E and across quintile groups in subsection F. 40. In this subsection, a different type of comparison will be made. By using dummies to identify population characteristics, we are able to group households according to whether or not they share a particular characteristic. In each case, the resulting comparison will tell us, whenever the dummy is statistically significant, whether or not a group with a particular characteristic demands more or less of a given commodity than the average demanded quantities of the other household groups that do not possess the characteristic in question. 41. We start by investigating which municipalities consume more or less of each product. We shall refer to this analysis as "municipality effects." In fact, the terminology "effects" seems appropriate in each of the cases considered in what follows. The second subsection will consider quintile effects. We will then examine asset effects, residence type effects, water source effects, water collector effects, water collection distance effects, energy source utilization effects, energy source access effects, school transportation mode effects, health center transportation mode effects and expenditure on publicly provided services effects. This thorough analysis will provide us with important pieces of information, which will be useful later when we discuss the policy implications of all results obtained in our study. G1. Municipality Effects 42. A dummy for each municipality is used to identify residents of that municipality. If, for example, a household resides in Cacuaco this household is given the value "1" in the dummy for Cacuaco and the value "0" in the dummies for all other municipalities. 35 Tables A24 through A28 inform us which municipalities demand more or less of each product. It turns out that for each product there is at least one statistically significant municipality dummy. 43. In Table A24, we find nine estimated water demand functions. The estimations are distinguished according to the municipality dummy which is included as an explanatory variable in addition to expenditure and water price estimators. Notice that, because the dependent variables are measured in terms of logs, the coefficients of the dummy variables must be transformed prior to interpreting them in terms of percentage additions or subtractions relative to the average values of the respective counterparts. The coefficient values are transformed by taking anti-logs, subtracting the implied anti-logs values by one and then multiplying the results by one hundred. Although we do not report the results of the transformations in the tables that follow, we will refer to the transformed values when analyzing the estimated equations. 36 Table A24: Estimated Water Demand Functions Municipality Effects Dependent Variables Log Log Log Log Log Log Log Log Log Water Water Water Water Water Water Water Water Water Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 2.079* 2.016* 2.039* 2.059* 2.181* 2.106* 2.018* 2.114* 1.932* (0.190) (0.187) (0.182) (0.187) (0.187) (0.187) (0.188) (0.185) (0.184) Log 0.466* 0.472* 0.484* 0.467* 0.458* 0.462* 0.478* 0.458* 0.478* Expenditure (0.027) (0.026) (0.026) (0.026) (0.026) (0.027) (0.026) (0.026) (0.026) Log Water - 1.408* - 1.405* - 1.404* - 1.412* - 1.421* - 1.408* - 1.400* - 1.401* - 1.419* Price (0.035) (0.035) (0.035) (0.036) (0.035) (0.035) (0.035) (0.035) (0.035) 0.023 - - - - - - - - Cacuaco (0.091) - 0.334* - - - - - - - Cazenga (0.112) - - - 0.448* - - - - - - Ingombota (0.073) - - - 0.123 - - - - - Kilamba (0.089) - - - - - 0.243* - - - - Maianga (0.078) - - - - - 0.115 - - - Rangel (0.142) - - - - - - - 0.302* - - Samba (0.122) - - - - - - - 0.289* - Sambizanga (0.106) - - - - - - - - 0.484* Viana (0.084) Observations 713 713 713 713 713 713 713 713 713 F 672.57 683.87 721 674.92 684.97 673.72 680.32 681.95 714.77 Adjusted R2 0.74 0.74 0.75 0.74 0.74 0.74 0.74 0.74 0.75 44. The statistically significant expenditure and water price estimates vary slightly across estimations, with all price estimates being higher than the ones presented in Table A17. Of all municipality dummies, only those for Cacuaco, Kilamba and Rangel are not statistically significant. After transforming the coefficient values of the dummy variables, we find that Cazenga's water consumption is 39.65% higher than the average of the other eight municipalities. Likewise, the water consumption level in Sambizanga is 33.51% higher than the average consumption of the other eight municipalities and in Viana is 62.26% higher than the average consumption of the other eight municipalities. Ingombota (-56.52%), Maianga (-27.51%) and Samba (-35.26%) consume less water than the average consumption of their counterparts. 37 Table A25: Estimated LPG Demand Functions Municipality Effects Dependent Variables Log Log Log Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG LPG LPG LPG Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant - 2.001* - 1.974* - 1.992* - 2.023* - 1.957* - 1.958* - 1.991* - 1.992* - 1.873* (0.221) (0.218) (0.221) (0.217) (0.219) (0.219) (0.220) (0.224) (0.220) Log 0.283* 0.279* 0.280* 0.289* 0.281* 0.280* 0.285* 0.281* 0.281* Expenditure (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG - 0.226* - 0.223* - 0.226* - 0.233* - 0.229* - 0.229* - 0.228* - 0.225* - 0.239* Price (0.031) (0.031) (0.031) (0.030) (0.031) (0.031) (0.031) (0.032) (0.031) 0.094* - - - - - - - - Cacuaco (0.057) - - 0.241* - - - - - - - Cazenga (0.061) - - 0.044 - - - - - - Ingombota (0.038) - - - 0.240* - - - - - Kilamba (0.045) - - - - - 0.035 - - - - Maianga (0.039) - - - - - 0.057 - - - Rangel (0.068) - - - - - - - 0.129* - - Samba (0.067) - - - - - - - - 0.039 - Sambizanga (0.056) - - - - - - - - - 0.151* Viana (0.051) Observations 983 983 983 983 983 983 983 983 983 F 169.72 176.20 169.01 182.54 168.73 168.70 170.32 168.58 172.83 Adjusted R2 0.34 0.35 0.34 0.36 0.34 0.34 0.34 0.34 0.34 45. Table A25 describes nine estimated gas demand functions. As in the previous case, the expenditure and price estimators are all statistically significant and the estimates vary only so slightly across estimations. The dummies for Ingombota, Maianga, Rangel and Sambizanga are not statistically significant. The municipalities whose gas quantities demanded are higher than the average demands of their counterparts are Cacuaco (9.86%) and Kilamba (27.12%). Those whose gas quantities demanded are smaller than the average demands of their counterparts are Cazenga (-27.25%), Samba (-13.77%) and Viana (-16.30%). 38 Table A26: Estimated Gasoline Demand Functions Municipality Effects Dependent Variables Log Log Log Log Log Log Log Log Log Gasol. Gasol. Gasol. Gasol. Gasol. Gasol. Gasol. Gasol. Gasol. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 2.087* 2.070* 2.149* 2.304* 2.257* 2.062* 1.945* 2.038* 2.024* (0.619) (0.620) (0.610) (0.589) (0.630) (0.609) (0.617) (0.618) (0.611) Log 0.462* 0.463* 0.458* 0.430* 0.454* 0.472* 0.482* 0.462* 0.466* Expenditure (0.058) (0.058) (0.057) (0.055) (0.058) (0.057) (0.058) (0.057) (0.057) Log Gasoline - 0.637* - 0.632* - 0.612* - 0.672* - 0.671* - 0.643* - 0.616* - 0.622* - 0.638* Price (0.235) (0.239) (0.231) (0.222) (0.235) (0.230) (0.232) (0.233) (0.231) 0.028 - - - - - - - - Cacuaco (0.267) - 0.082 - - - - - - - Cazenga (0.229) - - - 0.332* - - - - - - Ingombota (0.154) - - - 0.841* - - - - - Kilamba (0.198) - - - - - 0.221 - - - - Maianga (0.180) - - - - - - 0.648* - - - Rangel (0.287) - - - - - - - 0.417* - - Samba (0.224) - - - - - - - 0.236 - Sambizanga (0.201) - - - - - - - - 0.450* Viana (0.220) Observations 171 171 171 171 171 171 171 171 171 F 23.23 23.29 25.43 31.77 23.94 25.64 24.86 23.88 25.21 Adjusted R2 0.28 0.28 0.30 0.35 0.29 0.30 0.30 0.29 0.30 46. In Table A26, nine estimated gasoline demand functions are considered. All expenditure and price estimators are statistically significant, and price estimates are higher than the ones presented in Table A17. The dummies for Cacuaco, Cazenga, Maianga and Sambizanga are not statistically significant. Kilamba (131.87%), Sambizanga (26.62%) and Viana (56.83%) each demands more gasoline than the average demands of its counterparts. Ingombota (-39.38%), Maianga (-24.73%) and Samba (- 51.74%) each demands less than the average demands of its counterparts. 39 Table A27: Estimated Diesel Demand Functions Municipality Effects Dependent Variables Log Log Log Log Log Log Log Log Log Diesel Diesel Diesel Diesel Diesel Diesel Diesel Diesel Diesel Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 0.742 0.526 0.731 1.055 0.715 0.701 1.071 1.104 1.017 (1.841) (1.994) (1.864) (1.797) (1.840) (1.829) (1.840) (1.710) (1.870) Log 0.683* 0.693* 0.678* 0.653* 0.691* 0.698* 0.671* 0.591* 0.634* Expenditure (0.150) (0.158) (0.150) (0.145) (0.152) (0.150) (0.147) (0.142) (0.160) Log Diesel - 0.991* - 0.957* - 0.978* - 1.049* - 1.007* - 1.005* - 1.071* - 0.826* - 0.971* Price (0.358) (0.366) (0.366) (0.348) (0.360) (0.355) (0.360) (0.335) (0.355) - 0.334 - - - - - - - - Cacuaco (1.098) - 0.156 - - - - - - - Cazenga (0.546) - - 0.018 - - - - - - Ingombota (0.380) - - - 0.656* - - - - - Kilamba (0.389) - - - - 0.362 - - - - Maianga (0.803) - - - - - - 0.667 - - - Rangel (0.783) - - - - - - 0.730 - - Samba (0.648) - - - - - - - - 0.953* - Sambizanga (0.346) - - - - - - - - 0.439 Viana (0.606) Observations 49 49 49 49 49 49 49 49 49 F 11.01 11.00 10.95 12.60 11.07 11.37 11.69 15.34 11.26 Adjusted R2 0.38 0.38 0.38 0.42 0.39 0.39 0.40 0.47 0.39 47. Nine estimated diesel demand functions are displayed in Table A27. They inform us that expenditure and price estimators are all statistically significant, although all price estimates are substantially smaller than the ones given in Table A17. Only two municipality dummies are statistically significant ­ Kilamba consumes 92.70% more diesel than the average consumption of its counterparts and Sambizanga consumes 159.35% less diesel than the average consumption of its counterparts. 40 Table A28: Estimated Electricity Demand Functions Municipality Effects Dependent Variables Log Log Log Log Log Log Log Log Log Electr. Electr. Electr. Electr. Electr. Electr. Electr. Electr. Electr. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant - 0.160 - 0.589 - 0.555 - 0.597 - 0.824 - 0.636 - 0.504 - 0.400 - 0.658 (0.558) (0.540) (0.548) (0.543) (0.546) (0.549) (0.541) (0.562) (0.549) Log 0.440* 0.446* 0.466* 0.465* 0.477* 0.460* 0.436* 0.461* 0.461* Expenditure (0.034) (0.034) (0.034) (0.034) (0.034) (0.035) (0.035) (0.034) (0.034) Log LPG 0.189* 0.241* 0.216* 0.223* 0.229* 0.229* 0.229* 0.191* 0.229* Price (0.073) (0.072) (0.073) (0.073) (0.073) (0.073) (0.072) (0.076) (0.073) Log Water 0.174* 0.161* 0.164* 0.179* 0.181* 0.165* 0.150* 0.177* 0.164* Price (0.047) (0.047) (0.047) (0.047) (0.047) (0.047) (0.047) (0.048) (0.047) - 0.455* - - - - - - - - Cacuaco (0.133) - - 0.710* - - - - - - - Cazenga (0.202) - - - 0.192* - - - - - - Ingombota (0.103) - - - - 0.366* - - - - - Kilamba (0.135) - - - - 0.279* - - - - Maianga (0.094) - - - - - 0.070 - - - Rangel (0.165) - - - - - - 0.464* - - Samba (0.129) - - - - - - - 0.216* - Sambizanga (0.119) - - - - - - - - 0.112 Viana (0.170) Observations 386 386 386 386 386 386 386 386 386 F 58.63 58.88 55.40 56.95 57.49 54.12 59.13 55.35 54.22 Adjusted R2 0.37 0.38 0.36 0.37 0.37 0.36 0.38 0.36 0.36 48. Estimated electricity demand functions are given in Table A28. All expenditure and price estimators are statistically significant. The dummies for Rangel and Viana are the only ones that are not statistically significant. Electricity consumption higher than the average consumption of its counterparts is observed in Maianga (32.18%), Samba (59.04%) and Sambizanga (24.11%). Electricity demand is lower than the average demand of its counterparts in Cacuaco (-57.62%), Cazenga (-103.40%), Ingombota (- 21.17%) and Kilamba (-44.20%). 41 49. In sum, relative to the average consumption of its counterparts: (i) Cacuaco consumes 9.86% more gas and 57.62% less electricity; (ii) Cazenga consumes 39.65% more water, 27.25% less gas and 103.40% less electricity; (iii) Ingombota consumes 56.52% less water, 39.38% less gasoline and 21.17% less electricity; (iv) Kilamba consumes 27.12% more gas, 131.87% more gasoline, 92.71% more diesel and 44.20% less electricity; (v) Maianga consumes 27.51% more water and 32.18% more electricity; (vi) Rangel consumes 91.17% less gasoline; (vii) Samba consumes 35.26% less water, 13.77% less gas, 51.74% less gasoline and 59.04% more electricity; (viii) Sambizanga consumes 33.51% more water, 159.35% less diesel and 24.11% more electricity; and (ix) Viana consumes 62.26% more water, 16.30% less gas and 56.83% more gasoline. G2. Quintile Effects 50. Having compared product consumption levels across municipalities, we now compare product consumption levels across quintiles. The statistically significant dummies will capture differences in behavior across income groups. Interestingly, there is no statistically significant quintile dummy in the estimated water and gasoline demand functions. This implies that water and gasoline consumption levels across quintiles are not sufficiently different from each other. 42 Table A29: Estimated LPG Demand Functions Quintile Effects Dependent Variables Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 2.130* - 1.941* - 1.951* - 1.998* - 1.760* - 1.622* (0.245) (0.221) (0.220) (0.220) (0.228) (0.348) Log 0.298* 0.278* 0.281* 0.286* 0.244* 0.246* Expenditure (0.017) (0.013) (0.013) (0.013) (0.017) (0.031) Log LPG - 0.224* - 0.229* - 0.229* - 0.227* - 0.225* - 0.225* Price (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) 0.085 - - - - - 0.146 Quintile 1 (0.055) (0.117) - - 0.030 - - - - 0.169* Quintile 2 (0.040) (0.086) - - - 0.036 - - - 0.154* Quintile 3 (0.038) (0.071) - - - - 0.067* - - 0.155* Quintile 4 (0.039) (0.059) - - - - 0.160* - Quintile 5 (0.052) Observations 983 983 983 983 983 983 F 169.55 168.60 168.81 169.79 173.12 86.35 Adjusted R2 0.34 0.34 0.34 0.34 0.34 0.34 51. In Table A29, we collect the results of six estimations for the gas demand function. The first five estimations include only one dummy, a dummy for each quintile. In the last estimation, four dummies are included; namely, the dummies for all quintiles except for the fifth one. The intention here is to compare the consumption levels of the four included groups with the consumption level of the excluded one ­ that is, we wish to find out whether each of the included groups consumes more or less than the richest 20% of the population. 52. The dummies for the first three quintiles when considered separately are not statistically significant. In the fourth estimation, we find that the fourth quintile consumes 6.93% less gas than the average consumption level of the other four quintiles. The fifth quintile consumes 17.35% more gas than the average consumption level of its counterparts. The last estimation implies that relative to the gas consumption level of the fifth quintile, the gas consumption levels of the second, third and fourth quintiles are lower by 18.41%, 16.65% and 16.77%, respectively. 43 Table A30: Estimated Diesel Demand Functions Quintile Effects Dependent Variables Log Log Log Log Log Log Diesel Diesel Diesel Diesel Diesel Diesel Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 0.505 2.825 - 0.120 0.758 1.777 6.778* (1.961) (1.831) (1.841) (1.844) (2.218) (3.307) Log 0.704* 0.396* 0.755* 0.678* 0.521* - 0.049 Expenditure (0.166) (0.167) (0.150) (0.149) (0.242) (0.351) Log Diesel - 0.974* - 0.886* - 0.936* - 0.981* - 0.985* - 0.893* Price (0.357) (0.328) (0.345) (0.356) (0.354) (0.326) 0.305 - - - - - 2.045 Quintile 1 (0.861) (1.280) - - 1.385* - - - - 2.468* Quintile 2 (0.472) (0.874) - - 0.857* - - - 0.555 Quintile 3 (0.471) (0.718) - - - - 0.063 - - 0.768 Quintile 4 (0.396) (0.488) - - - - 0.412 - Quintile 5 (0.501) Observations 49 49 49 49 49 49 F 11.03 15.92 12.86 10.97 11.34 8.73 Adjusted R2 0.39 0.48 0.43 0.38 0.39 0.49 53. Table A30 describes estimated diesel demand functions. When considered separately, the dummies for the first, fourth and fifth quintiles are not statistically significant. In the second estimation, we find that the second quintile demands 299.48% less diesel than the average of its counterparts. The third quintile, however, consumes 135.61% more diesel than the average of its counterparts. In the last estimation, we find that the second quintile consumes 1079.88% less diesel than the fifth quintile! 54. Table A31 describes estimated electricity demand functions. Only the dummy for the second quintile is statistically significant. The second quintile consumes 26.24% less electricity than the average of its counterparts. 44 Table A31: Estimated Electricity Demand Functions Quintile Effects Dependent Variables Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 0.886 - 0.420 - 0.697 - 0.578 - 0.812 - 0.754 (0.607) (0.555) (0.552) (0.552) (0.594) (0.936) Log 0.481* 0.430* 0.466* 0.459* 0.492* 0.466* Expenditure (0.040) (0.037) (0.035) (0.035) (0.053) (0.084) Log Water 0.165* 0.169* 0.167* 0.168* 0.167* 0.170* Price (0.047) (0.047) (0.047) (0.047) (0.047) (0.047) Log LPG 0.241* 0.237* 0.228* 0.219* 0.224* 0.235* Price (0.075) (0.073) (0.073) (0.074) (0.073) (0.075) 0.141 - - - - 0.134 Quintile 1 (0.152) (0.307) - - 0.233* - - - - 0.118 Quintile 2 (0.109) (0.234) - - 0.081 - - 0.099 Quintile 3 (0.095) (0.187) - - - 0.092 - 0.106 Quintile 4 (0.097) (0.153) - - - - - 0.102 - Quintile 5 (0.138) Observations 386 386 386 386 386 386 F 54.39 55.85 54.33 54.40 54.27 31.77 Adjusted R2 0.36 0.36 0.36 0.36 0.36 0.36 55. As in the previous subsection, we provide a summary of the results below. Relative to the average consumption level of its counterparts: (i) the second quintile consumes 299.48% less diesel and 26.24% less electricity; (ii) the third quintile consumes 135.61% more diesel; (iii) the fourth quintile consumes 6.93% less gas; and (iv) the fifth quintile consumes 17.35% more gas. Relative to the consumption level of the fifth quintile: (i) the second quintile consumes 18.41% less gas and 1079.88% less diesel; (ii) the third quintile consumes 16.65% less gas; and (iii) the fourth quintile consumes 16.77% less gas. 45 G3. Asset Effects 56. The previous subsection compared individual behavior according to income. We now make behavioral comparisons based on assets. There are seven types of assets: (i) radio; (ii) television; (iii) refrigerator; (iv) automobile; (v) computer; (vi) residential property; and (vii) generator. Of 1,108 households, 989 own radios, 858 own televisions, 683 own refrigerators, 218 own automobiles, 157 own computers, 722 own residential properties and 207 own generators. The idea here is that possession of certain types of assets, such as automobile, residential property, computers and generators, may be viewed as a proxy for wealth. In addition, as some households among the poor are known to possess radios, televisions and refrigerators, one may argue that households who do not have such items generally face extreme levels of poverty. It seems logical, therefore, to make two hypotheses: (i) the higher a household's wealth, as demonstrated by its possession of automobiles, residential property, computers or generators, the higher should be its demands of fuels and utilities; and (ii) households at extreme levels of poverty, as measured by lack of radio, television or refrigerator possession, should consume less of all products than their counterparts. 46 Table A32: Estimated Water Demand Functions Assets Effects Dependent Variables Log Log Log Log Log Log Log Water Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) (7) Constant 2.332* 2.218* 2.074* 1.944* 2.090* 1.986* 1.990* (0.204) (0.184) (0.184) (0.215) (0.187) (0.190) (0.202) Log Expenditure 0.467* 0.483* 0.487* 0.473* 0.465* 0.465* 0.468* (0.026) (0.026) (0.026) (0.027) (0.026) (0.026) (0.026) Log Water Price - 1.404* - 1.411* - 1.409* - 1.409* - 1.408* - 1.417* -1.411* (0.035) (0.035) (0.035) (0.035) (0.035) (0.036) (0.036) Radio - 0.281* - - - - - - (0.010) Television - - 0.346* - - - - - (0.066) Refrigerator - - - 0.241* - - - - (0.059) Automobile - - - 0.107 - - - (0.079) Computer - - - - 0.006 - - (0.094) Real Estate - - - - - 0.152* - (0.063) Generator - - - - - - 0.095 (0.075) Observations 713 713 713 713 713 713 713 F 682.74 707.69 693.55 674.80 672.49 680.09 674.54 Adjusted R2 0.74 0.75 0.74 0.74 0.74 0.74 0.74 57. Estimated demand functions for gasoline and diesel produced no statistically significant dummy. Table A32 informs us about seven estimated water demand functions. All expenditure and price estimators are statistically significant. The price estimates are all higher than the ones presented in Table A17. The dummies for automobile, computer and generator are not statistically significant. While the dummy estimate for homeownership is consistent with our first hypothesis, the dummy estimates for radio, television and refrigerator refute our second hypothesis. Homeowners consume 16.42% more water than non-owners. However, those who possess radios consume 32.45% less water than those who do not, those who have televisions consume 41.34% less water than those who do not, and those who have refrigerators consume 27.25% less water than those who do not. 47 Table A33: Estimated LPG Demand Functions Assets Effects Dependent Variables Log Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) (7) Constant - 2.006* - 1.974* - 1.961* - 2.121* - 1.994* - 1.972* - 1.984* (0.225) (0.220) (0.219) (0.225) (0.219) (0.220) (0.223) Log Expenditure 0.281* 0.280* 0.283* 0.290* 0.287* 0.282* 0.282* (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG Price - 0.229* - 0.231* - 0.228* - 0.229* - 0.229* - 0.231* - 0.229* (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) Radio 0.051 - - - - - - (0.054) Television - 0.032 - - - - - (0.040) Refrigerator - - - 0.039 - - - - (0.033) Automobile - - - 0.118* - - - (0.039) Computer - - - - - 0.100* - - (0.044) Real Estate - - - - - 0.026 - (0.033) Generator - - - - - - 0.023 (0.039) Observations 983 983 983 983 983 983 983 F 168.78 168.65 169.03 173.01 170.84 168.64 168.51 Adjusted R2 0.34 0.34 0.34 0.34 0.34 0.34 0.34 58. Table A33 presents seven estimated gas demand functions. Expenditure and price estimators are all statistically significant. Automobile and computer dummies are the only ones that are statistically significant. Automobile owners demand 12.52% more gas than non-owners and computer owners demand 10.52% less gas than non-owners. 59. Estimated electricity demand functions are displayed in Table A34. All expenditure and price estimators are statistically significant. Refrigerator, automobile and computer are the statistically significant dummy estimators. Relative to the electricity consumption levels of non-owners, the electricity consumption levels of owners of refrigerators and computers are in excess of 37.85% and 26.62%, respectively, and the electricity consumption level of owners of automobiles falls short by 19.24%. 48 Table A34: Estimated Electricity Demand Functions Assets Effects Dependent Variables Log Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant - 0.419 - 0.379 - 0.910* - 0.469 - 0.609 - 0.618 - 0.627 (0.569) (0.575) (0.548) (0.552) (0.545) (0.549) (0.572) Log Expenditure 0.468* 0.467* 0.455* 0.451* 0.453* 0.460* 0.462* (0.035) (0.034) (0.034) (0.035) (0.034) (0.035) (0.035) Log Water Price 0.165* 0.166* 0.178* 0.165* 0.163* 0.169* 0.166* (0.047) (0.047) (0.047) (0.047) (0.047) (0.048) (0.047) Log LPG Price 0.223* 0.221* 0.237* 0.223* 0.226* 0.233* 0.228* (0.073) (0.073) (0.072) (0.073) (0.073) (0.074) (0.074) - 0.243 - - - - - - Radio (0.170) - - 0.264 - - - - - Television (0.178) - - 0.321* - - - - Refrigerator (0.101) - - - - 0.176* - - - Automobile (0.090) - - - - 0.236* - - Computer (0.100) - - - - - - 0.061 - Real Estate (0.087) - - - - - - - 0.010 Generator (0.105) Observations 386 386 386 386 386 386 386 F 54.85 54.91 58.01 55.56 56.22 54.25 54.06 Adjusted R2 0.36 0.36 0.37 0.36 0.36 0.36 0.36 60. Therefore, the results do not support either hypothesis. It is possible that our proxies for wealth and extreme poverty are inadequate, as poor and wealthy households may be endowed with automobiles, residential property, computers and generators and wealthy households may not possess radios, televisions or refrigerators. In any event, we conclude that relative to non-owners: (i) radio owners consume 32.45% less water; (ii) television owners consume 41.34% less water, (iii) refrigerator owners consume 27.25% less water and 37.85% more electricity; (iv) automobile owners consume 12.52% more gas and 19.24% less electricity; (v) computer owners consume 10.525 less gas and 26.62% more electricity; and (vi) homeowners consume 16.42% more water. 49 G4. Residence Type Effects 61. Comparisons among groups of consumers distinguished by their type of residence may reveal some valuable pieces of information for policy making, particularly for taxing or taxing exemption purposes. In our sample, there are 629 houses, 197 apartments, 153 multiple-floor houses and 128 annexes. Table A35: Estimated Water Demand Functions Residence Type Effects Dependent Variables Log Log Log Log Water Water Water Water Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) Constant 2.076* 1.998* 2.017* 2.102* (0.187) (0.186) (0.202) (0.186) Log 0.466* 0.484* 0.471* 0.460* Expenditure (0.026) (0.026) (0.027) (0.026) Log Water - 1.408* - 1.390* - 1.405* - 1.414* Price (0.036) (0.036) (0.036) (0.036) 0.049 - - - Annex (0.091) - - 0.298* - - Apartment (0.084) - - 0.051 - House (0.062) Multiple-Floor - - - 0.167* House (0.087) Observations 712 712 712 712 F 673.16 688.81 673.68 677.49 Adjusted R2 0.74 0.74 0.74 0.74 62. Table A35 presents four estimated water demand functions. The dummies for annex and house are not statistically significant. Relative to the average water consumption levels of their respective counterparts, the water consumption level is: (i) 34.72% lower for those who reside in apartments; and (ii) 18.18% higher for those who reside in multiple-floor houses. 63. Table A36 provides us with information about estimated gas and gasoline demand functions. From equations (1) through (4), we can see that the sole statistically significant dummy in the estimated gas demand functions is the house dummy. Accordingly, those households who reside in houses consume 7.25% more gas than the average gas consumption level of their counterparts. As for the estimated gasoline demand functions, equations (5) through (8), we observe that the apartment dummy is the only one that is 50 statistically significant. Those who reside in apartments consume 32.18% less gasoline than those who do not. Table A36: Estimated LPG and Gasoline Demand Functions Residence Type Effects Dependent Variables Log Log Log Log Log Log Log Log LPG LPG LPG LPG Gasoline Gasoline Gasoline Gasoline Quantity Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) Constant - 1.962* - 1.985* - 2.084* - 1.959* 2.079* 2.079* 2.020* 2.033* (0.219) (0.220) (0.226) (0.220) (0.622) (0.612) (0.632) (0.617) Log 0.282* 0.284* 0.287* 0.281* 0.463* 0.473* 0.467* 0.458* Expenditure (0.013) (0.013) (0.013) (0.013) (0.058) (0.057) (0.059) (0.057) Log LPG - 0.229* - 0.228* - 0.223* - 0.230* - - - - Price (0.031) (0.031) (0.031) (0.031) Log Gasoline - - - - - 0.632* - 0.639* - 0.638* - 0.614* Price (0.234) (0.232) (0.234) (0.233) - 0.078 - - - - 0.044 - - - Annex (0.050) (0.256) - - 0.062 - - - - 0.279* - - Apartment (0.040) (0.157) - - 0.070* - - - 0.068 - House (0.032) (0.132) Multiple-Floor - - - 0.001 - - - 0.210 House (0.044) (0.162) Observations 983 983 983 983 171 171 171 171 F 169.30 169.28 169. 169. 23.24 24.72 23.78 24.02 Adjusted R2 0.34 0.34 0.34 0.34 0.28 0.30 0.28 0.29 64. Three estimated electricity demand functions are shown in Table A37. Apartment and house dummies are statistically significant. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 58.72% higher for those who reside in apartments; and (ii) 27.38% lower for those who reside in multiple-floor houses. 65. Thus, relative to the average consumption of their counterparts: (i) apartment residents consume 34.72% less water, 32.18% less gasoline and 58.72% more electricity; (ii) house residents consume 7.25% more gas; and (iii) multiple-floor-house residents consume 18.18% more water and 27.38% less electricity. 51 G5. Water Source Effects 66. The difficulty of accessing publicly delivered services, such as water and electricity, may substantially reduce overall household consumption of these essential goods due to additional marketed and non-marketed costs faced by households who have limited or no access to these services. These households may have to procure the needed products in secondary or black markets and/or spend considerable amounts of time waiting to be serviced, searching or gathering water and energy items from alternative sources. Since the additional costs increase the actual price faced by a household to consume any one of such essential commodities and also lower real household income due to the taxing effects that searching, gathering and waiting times have on earnings and leisure, as access costs rise, household consumption of a hardly accessible good should fall (negative own price effect), consumption of any close substitute, if there is any at all, should rise (positive cross price effect) and consumption of all other goods and services should fall (income effect). Table A37: Estimated Electricity Demand Functions Residence Type Effects Dependent Variables Log Log Log Electric. Electric. Electric. Quantity Quantity Quantity Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) Constant - 0.504 - 0.269 - 0.643 (0.532) (0.557) (0.548) Log 0.443* 0.441* 0.464* Expenditure (0.034) (0.035) (0.034) Log Water 0.113* 0.136* 0.169* Price (0.047) (0.048) (0.047) Log LPG 0.211* 0.215* 0.228* Price (0.071) (0.073) (0.073) 0.462* - - Apartment (0.091) - - 0.242* - House (0.082) Multiple-Floor - - - 0.094 House (0.118) Observations 386 386 386 F 64.11 57.48 54.30 Adjusted R2 0.40 0.37 0.36 67. In this subsection, we study some of the effects caused by lack of access to publicly provided water service. The majority of the sampled population does not have access to tap water at home. Of 1,108 households, only 401 have the luxury of receiving water at home through EPAL's piping system. 327 households consume water from 52 public taps, 273 households purchase it from water trucks, 69 households consume it from their neighbors, 37 households purchase it from small street vendors and only one household gathers water from a well. 68. In what follows, we will compare water and electricity consumption levels of these various groups. Unfortunately, the estimations for the other products do not produce statistically significant dummies. Table A38 gives us six estimated water demand functions. The dummies for public taps and wells are not statistically significant. Relative to the average consumption levels of their respective counterparts, the water consumption level is: (i) 26.36% higher for those who purchase water from water trucks; (ii) 51.59% higher for those who purchase from small street vendors; and (iii) 30.87% lower for those who consume tap water at home. Table A38: Estimated Water Demand Functions Water Source Effects Dependent Variables Log Log Log Log Log Log Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 2.175* 2.029* 2.072* 2.101* 2.093* 2.035* (0.191) (0.183) (0.183) (0.183) (0.183) (0.182) Log 0.455* 0.462* 0.464* 0.460* 0.463* 0.482* Expenditure (0.026) (0.025) (0.025) (0.026) (0.026) (0.026) Log Water - 1.382* - 1.351* - 1.376* - 1.389* - 1.389* - 1.359* Price (0.035) (0.036) (0.035) (0.035) (0.035) (0.035) - 0.083 - - - - - Public Taps (0.062) - 0.234* - - - - Trucks (0.065) - - 0.416* - - - Small Vendors (0.157) - - - 0.242* - - Neighbors (0.144) - - - - - 0.659 - Wells (0.787) - - - - - - 0.269* Home Taps (0.066) Observations 741 741 741 741 741 741 F 681.98 696.12 688.52 683.30 680.63 700.64 Adjusted R2 0.73 0.74 0.74 0.73 0.73 0.74 69. Table A39 collects five estimated electricity demand functions. The dummies for water trucks and small street vendors are not statistically significant. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 76.47% lower for those who consume water from public taps; 53 (ii) 41.76% lower for those who consume water available at neighboring residences; and (iii) 60.80% higher for those who consume tap water at home. 70. In sum, relative to the average consumption of their counterparts: (i) households whose primary water source is public taps consume 76.47% less electricity; (ii) households who mainly purchase water from water trucks consume 26.36% more water; (iii) households who mostly purchase water from street vendors consume 51.59% more water; (iv) households who have access to water from their neighbors consume 41.76% less electricity; (v) households who have tap water at their homes consume 30.87% less water and 60.80% more electricity. 54 Table A39: Estimated Electricity Demand Functions Water Source Effects Dependent Variables Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) Constant - 0.117 - 0.639 - 0.634 - 0.593 - 0.477 (0.534) (0.549) (0.549) (0.546) (0.525) Log 0.429* 0.463* 0.463* 0.465* 0.450* Expenditure (0.033) (0.035) (0.035) (0.034) (0.033) Log Water 0.170* 0.164* 0.168* 0.173* 0.101* Price (0.045) (0.049) (0.048) (0.047) (0.046) Log LPG 0.190* 0.228* 0.226* 0.220* 0.186* Price (0.070) (0.073) (0.074) (0.073) (0.070) - 0.568* - - - - Public Taps (0.093) - - 0.017 - - - Trucks (0.093) - - 0.080 - - Small Vendors (0.240) - - - - 0.349* - Neighbors (0.182) - - - - 0.475* Home Taps (0.078) Observations 386 386 386 386 386 F 68.61 54.06 54.10 54.06 55.50 Adjusted R2 0.41 0.36 0.36 0.36 0.37 G6. Water Collector Effects 71. Time spent collecting water may be very costly to households, especially if water collectors have to divert time away from earning or schooling activities in order to fetch water. The survey reveals that water collection is undertaken by various members of the household, and sometimes by individuals other than household members. Of 1,107 households, 456 make this service the responsibility of male teenagers, 254 require every household member to carry out this task, 184 put their women in charge of it, 79 get it from outsiders, 75 use female teenagers to execute it, 38 make their children do it and 21 put their men in charge of it. Tables A40, A41 and A42 display estimated demand functions for water, gas and electricity, respectively. The estimations for gasoline and diesel yield no statistically significant dummy. 72. In Table A40, we see that the dummies for children, women and male teens are statistically significant. Relative to the average water consumption levels of their respective counterparts, the water consumption level is: (i) 40.64% lower for those households who use children to collect water; (ii) 22.88% higher for those households 55 who put their women in charge of the service; and (iii) 34.18% lower for those households who make it the responsibility of their male teenagers. Table A40: Estimated Water Demand Functions Water Collector Effects Dependent Variables Log Log Log Log Log Log Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 2.100* 2.091* 2.070* 2.090* 2.079* 2.064* (0.186) (0.187) (0.186) (0.186) (0.187) (0.184) Log 0.466* 0.465* 0.463* 0.464* 0.466* 0.483* Expenditure (0.026) (0.026) (0.026) (0.026) (0.026) (0.026) Log Water - 1.416* - 1.408* - 1.403* - 1.403* - 1.408* - 1.378* Price (0.036) (0.036) (0.035) (0.036) (0.036) (0.036) - 0.341* - - - - - Children (0.139) - - 0.032 - - - - Men (0.228) - - 0.206* - - - Women (0.080) - - - 0.063 - - Outsiders (0.104) - - - - 0.059 - Female Teens (0.115) - - - - - - 0.294* Male Teens (0.062) Observations 712 712 712 712 712 712 F 678.82 671.13 679.62 671.59 671.44 699.56 Adjusted R2 0.74 0.74 0.74 0.74 0.74 0.75 73. In Table A41, we observe that the dummies for outsiders and female teens are the only ones that are statistically significant. Relative to the average gas consumption levels of their respective counterparts, the gas consumption level is: (i) 17.82% lower for those households who use the service of outsiders; and (ii) 18.29% higher for those households who assign it to their female teenagers. 74. Children and male teens are the sole statistically significant dummies in the estimated electricity demand functions ­ see Table A42. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 99.17% lower for those households who use their children to collect water; and (ii) 48.14% higher for those households who assign it to their male teenagers. 56 75. We conclude that relative to the average consumption of their counterparts: (i) households whose main water collectors are children consume 40.64% less water and 99.17% less electricity; (ii) households whose main water collectors are women consume 22.88% more water; (iii) households whose main water collectors are non-household members consume 17.82% less gas; (iv) households whose main water collectors are female teenagers consume 18.29% more gas; and (v) households whose main water collectors are male teenagers consume 34.18% less water and more 48.14% more electricity. Table A41: Estimated LPG Demand Functions Water Collector Effects Dependent Variables Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 1.968* - 1.955* - 1.965* - 1.980* - 2.041* - 1.963* (0.220) (0.220) (0.220) (0.219) (0.221) (0.220) Log 0.281* 0.281* 0.281* 0.282* 0.283* 0.281* Expenditure (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG - 0.228* - 0.230* - 0.229* - 0.226* - 0.222* - 0.229* Price (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) 0.086 - - - - - Children (0.090) - - 0.052 - - - - Men (0.112) - - 0.021 - - - Women (0.041) - - - - 0.164* - - Outsiders (0.059) - - - - 0.168* - Female Teens (0.065) - - - - - - 0.006 Male Teens (0.032) Observations 983 983 983 983 983 983 F 168.79 168.44 168.46 172.19 171.71 168.35 Adjusted R2 0.34 0.34 0.34 0.34 0.34 0.34 G7. Effects from Water Collection Distance 76. The cost incurred by a household to fetch water should increase with the distance it must cover to do so. Hence, it is seems reasonable to hypothesize that household water and fuel consumption should be negatively related to the distance between home and 57 water source. The majority of households in our sample have to fetch water at some place outside of their homes. Of 1,108 households, 493get water from places located within 200 meters of their homes, 399 do not collect it, 173 get it from places located between 200 to 400 meters from home, 40 have to cover more than 500 but less than 1000 meters to get it and 3 have to travel beyond 1000 meters to obtain water. 77. Table A43 offers five estimated water demand functions. The dummy for less than 200 meters is the sole one that is not statistically significant. Relative to the average water consumption levels of their respective counterparts, the water consumption level is: (i) 27.51% higher for those households who get water somewhere between 200 to 400 meters from home; (ii) 63.72% lower for those households who get water somewhere between 500 to 1000 meters from home; (iii) 323.34% lower for those households who go beyond 1000 meters from home to get water; and (iv) 34.72% lower for those households who do not collect water. Hence, water consumption appears to be negatively related to the cost of collecting water, as measured by distance between home and water source. 58 Table A42: Estimated Electricity Demand Functions Water Collector Effects Dependent Variables Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 0.605 - 0.641 - 0.648 - 0.682 - 0.556 - 0.646 (0.544) (0.549) (0.549) (0.549) (0.552) (0.531) Log 0.462* 0.462* 0.463* 0.462* 0.460* 0.454* Expenditure (0.034) (0.034) (0.035) (0.034) (0.034) (0.033) Log Water 0.155* 0.166* 0.165* 0.154* 0.167* 0.116* Price (0.047) (0.047) (0.047) (0.048) (0.047) (0.047) Log LPG 0.225* 0.228* 0.229* 0.235* 0.220* 0.210* Price (0.073) (0.073) (0.073) (0.074) (0.074) (0.071) - 0.689* - - - - - Children (0.259) - - 0.018 - - - - Men (0.318) - - - 0.029 - - - Women (0.103) - - - - 0.162 - - Outsiders (0.144) - - - - - 0.202 - Female Teens (0.167) - - - - - 0.393* Male Teens (0.078) Observations 386 386 386 386 386 386 F 56.81 54.05 54.08 54.55 54.62 63.92 Adjusted R2 0.37 0.36 0.36 0.36 0.36 0.40 78. As for the estimated gas demand functions, Table A44 demonstrates that only two dummies are statistically significant, the dummy for those who get water less than 200 meters from home and the dummy for those who do not collect water. The group who gets water less than 200 meters from home consumes 8.87% more gas than the average gas consumption for the other groups. As for the group who does not collect water, its gas consumption is 620.66% lower than the average gas consumption for the other groups. 79. As Table A45 reveals, all dummies are statistically significant in the estimated electricity demand functions. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 31.78% lower for those who get water within 200 meters from their homes; (ii) 29.95% higher for those who get it between 200 and 400 meters from home; (iii) 52.81% lower for those who obtain it between 500 to 1000 meters from home; (iv) 272.85% lower for those who go beyond 1000 meters from home to fetch it; and (iv) 59.52% higher for those who do not collect water. A clear distance pattern is apparent in this case for those who travel more 59 than 200 meters to fetch water ­ the farther away from home a household must go to do so, the smaller it is its electricity consumption relative to the average! Table A43: Estimated Water Demand Functions Water Collection Distance Effects Dependent Variables Log Log Log Log Log Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) Constant 2.071* 1.832* 2.120* 2.084* 2.005* (0.186) (0.194) (0.185) (0.186) (0.184) Log Expenditure 0.465* 0.484* 0.463* 0.466* 0.489* (0.026) (0.026) (0.026) (0.026) (0.026) Log Water Price - 1.405* - 1.392* - 1.415* - 1.410* - 1.377* (0.035) (0.035) (0.035) (0.035) (0.036) 0.113 - - - - Less than 200m (0.079) - 0.243* - - - 200m to 400m (0.059) - - - 0.493* - - 500m to 1000m (0.161) - - - - 1.443* - More than 1000m (0.778) - - - - - 0.298* Do not Collect Water (0.067) Observations 713 713 713 713 713 F 675.13 684.49 676.90 694.29 698.22 Adjusted R2 0.74 0.74 0.74 0.75 0.75 80. In conclusion, our findings about water collection distance reveal that relative to the average consumption levels of their counterparts: (i) households who travel less than 200 meters to collect water consume 8.87% more gas and 31.78% less electricity; (ii) households who travel from 200 to 400 meters to collect water consume 27.51% more water and 29.95% more electricity; (iii) households who travel from 500 to 1000 meters to collect water consume 63.72% less water and 52.81% more electricity; (iv) households who travel more than 1000 meters to collect water consume 323.24% less water and 272.85% less electricity; (v) households who do not collect water consume 34.72% less water, 620.66% less gas and 59.52% more electricity. 60 Table A44: Estimated LPG Demand Functions Water Collection Distance Effects Dependent Variables Log Log Log Log Log LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) Constant - 1.988* - 1.916* - 1.966* - 1.958* - 1.975* (0.220) (0.224) (0.220) (0.219) (0.221) Log Expenditure 0.283* 0.280* 0.281* 0.281* 0.282* (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG Price - 0.229* - 0.232* 0.229* 0.230* 0.228* (0.031) (0.031) (0.031) (0.031) (0.031) 0.085* - - - - Less than 200m (0.043) - 0.031 - - - 200m to 400m (0.031) - - 0.059 - - 500m to 1000m (0.095) - - - - 0.308 - More than 1000m (0.343) - - - - - 1.975* Do not Collect Water (0.221) Observations 983 983 983 983 983 F 170.25 168.81 168.53 168.74 168.50 Adjusted R2 0.34 0.34 0.34 0.34 0.34 G8. Effects from Energy Source Utilization Habits and Access 81. For policy making purposes, it is crucial that public authorities know information about the population's energy source utilization habits as well as whether or not access to the source utilized is difficult. Unlike water, a household may be able to find a close substitute to its most preferred energy source if it is not accessible. Gas, electricity and coal may be deemed close substitutes when used for cooking. Similarly, electricity and kerosene are viewed by some households as close substitutes when these energy sources are used for lighting. However, our previous observations regarding how access costs affect prices and household real incomes and their command over consumption bundles are also applicable here. All else held constant, the higher the access cost faced by a household to consume an energy commodity, the lower should be its quantity demanded of such a commodity (negative own price effect), the higher should be its consumption of close substitutes (positive cross price effect) and the lower should be its consumption of all other commodities (income effect). 61 Table A45: Estimated Electricity Demand Functions Water Collection Distance Effects Dependent Variables Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) Constant - 0.750 - 0.279 - 0.534 - 0.573 - 0.485 (0.547) (0.553) (0.548) (0.548) (0.525) Log Expenditure 0.466* 0.454* 0.458* 0.464* 0.451* (0.034) (0.034) (0.034) (0.034) (0.033) Log Water Price 0.152* 0.151* 0.157* 0.163* 0.103* (0.047) (0.047) (0.047) (0.047) (0.046) Log LPG Price 0.243* 0.198* 0.219* 0.217* 0.187* (0.073) (0.073) (0.073) (0.073) (0.071) - 0.276* - - - - Less than 200m (0.119) - - 0.262* - - - 200m to 400m (0.081) - - - 0.424* - - 500m to 1000m (0.211) - - - - 1.316* - More than 1000m (0.775) - - - - 0.467* Do not Collect Water (0.078) Observations 386 386 386 386 386 F 56.16 58.15 55.63 55.18 67.98 Adjusted R2 0.36 0.37 0.36 0.36 0.41 82. When asked about their energy source utilization habits, the numbers of respondents who claimed they frequently use: (i) coal, are 182 out of 1,122 households (16.22%); (ii) kerosene, are 265 out of 1,120 households (23.66%); (iii) gasoline, are 203 out of 1,129 households (17.98%); (iv) diesel, are 62 out 1,120 households (5.54%); (v) gas, are 1,083 out of 1,157 households (93.60%); (vi) electricity supplied by public company, are 778 out of 1,159 households (67.13%); (vii) electricity from generator, are 164 out of 1,112 households (14.75%). When asked about whether access to energy sources is easy or difficult, the numbers of respondents who claimed it was difficult to access (i) coal, are 20 out of 170 households (11.76%); (ii) kerosene, are 111 out of 253 households (43.87%); (iii) gasoline, are 77 out of 186 households (41.40%); (iv) diesel, are 15 out 58 households (25.86%); (v) gas, are 463 out of 1,031 households (44.91%); 62 (vi) electricity supplied by public company, are 254 out of 737 households (34.46%); (vii) electricity from generator, are 48 out of 148 households (32.43%). Table A46: Water Demand Functions Energy Source Utilization Effects Dependent Variables Log Log Log Log Log Log Log Water Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant 2.695* 2.280* 2.101* 2.454* 2.084* 1.923* 2.246* (0.208) (0.190) (0.205) (0.247) (0.185) (0.194) (0.211) Log 0.421* 0.455* 0.445* 0.436* 0.463* 0.478* 0.439* Expenditure (0.026) (0.027) (0.027) (0.027) (0.026) (0.026) (0.027) Log Water - 1.378* - 1.392* - 1.396* - 1.390* - 1.389* - 1.384* - 1.390* Price (0.035) (0.035) (0.036) (0.036) (0.035) (0.035) (0.036) 0.377* - - - - - - Coal (0.080) - 0.203* - - - - - Kerosene (0.065) - - - 0.128* - - - - Gasoline (0.074) - - - 0.192 - - - Diesel (0.145) - - - - - 0.073 - - LPG (0.114) - - - - - - 0.164* - Electricity (0.061) - - - - - - 0.003 Generator (0.083) Observations 716 715 720 715 741 741 709 F 663.19 650.85 651.43 641.15 680.27 688.71 639.68 Adjusted R2 0.74 0.73 0.73 0.73 0.73 0.74 0.73 83. To study the effects promoted by energy source utilization habits on quantities demanded, seven dummies are included in the estimations of demand functions for water, gas and electricity. The estimations for gasoline and diesel demand functions do not provide us with any statistically significant energy-source-utilization dummy. As it can be observed in Table A44, the diesel, gas and generator dummies are not statistically significant. Relative to the average water consumption levels of their respective counterparts, the water consumption level is: (i) 45.79% higher for those households who frequently use coal; (ii) 22.51% higher for those households who frequently use kerosene; (iii) 13.66% lower for those households who frequently use gasoline; and (iv) 17.82% lower for those who frequently use publicly provided electricity. 63 84. Turning our attention to estimated gas demand functions, we notice from Table A47 that only coal and gasoline dummies are statistically significant. Relative to the average gas consumption levels of their respective counterparts, the gas consumption level is: (i) 7.90% lower for those households who frequently use coal; and (ii) 14.34% lower for those households who frequently use gasoline. Coal and gasoline consumption may be viewed as substitutes to gas consumption. Table A47: Estimated LPG Demand Functions Energy Source Utilization Effects Dependent Variables Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 2.036* - 1.935* - 2.033* - 2.068* - 2.006* - 1.834* (0.226) (0.219) (0.221) (0.233) (0.224) (0.226) Log 0.274* 0.273* 0.283* 0.274* 0.282* 0.275* Expenditure (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG - 0.221* - 0.224* - 0.236* - 0.222* - 0.235* - 0.236* Price (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) - 0.076* - - - - - Coal (0.044) - 0.013 - - - - Kerosene (0.039) - - - 0.134* - - - Gasoline (0.039) - - - - 0.108 - - Diesel (0.068) - - - - - 0.046 - Electricity (0.035) - - - - - 0.044 Generator (0.043) Observations 994 992 1001 992 1029 985 F 153.86 154.00 164.90 155.20 172.44 157.02 Adjusted R2 0.32 0.32 0.33 0.32 0.33 0.32 85. Kerosene and generator are the only statistically significant dummies in the estimated electricity demand functions, as it can be immediately observed in Table A48. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 82.21% lower for those households who frequently use kerosene; and (ii) 23.24% higher for those households who frequently use generators. While kerosene may be viewed as substitute, private electricity generation may be seen as complement to publicly provided electricity. 86. The effects promoted by the difficulty of accessing the frequently used energy sources on the estimated water, gas, gasoline and electricity demand functions are 64 demonstrated below in Tables A49 ­ A51. Consider first the estimated water demand functions. Coal, kerosene and diesel dummies are not statistically significant. Relative to the water consumption levels of their respective counterparts (groups who have easy access to energy sources), the water consumption level of households who deem access to the source difficult is: (i) 52.35% lower for those households who frequently use gasoline; (ii) 18.65% lower for those households who frequently use gas; (iii) 22.75% lower for those households who frequently use publicly provided electricity; and (iv) 86.82% lower for those who frequently use privately provided electricity (generators). One potential interpretation of these results is that households who systematically use coal for cooking and kerosene for lighting tend also to consume more water than the average. Table A48: Estimated Electricity Demand Functions Energy Source Utilization Effects Dependent Variables Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant - 0.392 - 1.161* - 0.643 - 0.735 - 0.657 - 0.385 (0.601) (0.581) (0.569) (0.611) (0.548) (0.607) Log 0.452* 0.454* 0.458* 0.462* 0.460* 0.448* Expenditure (0.038) (0.037) (0.037) (0.037) (0.034) (0.038) Log Water 0.182* 0.169* 0.168* 0.169* 0.163* 0.166* Price (0.050) (0.049) (0.050) (0.050) (0.047) (0.051) Log LPG 0.223* 0.226* 0.242* 0.238* 0.233* 0.230* Price (0.076) (0.075) (0.076) (0.076) (0.0730 (0.077) 0.180 - - - - - Coal (0.113) - - 0.600* - - - - Kerosene (0.213) - - 0.093 - - - Gasoline (0.101) - - - - 0.008 - - Diesel (0.174) - - - - 0.321 - Electricity (0.277) - - - - - 0.209* Generator (0.124) Observations 364 364 368 364 386 358 F 48.37 50.43 48.63 47.40 54.58 47.72 Adjusted R2 0.34 0.35 0.34 0.34 0.36 0.34 87. As for the impacts on gas demand, Table A50 informs us that gasoline, diesel, gas and generator are not statistically significant dummies. Relative to the gas consumption 65 levels of their respective counterparts (groups who have easy access to energy sources), the gas consumption level of households who deem access to the source difficult is: (i) 55.74% higher for those households who frequently use coal; (ii) 37.44% lower for those households who frequently use kerosene; and (iii) 10.52% higher for those households who frequently use publicly provided electricity. Hence, gas consumption is apparently higher for those groups who consider either coal or electricity as a close substitute to gas for cooking purposes and lower for those groups who utilize gas and kerosene together, the first for cooking and the latter for lighting. Table A59: Estimated Water Demand Functions Energy Source Access Effects Dependent Variables Log Log Log Log Log Log Log Water Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant 2.254* 2.207* 2.663* 0.777 2.231* 1.926* 2.712* (0.303) (0.348) (0.523) (0.998) (0.199) (0.251) (0.565) Log 0.482* 0.458* 0.397* 0.663* 0.456* 0.489* 0.411* Expenditure (0.041) (0.054) (0.069) (0.133) (0.027) (0.033) (0.074) Log Water - 1.475* - 1.456* - 1.350* - 1.709* - 1.425* - 1.403* - 1.399* Price (0.064) (0.071) (0.091) (0.211) (0.037) (0.044) (0.107) - 0.180 - - - - - - Coal (0.169) - 0.144 - - - - - Kerosene (0.125) - - - 0.421* - - - - Gasoline (0.152) - - - - 0.038 - - - Diesel (0.277) - - - - - 0.171* - - LPG (0.061) - - - - - - 0.205* - Electricity (0.077) - - - - - - - 0.625* Generator (0.180) Observations 109 201 135 28 659 448 99 F 286.47 231.73 81.42 41.20 629.40 438.43 67.46 Adjusted R2 0.89 0.78 0.64 0.82 0.74 0.75 0.67 88. Table A51 presents seven estimated gasoline demand functions. Kerosene and electricity are the sole statistically significant dummies. Relative to the gasoline consumption levels of their respective counterparts (groups who have easy access to energy sources), the gasoline consumption level of households who deem access to the source difficult is: (i) 157.28% lower for those households who frequently use kerosene; 66 (ii) 40.07% higher for those households who frequently use publicly provided electricity. Accordingly, groups that consume kerosene for lighting may view gasoline as a close substitute, since gasoline is utilized to fuel generators. Similarly, groups that utilize generators also use publicly provided electricity and do not consider generator usage as a substitute activity to usage of publicly provided electricity. Table A50: Estimated LPG Demand Functions Energy Sources Access Effects Dependent Variables Log Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (6) Constant - 2.055* - 2.241* - 2.928* - 2.696* - 1.957* - 2.205* - 1.371* (0.704) (0.459) (0.532) (1.231) (0.220) (0.272) (0.686) Log 0.298* 0.429* 0.392* 0.149* 0.281* 0.282* 0.308* Expenditure (0.040) (0.029) (0.034) (0.064) (0.013) (0.015) (0.040) Log LPG - 0.247* - 0.315* - 0.228* 0.005 - 0.230* - 0.202* - 0.342* Price (0.094) (0.068) (0.072) (0.155) (0.031) (0.037) (0.092) 0.443* - - - - - - Coal (0.251) - - 0.318* - - - - - Kerosene (0.062) - - 0.069 - - - - Gasoline (0.076) - - - - 0.076 - - - Diesel (0.193) - - - - 0.005 - - LPG (0.031) - - - - - 0.100* - Electricity (0.040) - - - - - - 0.122 Generator (0.111) Observations 138 191 180 53 983 700 143 F 20.20 73.93 46.63 1.83 168.34 115.86 22.76 Adjusted R2 0.30 0.54 0.43 0.05 0.34 0.33 0.32 89. The estimated electricity demand effects are shown in Table A52. Kerosene, gasoline, diesel and generator are not statistically significant dummies. Relative to the electricity consumption levels of groups who have easy access to energy sources, the electricity consumption level of households who deem access to the source difficult is: (i) 314.54% lower for those households who frequently use coal; (ii) 19.96% lower for those households who frequently use gas; (iii) 20.08% lower for those households who frequently use publicly provided electricity. These results suggest that coal and gas users may see electricity as a close substitute energy source for cooking, and those households 67 who typically rely on publicly provided electricity for various purposes are also high users of an alternative energy source, such as generators. Table A51: Estimated Gasoline Demand Functions Energy Source Access Effects Dependent Variables Log Log Log Log Log Log Log Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant - 2.103 - 7.764* 1.945* - 10.460* 1.959* 1.937* 0.829 (1.918) (2.552) (0.624) (4.848) (0.627) (1.116) (0.981) Log 0.600* 0.594* 0.472* 0.577* 0.463* 0.500* 0.467* Expenditure (0.112) (0.132) (0.058) (0.250) (0.057) (0.070) (0.072) Log Gasoline 0.901 4.031* - 0.638* 5.227* - 0.602* - 0.771 0.025 Price (0.879) (1.086) (0.233) (2.407) (0.235) (0.515) (0.394) 0.522 - - - - - - Coal (0.724) - - 0.945* - - - - - Kerosene (0.408) - - 0.176 - - - - Gasoline (0.132) - - - 0.646 - - - Diesel (0.716) - - - - 0.146 - - LPG (0.132) - - - - - 0.337* - Electricity (0.176) - - - - - - - 0.189 Generator (0.176) Observations 30 9 171 13 171 120 95 F 10.35 8.69 24.07 3.85 23.81 19.08 14.73 Adjusted R2 0.49 0.74 0.29 0.42 0.29 0.31 0.30 G9. Effects from Transportation Modes Frequently Used to Go to School and Health Center 90. Information about the transportation modes the population frequently uses to go to schools and health centers is of vital importance for policy makers who seek alternative venues for utilizing public funds saved with the removal of fuel and utility price subsidies. In this subsection, we provide this information and also examine the effects that are promoted on the demands for water, gas, gasoline, diesel and electricity of introducing dummies that identify the various transportation mode groups. 68 Table A52: Estimated Electricity Demand Functions Energy Source Access Effects Dependent Variables Log Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant - 1.378 - 3.304 - 0.295 - 2.162 - 0.598 - 0.475 0.774 (0.993) (2.052) (1.2670 (1.917) (0.545) (0.574) (1.423) Log 0.418* 0.411* 0.472* 0.272* 0.449* 0.451* 0.423* Expenditure (0.057) (0.107) (0.090) (0.141) (0.035) (0.085) (0.089) Log Water 0.322* 0.135 0.210 - 0.098 0.171* 0.149* 0.107 Price (0.096) (0.116) (0.132) (0.209) (0.047) (0.048) (0.156) Log LPG 0.406* 0.585* 0.192 0.616* 0.246* 0.227* 0.108 Price (0.114) (0.319) (0.171) (0.229) (0.073) (0.074) (0.165) - 1.422* - - - - - - Coal (0.557) - - 0.592 - - - - - Kerosene (0.365) - - - 0.299 - - - - Gasoline (0.208) - - - 0.190 - - - Diesel (0.269) - - - - - 0.182* - - LPG (0.080) - - - - - - 0.183* - Electricity (0.085) - - - - - - - 0.369 Generator (0.239) Observations 63 14 80 22 386 373 50 F 19.96 9.75 11.16 3.13 56.05 54.13 6.57 Adjusted R2 0.55 0.73 0.34 0.29 0.36 0.36 0.31 91. 1,042 households answered questions about transportation modes frequently used to go to school. The choices made by these households in descending order of importance were: (i) walking, for 603 households (57.87%); (ii) vans, for 288 households (27.64%); (iii) automobiles, for 96 households (9.21%); (iv) buses, for 32 households (3.07%); (v) other modes, for 19 households (1.82%); (vi) bicycles, for 3 households (0.29%); (vii) motorcycle, 1 for household (0.10%). 69 92. Furthermore, of 1,103 households, the questionnaire choices made by households regarding the frequently used transportation mode to go health centers in descending order of importance were: (i) walking, for 586 households (53.13%); (ii) vans, for 339 households (30.73%); (iii) automobiles, for 144 households (13.06%); (iv) buses, for 25 households (2.27%); (v) motorcycle, for 5 households (0.45%); (vi) bicycles, for 3 households (0.27%); (vii) other modes, for 1 household (0.09%). 93. In sum, the most frequently used transportation mode to go to schools and health centers is walking, followed by van service and automobiles. Public buses are rarely used. Table A53: Estimated Water Demand Functions Transportation Mode to School Effects Dependent Variables Log Log Log Log Log Log Water Water Water Water Water Water Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 2.007* 1.975* 1.982* 1.951* 1.743* 1.976* (0.188) (0.188) (0.188) (0.190) (0.197) (0.189) Log 0.477* 0.480* 0.484* 0.484* 0.494* 0.479* Expenditure (0.026) (0.026) (0.026) (0.027) (0.026) (0.026) Log Water - 1.406* - 1.402* - 1.409* - 1.403* - 1.417* - 1.403* Price (0.036) (0.036) (0.036) (0.036) (0.036) (0.036) - 0.420* - - - - - Bus (0.153) - - 0.117 - - - - Bicycle (0.545) - - - 0.138* - - - Van (0.066) - - - - 0.099 - - Automobile (0.107) - - - - 0.217* - Walk (0.060) - - - - - - 0.017 Other Modes (0.201) Observations 686 686 686 686 686 686 F 673.91 664.15 669.76 665.20 681.19 664.10 Adjusted R2 0.75 0.74 0.75 0.74 0. 75 0.74 70 94. Let us now consider the estimated demand functions. Table A53 presents six estimated water demand functions and the effects associated with the dummies that identify groups of consumers according to the transportation mode used to go to school. The dummies for bicycle, automobile and other modes are not statistically significant. Relative to the average level of water consumption of their respective counterparts, the water consumption level is: (i) 52.20% lower for those who use buses to go to school; (ii) 14.80% lower for those who use vans; and (iii) 24.23% higher for those who walk to school. Thus, households who use buses and vans to go to school are relatively low water demanders, while households who walk to school are relatively high water demanders. Table A54: Estimated Water Demand Functions Transportation Mode to Health Center Effects Dependent Variables Log Log Log Log Log Log Water Water Water Water Water Water Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 2.048* 2.045* 2.062* 2.064* 2.040* 1.779* (0.185) (0.186) (0.186) (0.186) (0.187) (0.200) Log 0.472* 0.472* 0.474* 0.468* 0.473* 0.492* Expenditure (0.026) (0.026) (0.026) (0.026) (0.026) (0.026) Log Water - 1.410* - 1.405* - 1.411* - 1.404* - 1.406* - 1.424* Price (0.035) (0.035) (0.036) (0.036) (0.035) (0.036) - 0.552* - - - - - Bus (0.203) - - 1.315* - - - - Bicycle (0.549) - - - 0.121* - - - Van (0.063) - - - - 0.065 - - Motorcycle (0.451) - - - - - 0.106 - Automobile (0.094) - - - - - 0.224* Walk (0.060) Observations 712 712 712 712 712 712 F 682.88 680.70 678.10 673.35 674.94 691.10 Adjusted R2 0.74 0.74 0.74 0.74 0.74 0.74 95. Table A54 collects estimated water demand functions and the effects of the dummies that identify groups of consumers according to the transportation mode used to go to health centers. Only motorcycle and automobile dummies are not statistically significant. Relative to the average level of water consumption of their respective counterparts, the water consumption level is: (i) 73.67% lower for those who use buses; (ii) 272.48% lower for those who use bicycles; (iii) 12.86% lower for those who use 71 vans; and (iv) 25.11% higher for those who walk. The behavior patterns of those households who use buses and vans to go to health centers is similar to those who use the same modes of transportation to go to school in that their water demands are lower than the average demands of their respective counterparts. Those households who use bicycles to go to health centers are also included among the groups of relatively low water demanders. Households who walk to health center behave similarly to households who walk to school in the sense that their water demands exceed the average demands of their respective counterparts. Table A55: Estimated LPG Demand Functions Transportation Mode to School Effects Dependent Variables Log Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG LPG Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant - 2.167* - 2.165* - 2.168* - 2.165* - 2.194* - 2.195* - 2.163* (0.225) (0.225) (0.225) (0.225) (0.225) (0.226) (0.225) Log 0.283* 0.282* 0.282* 0.283* 0.289* 0.284* 0.282* Expenditure (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG - 0.201* - 0.201* - 0.202* - 0.201* - 0.202* - 0.202* - 0.201* Price (0.032) (0.032) (0.032) (0.032) (0.032) (0.032) (0.032) 0.013 - - - - - - Bus (0.088) - 0.048 - - - - Bicycle (0.278) - - 0.019 - - - Van (0.035) - - - - 0.005 - - Motorcycle (0.481) - - - - - 0.116* - - Automobile (0.053) - - - - - 0.034 Walk (0.032) - - - - - - 0.169 Other Modes (0.129) Observations 932 932 932 932 932 932 932 F 160.05 160.06 160.19 160.04 162.42 160.62 160.91 Adjusted R2 0.34 0.34 0.34 0.34 0.34 0.34 0.34 96. Estimated gas demand functions, which capture the effects of school transportation mode dummies, are demonstrated in Table A55. Only the automobile dummy is statistically significant. Relative to the average gas consumption level of its counterparts, the gas consumption level is 12.30% lower for households who use automobiles to go to school. 72 97. As for the discussion of the effects of dummies that identify groups of gas consumers according to the transportation mode used to go to health centers, we first notice in Table A56 that automobile and walk dummies are the only ones that are statistically significant. Relative to the average gas consumption levels of their counterparts, the gas consumption level is: (i) 17.12% lower for households who use automobiles; and (ii) 8.11% higher for those who walk. Hence, the behavior of households who use automobiles to go to health centers is similar to the behavior of households who use automobiles to go to school in that their gas consumptions fall below the average gas consumptions of their respective counterparts. Table A56: Estimated LPG Demand Functions Transportation Mode to Health Center Effects Dependent Variables Log Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG LPG Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) Constant - 1.955* - 1.954* - 1.953* - 1.958* - 1.977* - 2.064* - 1.955* (0.220) (0.220) (0.220) (0.220) (0.219) (0.224) (0.220) Log 0.281* 0.280* 0.280* 0.281* 0.290* 0.286* 0.280* Expenditure (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) (0.013) Log LPG - 0.230* - 0.230* - 0.230* - 0.229* - 0.232* - 0.225* - 0.230* Price (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) (0.031) - 0.094 - - - - - - Bus (0.102) - 0.032 - - - - - Bicycle (0.281) - - 0.006 - - - - Van (0.033) - - - 0.110 - - - Motorcycle (0.244) - - - - - 0.158* - - Automobile (0.045) - - - - - 0.078* - Walk (0.032) - - - - - - - 0.021 Other (0.486) Observations 980 980 980 980 980 980 980 F 166.93 166.52 166.53 166.61 172.70 169.58 166.51 Adjusted R2 0.34 0.34 0.34 0.34 0.34 0.34 0.34 98. Turning now our attention to estimated gasoline demand functions, it is clear from Table A57 that the bus dummy is the sole statistically significant dummy. Relative to the average gasoline consumption level of its counterparts, the gasoline consumption level is 138.45% lower for those who use buses to go to school. 73 Table A57: Estimated Gasoline Demand Functions Transportation Mode to School Effects Dependent Variables Log Log Log Log Log Log Gasol. Gasol. Gasol. Gasol. Gasol. Gasol. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) Constant 2.149* 2.103* 2.135* 2.077* 1.911* 2.058* (0.623) (0.633) (0.634) (0.634) (0.657) (0.633) Log 0.461* 0.463* 0.456* 0.473* 0.474* 0.471* Expenditure (0.059) (0.060) (0.060) (0.061) (0.062) (0.061) Log Gasoline - 0.632* - 0.640* - 0.649* - 0.636* - 0.624* - 0.639* Price (0.231) (0.235) (0.235) (0.235) (0.235) (0.231) - 0.869* - - - - - Bus (0.323) - 1.173 - - - - Bicycle (0.853) - - 0.173 - - - Van (0.152) - - - - 0.235 - - Automobile (0.165) - - - - 0.178 - Walk (0.138) - - - - - - 1.313 Other Modes (0.858) Observations 160 160 160 160 160 160 F 24.30 21.80 21.52 21.87 21.70 22.01 Adjusted R2 0.31 0.28 0.28 0.28 0.28 0.28 99. As illustrated in Table A58, the bus and walk dummies, which identify the groups of gasoline consumers who take buses and walk to health centers, respectively, are the only ones which are statistically significant in the estimated gasoline demand functions. Relative to the average gasoline consumption levels of their respective counterparts, the gasoline consumption level is: (i) 77.54% lower for those who take buses; and (ii) 43.48% higher for those who walk. Table A58 also informs us about four estimated diesel demand functions. The walk dummy is the sole one which is statistically significant. Relative to the average diesel consumption level of its counterparts, the diesel consumption level is 188.06% lower for those households who walk. Hence, the behavior of households who walk to school is similar to the behavior of households who walk to health centers in that their electricity consumptions are lower than the average electricity consumptions of their respective counterparts. 74 Table A58: Estimated Gasoline and Diesel Demand Functions Transportation Mode to Health Center Effects Dependent Variables Log Gasoline Quantity Log Diesel Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) Constant 2.098* 2.179* 2.015* 2.087* 1.976* 0.664 0.575 0.409 0.489 (0.618) (0.627) (0.629) (0.621) (0.611) (1.989) (1.963) (2.001) (1.864) Log 0.463* 0.466* 0.472* 0.469* 0.481* 0.687* 0.695* 0.673* 0.656* Expenditure (0.058) (0.058) (0.060) (0.059) (0.058) (0.162) (0.159) (0.160) (0.151) Log Gasoline - 0.630* - 0.674* - 0.637* - 0.637* - 0.721* - - - - Price (0.232) (0.236) (0.234) (0.234) (0.232) Log Diesel - - - - - - 0.978* - 0.998* - 0.924* - 0.817* Price (0.366) (0.362) (0.369) (0.349) - 0.574* - - - - - 0.261 - - - Bus (0.325) (1.105) - - 0.170 - - - - 0.431 - - Van (0.151) (0.383) - - 0.665 - - - - - - Motorcycle (0.870) - - - - 0.151 - - - 0.271 - Automobile (0.140) (0.328) - - - - 0.361* - - - - 1.058* Walk (0.133) (0.426) Observations 170 170 170 170 170 48 48 48 48 F 24.13 23.26 22.94 23.21 26.11 10.35 11.04 10.71 13.83 Adjusted R2 0.29 0.28 0.28 0.28 0.31 0.37 0.39 0.38 0.45 100. Table A59 gives us the estimated electricity functions which account for the school and health center transportation mode dummies. Models (1) through (5) represent the estimations that include the school transportation mode dummies. Models (6) through (10) represent the estimations that include the health center transportation mode dummies. The automobile dummy is the sole statistically significant dummy in the first five models. Relative to the average electricity consumption level of its counterparts, the electricity consumption level is 28.66% higher for those households who utilize automobiles to go to school. As for the last five models, only the bicycle dummy is not statistically significant. Relative to the average electricity consumption levels of their respective counterparts, the electricity consumption level is: (i) 68.03% lower for those who take buses; (ii) 21.29% higher for those who take vans; (iii) 25.86% higher for those who use automobiles; and (iv) 33.64% lower for those who walk. 75 Table A59: Estimated Electricity Demand Functions Transportation Mode Effects Equations with School Equations with Health Center Transportation Mode Dummies Transportation Mode Dummies Log Log Log Log Log Log Log Log Log Log Electr Electr. Electr. Electr. Electr. Electr Electr. Electr. Electr. Electr. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Quant. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (1) (2) (3) (4) (5) (6) (7) (8) (9) (10) Constant - 0.681 - 0.692 - 0.685 - 0.684 - 0.637 - 0.656 - 0.607 - 0.574 - 0.639 - 0.268 (0.561) (0.562) (0.562) (0.558) (0.564) (0.546) (0.549) (0.546) (0.546) (0.550) Log 0.463* 0.462* 0.462* 0.452* 0.458* 0.463* 0.459* 0.455* 0.451* 0.439* Expenditure (0.035) (0.035) (0.035) (0.036) (0.034) (0.034) (0.034) (0.034) (0.034) (0.035) Log Water 0.165* 0.168* 0.167* 0.168* 0.174* 0.155* 0.162* 0.177* 0.166* 0.192* Price (0.049) (0.049) (0.049) (0.049) (0.049) (0.047) (0.047) (0.048) (0.047) (0.047) Log LPG 0.234* 0.236* 0.236* 0.240* 0.237* 0.231* 0.226* 0.216* 0.233* 0.215* Price (0.076) (0.076) (0.076) (0.075) (0.076) (0.073) (0.073) (0.073) (0.073) (0.072) - 0.239 - - - - - 0.519* - - - - Bus (0.218) (0.218) - 0.205 - - - - 0.625 - - - Bicycle (0.547) (0.774) - - - 0.028 - - - - 0.193* - - Van (0.087) (0.081) - - - 0.252* - - - - 0.230* - Automobile (0.114) (0.101) - - - - - 0.076 - - - - - 0.290* Walk (0.081) (0.082) Observations 370 370 370 370 370 385 385 385 385 385 F 52.24 51.83 51.81 53.70 52.12 54.90 52.95 54.90 54.71 54.90 Adjusted R2 0.36 0.36 0.36 0.36 0.36 0.36 0.35 0.36 0.36 0.36 G10. Education, Health and Public Transportation Expenditure Effects 101. We now investigate whether households who claimed they incur education, health care or public transportation expenses behave differently from those who did not make such claims. It is important to note that in the latter group there may be households who actually spend resources in education, health care or public transportation but preferred not to reveal such expenditures. Separating spenders from non-spenders and comparing the consumptions of these different groups yield valuable information for policy making purposes, but it does not allow us to forecast the impacts that increased expenditures will have on the quantities demanded of the various products. Hence, we will also try to capture the effects that education, health care and public transportation expenditures have on the consumption levels of water, electricity and fuels by disentangling household expenditures. We "sliced" total household expenditures to create six types of expenditure figures: (i) education expenditure; (ii) health care expenditure; (iii) public transportation expenditure; (iv) total expenditure net of education expenditure; (v) total expenditure net 76 of health care expenditure; and (vi) total expenditure net of public transportation expenditure. The findings of this subsection are relevant because policy makers may find it desirable to increase governmental expenditures in education, health care and public transportation in order to ameliorate the negative effects brought about by removal of fuel and utility subsidies. By eliminating such subsidies, the government will also eliminate its subsidy expenditures enabling it to employ some of the savings in alternative publicly provided service provision. Table A60: Water Demand Functions Education, Health and Public Education Issues Dependent Variables Log Log Log Log Log Log Water Water Water Water Water Water Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) 2.165* 2.104* 2.068* 2.370* 2.017* 1.781* Constant (0.189) (0.207) (0.186) (0.230) (0.189) (0.239) 0.470* - 0.458* - 0.463* - Log Total Expenditure (0.026) (0.026) (0.026) - 1.411* - 1.319* -1.408* - 1.370* - 1.410* - 1.414* Log Water Price (0.035) (0.047) (0.035) (0.047) (0.035) (0.045) Dummy ­ Education - 0.144* - - - - - Expenditure (0.070) Log Net of Education - 0.288* - - - - Expenditure (0.030) Log Education - 0.235* - - - - Expenditure (0.031) Dummy ­ Health - - 0.100 - - - Expenditure (0.067) Log Net of Health - - - 0.159* - - Expenditure (0.027) Log Health - - - 0.333* - - Expenditure (0.039) Dummy ­ Pub. Trans- - - - - 0.130* - portation Expenditure (0.063) Log Net of Pub. - - - - - 0.195* Transp. Expenditure (0.027) Log Pub. Transp. - - - - - 0.404* Expenditure (0.040) Observations 713 451 713 428 713 445 F 677.95 388.69 675.33 406.79 677.96 436.88 Adjusted R2 0.74 0.72 0.74 0.74 0.74 0.75 102. Table A60 provides us with six estimated water demand functions. Models (1), (3) and (5) include dummies for education spenders, health care spenders and public transportation spenders. Models (2), (4) and (6) do not include dummies. In each of these 77 models, there are two expenditure estimators, net of education expenditure and education expenditure in model (2), net of health expenditure and health expenditure in model (4), and net of public transportation expenditure and public transportation expenditure in model (6). Tables A61 ­ A63 are similarly organized. Table A61: LPG Demand Functions Education, Health and Public Education Issues Dependent Variables Log Log Log Log Log Log LPG LPG LPG LPG LPG LPG Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) - 1.963* - 2.285* - 2.061* - 2.313* - 1.933* - 2.013* Constant (0.218) (0.269) (0.215) (0.310) (0.217) (0.263) - 0.227* - 0.303* - 0.289* - Log Total Expenditure (0.031) (0.013) (0.013) - 0.227* - 0.184* - 0.217* - 0.131* - 0.228* - 0.233* Log LPG Price (0.031) (0.038) (0.030) (0.044) (0.030) (0.037) Dummy ­ Education - 0.105* - - - - - Expenditure (0.035) Log Net of Education - 0.243* - - - - Expenditure (0.014) Log Education - 0.074* - - - - Expenditure (0.014) Dummy ­ Health - - - 0.216* - - - Expenditure (0.033) Log Net of Health - - - 0.213* - - Expenditure (0.015) Log Health - - - 0.047* - - Expenditure (0.021) Dummy ­ Pub. - - - - - 0.153 * - Transp. Expenditure (0.032) Log Net of Pub. - - - - - 0.220* Transp. Expenditure (0.014) Log Pub. Transp. - - - - - 0.101* Expenditure (0.019) Observations 983 558 983 510 983 537 F 172.79 170.82 189.78 97.29 180.13 123.70 Adjusted R2 0.34 0.48 0.37 0.36 0.35 0.41 103. The dummies for education and public transportation spenders are statistically significant, but the dummy for health spenders is not. The estimates for the dummies in equations (1) and (5) imply that education spenders consume 15.49% less water than their counterparts and public transportation spenders consume 13.88% more water than their counterparts. 78 104. All expenditure estimators are statistically significant. In model (2), we find that a 10% increment in net of education expenditure yields an increase of 2.88% in water consumption, while a 10% expansion in education expenditure leads to an increase of 2.35% in water consumption. In model (4), we notice that water consumption expands by 1.59% and 3.33% if there are increases of 10% in net of health expenditure and health expenditure, respectively. In model (6), water consumption rises by 1.95% and 4.04% if there are increases of 10% in net of public transportation expenditure and public transportation expenditure, respectively. 105. The dummies for education, health and public transportation spenders are all statistically significant in the estimated gas demand functions ­ see Table A61. We find that all spenders consume less gas than their counterparts. Relative to the consumption levels of non-spenders, gas consumption is lower by: (i) 11.07% for education spenders; (ii) 24.11% for health spenders; and (iii) 16.53% for public transportation spenders. 79 Table A62: Gasoline Demand Functions Education, Health and Public Education Issues Dependent Variables Log Log Log Log Log Log Gasoline Gasoline Gasoline Gasoline Gasoline Gasoline Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) 2.040* 1.433* 2.102* 3.522* 2.182* 2.554* Constant (0.619) (0.673) (0.618) (0.792) (0.626) (0.678) 0.450* - 0.464* - 0.462* - Log Total Expenditure (0.059) (0.058) (0.058) -0.625* - 0.137 - 0.625* - 0.890* - 0.645* - 0.728* Log Gasoline Price (0.234) (0.273) (0.235) (0.289) (0.234) (0.234) Dummy ­ Education 0.144 - - - - - Expenditure (0.168) Log Net of Education - 0.279* - - - - Expenditure (0.063) Log Education - 0.204* - - - - Expenditure (0.062) Dummy ­ Health - - - 0.067 - - - Expenditure (0.148) Log Net of Health - - - 0.224* - - Expenditure (0.056) Log Health - - - 0.167* - - Expenditure (0.088) Dummy ­ Pub. - - - - - 0.113 - Transp. Expenditure (0.133) Log Net of Pub. - - - - - 0.202* Transp. Expenditure (0.062) Log Pub. Transp. - - - - - 0.301* Expenditure (0.074) Observations 171 97 171 99 171 97 F 23.58 20.58 23.32 14.54 23.57 14.53 Adjusted R2 0.28 0.38 0.28 0.29 0.28 0.30 106. All expenditure estimators are statistically significant. Gas consumption rises by 2.43% and 0.74% if net of education and education expenditures rise by 10%, respectively. Gas consumption increases by 2.13% and 0.047% if net of health and health expenditures increase by 10%, respectively. Gas consumption expands by 2.20% and 1.01% if net of public transportation and public transportation expenditures expand by 10%, respectively. 107. In the estimated gasoline demand functions, all dummies are statistically insignificant ­ see Table A62. All expenditure estimators, however, are statistically significant. Gasoline consumption rises by 2.79% and 2.04% if net of education and education expenditures, respectively, are incremented by 10%. Gasoline consumption 80 expands by 2.24% and 1.67% if net of health and health expenditures, respectively, are increased by 10%. Gasoline consumption expands by 2.02% and 3.01% if net of public transportation and public transportation expenditures expand by 10%, respectively. Table A63: Diesel Demand Functions Education, Health and Public Education Issues Dependent Variables Log Log Log Log Log Log Diesel Diesel Diesel Diesel Diesel Diesel Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) 0.575 0.819 1.253 7.262* 0.916 2.486 Constant 1.869) (2.657) 1.790) (2.706) 1.875) (3.990) 0.717* - 0.727* - 0.659* - Log Total Expenditure (0.168) (0.146) (0.154) - 0.963* - 0.802 - 1.070* - 2.120* - 0.966* - 0.581 Log Diesel Price (0.357) (0.610) (0.345) (0.566) (0.357) (0.781) Dummy ­ Education - 0.235 - - - - - Expenditure (0.467) Log Net of Education - 0.464* - - - - Expenditure (0.171) Log Education - 0.190 - - - - Expenditure (0.128) Dummy ­ Health - - - 0.731* - - - Expenditure (0.376) Log Net of Health - - - 0.180 - - Expenditure (0.175) Log Health - - - 0.210 - - Expenditure (0.190) Dummy ­ Pub. - - - - - 0.149 - Transp. Expenditure (0.321) Log Net of Pub. - - - - - 0.713* Transp. Expenditure (0.237) Log Pub. Transp. - - - - - 0.317 Expenditure (0.269) Observations 49 25 49 30 49 24 F 11.10 4.74 13.13 6.50 11.08 6.10 Adjusted R2 0.39 0.32 0.43 0.36 0.39 0.40 108. Table A63 informs us that only the dummy for health spenders is statistically significant in the estimated diesel demand functions. We find that health spenders consume 107.72% less diesel than non-spenders. Education, health and public transportation expenditure estimators are all insignificant. 109. In the estimated electricity demand functions, we notice that the dummy for health is statistically insignificant ­ see Table A64. We find that education and public 81 transportation spenders consume less electricity than their counterparts, 24.83% less for education spenders and 40.07% less for public transportation spenders. All expenditure estimators are statistically significant. Electricity consumption increases by: (i) 1.95% if education rises by 10%; (ii) 2.22% if health expenditure rises by 10%; and (iii) 2.28% if public transportation expenditure rises by 10%. Table A64: Electricity Demand Functions Education, Health and Public Education Issues Dependent Variables Log Log Log Log Log Log Electric. Electric. Electric. Electric. Electric. Electric. Quantity Quantity Quantity Quantity Quantity Quantity Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Independent Variables (1) (2) (3) (4) (5) (6) - 0.468 - 0.370 0.466* - 0.330 - 0.603 - 2.196* Constant (0.551) (0.727) (0.035) (0.735) (0.536) (0.711) 0.468* - 0.166* - 0.463* - Log Total Expenditure (0.034) (0.047) (0.034) 0.158* 0.218* 0.229* 0.195* 0.150* 0.113* Log Water Price (0.047) (0.068) (0.073) (0.070) (0.046) (0.060) 0.222* 0.231* - 0.228* 0.251* 0.426* Log LPG Price (0.073) (0.105) (0.104) (0.072) (0.095) Dummy ­ Education - 0.219* - - - - - Expenditure (0.101) Log Net of Education - 0.276* - - - - Expenditure (0.039) Log Education - 0.195* - - - - Expenditure (0.040) Dummy ­ Health - - - 0.058 - - - Expenditure (0.091) Log Net of Health - - - 0.252* - - Expenditure (0.039) Log Health - - - 0.222* - - Expenditure (0.051) Dummy ­ Pub. - - - - - 0.337* - Transp. Expenditure (0.080) Log Net of Pub. - - - - - 0.303* Transp. Expenditure (0.037) Log Pub. Transp. - - - - - 0.228* Expenditure (0.055) Observations 386 239 386 229 386 231 F 55.89 37.79 54.21 34.10 60.94 45.82 Adjusted R2 0.36 0.38 0.36 0.37 0.38 0.44 H. Budget Shares as Functions of Expenditure 82 110. Per capita household expenditure on goods and services is our measure of per capita household income. In our analysis of fuel and utility subsidization incidence, we were particularly interested in examining which quintile income group benefits from fuel and utility subsidies and whether there is a noticeable rising or declining tendency in terms of derived benefits as we move from the first quintile towards the last. This descriptive analysis generated useful results, which have been reported in the text. 111. In this subsection, we carry out another type of statistical exercise. We investigate whether fuel and utility budget shares decline with income. For comparison purposes, we also provide econometric results for the relationships between each food, education and public transportation budget shares and income. The hypothesis that budget shares of essential goods and services should decline with income is deeply rooted in microeconomic analysis of consumer behavior, since poor households should spend a proportionately larger portion of their incomes in consumption of such goods and services than better off households. As household income rises, households can afford consumption of non-essential goods and services and thus spend relatively less in consumption of essential goods and services. Table A65: Budget Shares as Functions of Expenditure Dependent Variables: Budget Shares in Logs Food LPG Water Educat. P. Tran. Electr. Keros. Gasol. Coal Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Coeff. Independent (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) (S. E.) Variables (1) (2) (3) (4) (5) (6) (7) (8) (9) 0.434* 2.166* 0.065 - 4.315* - 0.104 0.887* - 1.155* 0.893 1.673* Constant (0.207) (0.204) (0.309) (0.379) (0.384) (0.350) (0.627) (0.759) (0.760) Log - 0.160* - 0.630* - 0.391* 0.189* - 0.308* - 0.473* - 0.253* - 0.422* - 0.639* Expenditure (0.023) (0.023) (0.035) (0.043) (0.043) (0.039) (0.074) (0.083) (0.084) Observations 1021 980 851 772 677 576 249 209 178 F 46.91 746.70 124.80 19.83 51.07 149.66 11.62 25.98 58.43 Adjusted R2 0.04 0.43 0.13 0.02 0.07 0.21 0.04 0.11 0.25 112. Table A65 provides us with the results of nine regressions, ranging from the most representative dependent variable, food budget share, to the least one, coal budget share. The regressions for health and diesel budget shares did not produce statistically significant estimates. With the exception of the regression for education budget share, model (4), we see that the econometric results seem to support the hypothesis that budget shares of essential goods and services decline with income. Note that the lowest elasticities of income, in absolute terms, are observed for food and education budget shares, respectively. In other words, as income increases, the changes in food and education budget shares are lower than those for all the other budget shares considered. When considering these findings, one must not forget that food consumption includes consumption of essential and non-essential food items (e.g., luxurious foods such as prime beef) and education expenditure includes a gamma of expenditure items, ranging from expenditures on books and other school supplies to payments of matriculation and other fees at primary though tertiary private schools. Hence, the estimates of income elasticity for food and education budget shares are not very surprising. Furthermore, the 83 positive value for the estimate of income elasticity for education budget share indicates that richer households spend relatively more with education than poor ones. This is very consistent with the findings discussed in the text when we compared budget shares across quintile groups. 113. As for the other important public service considered here, the estimate of income elasticity for public transportation budget share, being low, implies that medium and high income households also benefit from public transportation services. As we discussed in the text, such benefits are very likely associated with usage of van service rather than bus service, since van service utilization is much larger than bus service utilization. 114. Although the largest income elasticity is for coal budget share, the most striking result for our purposes pertains to the second largest income elasticity, that for gas budget share, since gas consumption is widespread. In models (2) and (9), we see that a 10% rise in household income leads to decreases of 6.30% and 6.39% in gas and coal budget shares, respectively. In the text, we mentioned that the gas budget share is large for the first quintile and that it declines by substantial amounts as we move from the first toward the fifth quintile. Hence, the income elasticity estimate in the second model captures these two facts. 115. We also mentioned in the text that kerosene presented the largest reduction in budget share as we moved from the first to the fifth quintile. However, as we can observe in model (7), the income elasticity is the smallest among those for fuels and utilities ­ a 10% increase in income yields a 2.53% reduction in kerosene budget share. How can we explain this fact in light of our reasoning for the gas budget share's income elasticity? Not only kerosene consumption is more concentrated than gas consumption, but it is also less frequent. Most of the households who consume kerosene are within the first and second quintile groups. Hence, as income rises, the reduction in kerosene budget share is weaker than in the case of gas budget share. 116. Among fuel and utility budget shares, the income elasticity for water budget share is the lowest ­ see model (3) in Table A65. A 10% increase in income leads to a 3.91% reduction in water budget share. This low elasticity may be explained by the facts that there is widespread consumption of water sold by private providers, often at high prices, many poor households seem to favor consumption of water made available at public taps relative to alternative provision modes, that most of the population does not have access to water services provided by EPAL and that the quality of the water distributed by EPAL to the minority which has access to this service is generally poor due to contamination or leakage in the water distribution network. These facts may give rise to two phenomena that rationalize the low income elasticity: (i) poor households, within first and second quintiles, are more likely to consume water from public taps than better off households; and (ii) better off households are more likely to purchase water from private providers. 117. Finally, Table A65 also informs us that a 10% increase in income should lead to decreases of 4.73% and 4.22% in electricity and gasoline budget shares, respectively. These income response rate figures are situated within the middle of the fuel and utility 84 income response rate figures. Gasoline expenditure appears to be relatively more important for households within the fifth quintile than for households within the other four quintiles because this energy source is highly used by well off households in the generation of lighting or to fuel their automobiles. The income response rate figure for gasoline budget share is then highly influenced by the behavior of households who have generators or who own automobiles. APPENDIX III: WELFARE ANALYSIS 1. The two fundamental postulates used in this welfare analysis are that the preferences of all residents in Luanda are of the Gorman form (more precisely, Cobb- Douglas preferences) and that the social welfare function represents the preferences of the average resident in Luanda. The first postulate implies that social preferences can be formed by aggregating individual preferences. Indeed, the functional forms taken by the estimated demand functions of the previous subsection are perfectly consistent with this postulate. The second postulate is important not only for its algebraic functionality, since it makes it very easy to compute society's indirect utility function, but also because it enables us to put more weight on products of widespread consumption relative to products of restricted consumption. 2. The analysis will focus on the portions of an individual's indirect utility function that relate to the demand functions for water, electricity, gas, gasoline and diesel, since the other components of the indirect utility function are held constant. By summing all residents' indirect utility functions and then dividing the result by the population size we obtain the average resident's indirect utility function. Let log w denote this individual's indirect sub-utility function consisting of the sum of the weighted logs of the demand functions for water, electricity, gas, gasoline and diesel. Hence, we have log w = 1 log x1 +2 log x2 +3 log x3 +4 log x4 +5 log x5 , D D D D D where i is the weight assigned to the log of the demand function for commodity i, i = 1,...,5, x1 is the demand function for water, x2 is the demand function for electricity, D D x3 is the demand function for gas, x4 is the demand function for gasoline and x5 is the D D D demand function for diesel. In the solution to the consumer optimization problem, we know that the weight i corresponds to commodity i's budget share. Since we are dealing with the average individual, i should then be commodity i's average budget share. Hence, consistent with the information presented in the first column of Table 8, the weights should be: 1 = 0.0417, 2 = 0.0304, 3 = 0.0574, 4 = 0.0164 and 5 = 0.0028. Accordingly, the average Luanda resident is assumed to have the following indirect sub-utility function: 85 log w = 0.0417log x1 + 0.0304log x2 + 0.0574log x3 + 0.0164log x4 + 0.0028log x5 . D D D D D 3. Considering only the statistically significant estimates (including constants) and neglecting the error terms of the estimations, we are now able to obtain our fundamental equation: log w = 0.009 + 0.059loge -0.053log pWater - 0.006log pLPG - 0.010log pGasoline - 0.004log pDiesel, where e denotes expenditure, pWater is the price of water, pLPG is the price of gas, pGasoline is the price of gasoline and pDiesel is the price of diesel. The main results of the analysis so far are now immediate: (i) if expenditure increases by 10%, welfare rises by 0.59%; (ii) if the price of water increases by 10%, welfare shrinks by 0.53%; (iii) if the price of gas increases by 10%, welfare falls by 0.06%; (iv) if the price of gasoline increases by 10%, welfare drops by 0.10%; (v) if the price of diesel increases by 10%, welfare reduces by 0.04%. 4. Although the marginal welfare effects are all small in absolute terms, it is worth noting that the water price effect is significantly larger than every fuel price effect and that the gasoline price effect is larger than the gas price effect. The latter result follows from the fact that gas is viewed as a substitute to electricity ­ the demand for electricity is positively related to gas price ­, which then reduces the negative gas-price-welfare effect. 86 APPENDIX IV: INCEPTION REPORT, MARCH 2005 87 THE WORLD BANK ANGOLA Assessing the Impacts of Phasing Out Utility Price Subsidies Inception Report By Emilson Silva Consultant - The World Bank March 2005 88 Acknowledgments This inception report has been drafted by Emilson Silva, Associate Professor at Tulane University and consultant to the World Bank for this task. The consultations described in the main text took place during the period of February 14 to February 25, 2005 in Luanda. Camilla Rossaak, Social Scientist at the World Bank, was an integral member of the team. Her inputs during consultations, comments on the issues to be discussed, suggestions for changes in the approach adopted during the interviews, help with setting up appointments and overall contributions to the making of the questionnaire to be adopted in gathering household data were and have been extremely valuable. The financial support provided by the British Embassy in Luanda to this task is gratefully acknowledged. The wisdom and experience of Mr. John Thompson, British Ambassador in Luanda, have been very useful in refining the scope and the orientation of the work. Many thanks go also to all governmental officials who warmly welcomed the team in their offices and eagerly discussed with them the relevant issues. Many of these individuals also provided the team with key data sets, which will later be extremely important in the development of the economic analysis. The team kindly acknowledges the support from the Ministry of Finance through Vice-Minister Eduardo Severim de Morais, Manoel Neto da Costa, Director of Studies and International Relations, and Francisca Fortes, Director of Prices and Competition. The civil society organizations, donors and professors who made themselves available during the consultations offered very important pieces of information. This work would not have been possible without the cooperation and assistance of many experts and staff at both World Bank headquarters in Washington and the local office in Luanda. A great number of people at the headquarters helped the team in the conceptualization of the study and on the drafting of the questionnaire. The group of dedicated staff who helped in the early stages of this work includes Masami Kojima, Stefano Paternostro, Sarah Keener, and Louise Fox. As for the local World Bank team, Mr. Laurence Clark, Mr. Olivier Lambert, Ms. Domingas Pegado, Ms. Carla Balça and Mr. Lima da Silva, the team would like to say "muito obrigado por tudo". Finally, the team would like to thank Francisco Carneiro, Senior Country Economist for Angola, for his support at all stages of this work, since its inception in Washington, DC. It has been a real pleasure to conduct this work under his supervision. I. Background In a policy program scheduled to be completed by October 2005, the Government of Angola (GoA) has been gradually eliminating fuel price subsidies. When this program is complete, subsidies of public utilities (water and electricity) are supposed to be similarly phased out. By eliminating all subsidies, the GoA's fiscal position is expected to improve, since net revenues should increase ­ according to the Ministry of Finance, the GoA subsidy liability in 2004 amounted to US$ 1 billion ­ and an agreement with the 89 IMF should be facilitated. However, the increases in fuel and utility prices may lead to greater adversity for the vulnerable and poor. The adverse effects may be felt through a higher cost for the basic basket of goods and services or through increased hardship in having access to the various items of such basket. In the context of the Bank's Country Economic Memorandum and of the negotiations of a Staff Monitored Program between the Government and the IMF, with financial support from the British Embassy in Luanda, the Bank is commissioning an impact analysis of the Government's policy aimed at eliminating utility price subsidies. The main goal of the study is to estimate and evaluate the likely effects associated with the removal of such price subsidies on the distribution of household welfare in Luanda (both urban and rural), government accounts2 and inflation in Angola. The study will also capture the linkages [market (supply and demand sides), institutional and political (stakeholders)] that may exist between fuel and utility pricing, given the broader objective of examining removal of all price subsidies. Overall, the analysis will foster policy debate and feed relevant (quantitative and qualitative) information back into policy choices. This report highlights the issues raised in stakeholder consultations undertaken in Luanda during a two-week period ranging from February 14 to February 25, 2005. The group of stakeholders consulted consisted of governmental officials, civil society organizations, donors, researchers and economics professors. The meetings produced a gamma of inputs, data and suggestions for reshaping the scope of the study. For a better understanding of the changes promoted as a result of the consultations, the original scope of the analysis is presented next in what follows. The main issues raised during the stakeholder consultations are provided in section III. In section IV, the scope of the study is revised according to the inputs obtained during consultations. Section V discusses the methodology to be adopted, and section VI displays the timeline for the remainder of the project. 2In this instance, the study will complement earlier analysis carried out jointly by staffs of the World Bank and the IMF, see IMF (2003), Fiscal Subsidies in Angola, in Angola: Selected Issues and Statistical Appendix, 2003 Article IV Consultation Report, Washington, DC. 90 II. Original Scope The original scope of the study, which the team responsible for the impact analysis (Emilson Silva and Camilla Rossaak) explained to stakeholders in the beginning of the consultation period, was as follows. The analysis would consider the social and economic impacts of phasing out price subsidies for fuels and utilities. The study would investigate both macro and micro economic aspects of the problem at hand. The macroeconomic components of the study would include the possible impacts on governmental finances and inflation. The microeconomic impacts would be assessed by gathering information from households in what respects their demands for fuels and utility services (electricity and water). This would be done through a household survey, which would be conducted throughout the country, encompassing urban and rural areas. The original idea was to concentrate our efforts in understanding the impacts associated with the phasing out of fuel price subsidies, since the government had already started a program of eliminating such subsidies. We would also try to understand the issues pertaining to pricing of and access to utilities, but this investigation would not be the primary source of motivation for the study. There was also the intention of putting more emphasis on the social problems surrounding pricing of and access to energy sources rather than to water sources. In some consultations we also mentioned that the study would consider different scenarios concerning the speed of price adjustment and the trade off between rules and discretion. As for the speed of adjustment, the goal was to evaluate whether the adjustment should be made once and for all or gradually. As for the trade off between rules and discretion, the study would consider the pros and cons of establishing trigger rules of adjustment (e.g., tying within a narrow band domestic fuel prices to international prices) and contrast them with the current state of affairs where government authorities are given full discretion over price adjustments. The analysis in either case would fully account for political considerations, including suggestions for palliative measures that should be adopted in order to mitigate some of the negative effects associated with increases in fuel prices. In broad terms, the study would be structured to address the following main questions: I. What are the problems associated with subsidies? Who benefits from them? II. What are the likely impacts of eliminating/reducing subsidies on households and on the (macro) economy? Who are the likely winners and losers? III. How to design a program to compensate the poorest among the losers and also to moderate resistance? In order to address the questions above, it was deemed necessary to focus on the following issues: 1. Incidence of subsidies: What is the incidence of subsidies by income groups and per person in Angola? What are the characteristics (income and socioeconomic) of the various segments of the population affected by subsidies? How much do 91 Angolans pay for fuel and utility services in comparison to other countries in Africa and other parts of the world? 2. Access to fuels: Which energy sources do urban and rural households typically use for heating, lighting and cooking? How accessible are the fuels for the urban and rural poor? How substitutable are the fuels for the different types of consumers? 3. Size and consequences of price adjustments: By how much should prices be raised, and what are the timing issues involved (short and long term)? What are the estimated effects of rising fuel prices on food prices? Which segments of the population will be directly (indirectly through rising transportation and input costs) affected by rising fuel prices? Which segments of the population, if any, will not be affected at all by rising fuel and utility prices? 4. Secondary effects: Are there secondary markets (resale) for fuels and water in urban and rural areas? If so, how do they operate and what is the price premium charged? Are the officially set prices the effective prices charged for fuels? How does the distribution mechanism work? Are there unofficial intermediaries? 5. Operational efficiency: How does Sonangol operate as a distributor? What are Sonangol's operation costs? How can the pricing mechanism be more transparent? How can sector efficiency be enhanced, either through pricing or other mechanisms? How can political influence be minimized in price setting? How can consumers benefit from sector efficiency enhancement? 6. Use of fiscal savings: How much will the GoA save with the elimination of the subsidies? Are the savings obtained sufficiently large to compensate the poorer households and improve social protection? Should the GoA improve or augment existing social programs with the savings? How to phase fiscal savings relative to spending? 7. Monitoring and evaluation: Which monitoring and evaluation systems do exist and/or should be put in place to assess both the impacts of (i) the actual subsidy removal on households and on the economy; and (ii) the spending programs put in place to compensate the losers? III. Fruits from Stakeholder Consultations Starting on February 14 and running until February 25, 2005, the World Bank team in charge of the subsidy study conducted consultations with several governmental officials, civil society groups, donors and economics professors. The team met with the following governmental officials: 92 · Mr. António da G. Lopes Teixeira, General Associate Director, Angola's Roads Institute (Instituto de Estradas de Angola (INEA)). · Mr. Rui Augusto Tito, Vice Minister, Ministry of Energy and Water Affairs. · Mr. Paulo Matos, Director, Ministry of Energy and Water Affairs. · Mr. Bonifácio Manuel, Director, Ministry of Agriculture and Rural Development. · Ms. Luzia B. da Costa, Head, Department of Planning and Statistics, Ministry of Petroleum. · Mr. Mavinga B. David, Head, Department of Investments and Projects, Ministry of Petroleum. · Mr. Manuel da Costa, Director, Office of Studies and International Economic Relations, Ministry of Finance. · Ms. Francisca Fortes, Director, Office of Prices and Competition, Ministry of Finance. · Ms. Efigénia da Purificação S. S. Martins, Head, Department of Public Enterprises, Ministry of Finance. · Ms. Joana Cordeiro dos Santos, National Director of Accounting, Ministry of Finance. · Mr. Simão Neto, Director, Office of Informatics, Ministry of Finance. · Mr. Joaquim Flávio de Sousa Couto, General Director, National Statistics Institute (Instituto Nacional de Estatística (INE)). · Mr. Domingos Bernardo, Technical Associate, Office of the President of the Administration Council, Water Public Enterprise (Empresa Pública de Águas (EPAL)). · Mr. José Ambriz, Administrator, Office of the President of the Administration Council, Water Public Enterprise (Empresa Pública de Águas (EPAL)). Non-governmental and civil society organizations consulted included FAS, FMEA, CARE ­ Angola, YME, CEEA, DW, SAL, ADRA, COIEPA, FONGA, MEDAIR, AIA, Angola 2000. The donors consulted included British Embassy, UNDP, US Embassy, South African Embassy, WHO, Swedish Embassy, German Embassy, Norwegian Embassy, Italian Embassy, Spanish Embassy, UNFPA, European Commission, UN Human Rights. The team also met with Professors Laurinda Hoygaard, University Agostinho Neto, and Justino Pinto de Andrade and Manuel Alves da Rocha, Catholic University of Angola. The main issues raised during consultations with governmental officials are summarized below. The list is organized according to the chronology of the consultations. 1. INEA · Intercity roads are in very poor condition; around 90% of them are damaged. Access to some places inland is only possible via air (airplane or helicopter). · Transportation cost is very high. 93 · Food is much more expensive in the countryside than in coastal areas because of transportation costs. · Roads along the coast were de-mined; there is a "safe" perimeter ranging from 8 to 12 meters along these roads. · Massive investment is necessary in order to improve the country's road system. 2. Ministry of Petroleum · The refinery in Luanda produces 50% of the fuels utilized in the country. · Fuels produced in the refinery are sold to Sonangol Distribuidora, which is in charge of distributing the products to gas stations and other resale outlets. · Most gas stations belong to Sonangol. There are a few small private gas stations distributed across the country, mainly in urban areas. · The Ministry of Petroleum is currently working on implementing a law designed to increase competition in distribution and resale of fuels. · A large fraction of gasoline consumption is for generation of electricity (private generators). · Unavailability of gas stations in the countryside generates the conditions for the emergence of secondary markets. Hence, a policy that aims to increase the number of operating gas stations in remote areas in the provinces is of paramount importance. · The country is a net importer of gasoline. Its single refinery's productive capacity cannot match the market demand. · Fuels are not smuggled out of the country. (There appears to be no restriction in reselling the product abroad.) · The main reason for keeping low fuel prices is to compensate the population for low salaries. 3. Ministry of Finance · Due to the current transition period, the issue of eliminating fuel subsidies is very delicate. The practice of controlling prices has been inherited from the previous communist system, characterized by a command-and-control economy. 94 · Civil service salaries are low. Nobody knows for sure the distribution of salaries in the private sector. In any case, increases in fuel prices generate pressure to increase salaries in public and private sectors alike. One must then account for the inflationary impacts associated with such measures. · Last year, the government spent around US$ 1 billion in subsidies. There is no doubt that governmental finances would improve if subsidies were eliminated. · As for water and electricity, rises in prices would not solve the problems. The public companies are poorly administered; workers have low skills and are known to shirk their working duties. The services provided by these companies are of poor quality. Any attempt in increasing prices would then encounter public resistance. The public is willing to pay higher prices for these services only if their quality levels improve. · One common problem generated by a rise in fuel price is the subsequent increase in urban transportation fares, as the rise in operating cost is shifted in its entirety to consumers by van operators. Increases in transportation fares have in some occasions led to social unrest, protests and discontentment amongst the urban population that relies on this mode of transport. · The majority of Luanda's urban population lives in the periphery. It is mostly affected by the impacts of the rising fuel prices on transportation and secondary markets. · The government should not subsidize fuel prices, since such subsidies are regressive, benefiting the rich and not the poor. The government should instead subsidize industry and agriculture, the productive sectors. · One of the critical problems faced by the population which resides in the countryside is the fact that Sonangol has not systematically distributed fuels to the provinces. Due to high transportation costs, fuel prices are sometimes twice as large in the provinces as in Luanda (official prices). Sonangol often hires private transportation companies to distribute fuels in the provinces. · The electricity companies determine electricity charges. One must also account for cross subsidization, since electricity companies use subsidized fuels to generate electricity. · Public servants at public enterprises earn higher salaries than officials in other public sector areas. The wage bill in the public sector as a whole last year amounted to US$ 2 billion, which represented 10% of GDP. · Between February and May, 2005, adjustments in the prices of fuels will occur on a biweekly basis. 95 4. INE · The raining period will last until the end of April. As a result, access to the countryside is extremely difficult prior to the beginning of May. INE's field interviewers typically start their operations after the end of the raining period. In designing our strategy regarding the areas selected for the survey, we must take the restrictions imposed by the weather into account. · Due to the difficulties in accessing remote areas in the provinces, even during the dry season, nationwide surveys tend to be very costly. · The last census was carried out in 1970. There are no accurate or official estimates of population size and its regional distribution. · The Angolan population size is likely to be 12 million, of which 4 million is expected to reside in Luanda. · Most of the Angolan population is believed to reside in urban areas. · Luanda is different from other urban centers in many respects; however, the social indicators for Luanda are likely to be representative of the whole country. · Two household surveys were previously conducted, one in 1997 (4000 respondents) and another in 2001 (6600 respondents). While the first survey covered areas controlled by UNITA, the second did not. Such areas have not been surveyed since 1997. · A study which focuses on Luanda (urban and rural areas) would be very interesting, since it has not been done before. 5. Sonangol · Subsidy figures are calculated based on proposed Sonangol delivery prices for fuels in Luanda. Hence, transportation and other logistic costs to deliver products to provinces are typically ignored. This implies that the overall amount of subsidization is higher than the official figure. · Sonangol subcontracts with private transportation companies to deliver its products in the provinces. Sonangol does not have a sufficiently large number of trucks to supply fuels to all gas stations across the country. · There are 240 gasoline pumps in the whole country. These do not include pumps serving boats or offshore activities. 96 · Angola's single refinery was built in 1949. It lacks capacity to meet the market demands for fuels. 6. EPAL · New water treatment systems coexist with old ones in the provision of water. This coexistence negatively affects water quality delivered to the population. · Electricity and fuels are crucial inputs in the generation of water services. The company is in debt with Sonangol and electricity companies. · Operating plants provide water of good quality. However, there are problems in the distribution network. The network is old and allows infiltration because of its many holes. Since the sewage network system is located near to the water network and sewage spillovers are common, there is a high degree of contamination. Luanda's sewage system is controlled by the provincial government, which has failed to prevent the system's deterioration. · The World Bank financed a project called PRP ­ Programa de Reabilitacao Prioritaria da Rede Esgoto da Cidade de Luanda. The project was not carried out by the local authorities. · There is a law that enables a private operator to manage water provision and services. However, no firm has yet been granted the right to conduct such operations in the water sector. · Currently, one third of Luanda's population has water piped to their homes. · The company does not raise adequate revenues to finance its operating costs. This is due to a number of factors: (i) water charges are low; (ii) many clients fail to systematically pay their bills; (iii) company officials do not exert much effort in measuring clients' water consumption levels; and (iv) there are several clandestine connections to the water system, some of which are facilitated by company officials. The resulting revenue collected enables the company to only pay its wage bill. Other operating costs are typically financed by the central government. However, governmental transfers and subsidies are insufficient and irregular. Money disbursements do not appear to follow any systematic policy. · There are two types of water trucks responsible for delivering water to the population: (i) those under contract with the company; and (ii) those who are autonomous operators. Since there is no effort made to distinguish them, the water quality provided by autonomous operators is not regulated and cases of corruption are not uncommon, it is likely that the water quality provided in this market is low. 97 · Water charges were adjusted in May, August and October, 2003. The adjustments, however, were simply to keep up with inflation. Up to now there has been no attempt to increase charges in order to reduce the price subsidy. · The central government has been investing in expanding capacity. There has been no attempt to substitute old for new plants; hence, System III (new plant) is not a substitute for Systems I and II (older plants). The government is also spending some resources in the improvement of System II. · System III is not yet producing large social benefits because the distribution network is limiting and in poor condition. The expansion in capacity will only bring about greater regional water provision if the current distribution network is improved and expanded. · Urban population does not complain about the services provided by the company; no manifestations against increases in charges or in favor of improving water quality are observed. · People in rural areas typically pay more for water services. · Retail businesses should not benefit from price subsidies. · Water tariffs should be differentiated according to ability to pay, with the rich paying more than the poor for the same service and same quality. · The company is currently preparing a report, to be submitted at the end of March to the Ministry of Finance, in which it proposes differentiated tariffs. · The company faces some very serious managerial problems: (i) it lacks resources to commercialize and distribute its products; (ii) lacks skilled and motivated human resources in key areas, such as accounting and management; and (iii) it employs more workers than it needs. · Unskilled and skilled workers alike are unmotivated. There is a general perception that salaries are low, although they are higher than the average salaries paid in the public sector as a whole. Salaries paid by other public companies, however, are perceived to be higher. · Most employees belong to a politically active labor union. 7. Ministry of Energy and Water · The ministry is currently studying alternative ways to reduce the gaps between marginal costs and prices of utilities, in an effort motivated by the government's orientation in reducing subsidies. 98 · There does not seem to be logical to subsidize retail and service sectors, since these sectors are modern and competitive. · There is, perhaps, justification to keep price subsidies in utility services delivered to consumers. · It is common knowledge that the public utility companies are not efficient. · In the energy sector, privatization is thought to be a good strategy for development. · 20% of the Angolan population is estimated to have access to electricity. · Current average electricity consumption ranges between 90 to 100 Kw/h. There is a goal to increase this consumption level by 100% by 2015. However, new investments are needed in order to reach this target. · As for the water sector, there is a strategic program already in place and which should last until 2016. · Water treatment and distribution should be handled, perhaps, by a private company. · Water business is very lucrative. The informal sector raises annual revenues of about US$ 100 million. EPAL's annual revenues, on the other hand, are around US$ 10 million. · Shock treatment does not work: effective and sustainable price increases have to be gradual and of small magnitudes. · New investments in capacity expansion are sometimes fruitless because the new added capacity ends up retiring previously built capacity and the net effect is nil. The main issues raised during the meeting with the various civil society organizations are summarized below. Non-Governmental and Civil Society Organizations · Water and energy companies are inefficient. Clients find it difficult to make payments. · Access to water and energy sources is a problem of gigantic proportion, since it affects all social levels. The problem is more serious for the poor. There are, 99 for example, people in Luanda that have not had access to either water or electricity over the last three years. · Even the rich have private water tanks at home, purchase water from water trucks and have electricity generators at home. Public utility services are of poor quality. Much of the problems in poor service delivery are generated by the public companies themselves. · Price increases for utilities can be counterproductive in that only a few honest people would pay. In any event, prior to increasing tariffs, service quality must improve. · The inefficiencies created by the public utility companies in service delivery generate high costs in the production of goods and services in the private sector. New ventures in the private sector can only be realized if entrepreneurs are able to finance the costs of acquiring water tanks and electricity generators for their businesses. · Luanda's population has been growing at a very high rate. The public utility companies are unable to meet the demands of the expanding population. · The chaotic situation observed in Luanda caused by its rapidly growing population should be the focus of a study, which investigates the conditions that should be created to induce those who migrated to the city from rural areas to return to their original places of residence. Subsidies should be directed to rural development. · The production of a new demographic census is of extreme importance. The World Bank could help the Angolan government in this enterprise. · Fuel price subsidies favor only a small segment of society. Hence, the subsidies should be eliminated and hence fuel prices increased. The governmental resources saved with the elimination of fuel price subsidies should be used to: (i) invest in priority areas such as education; (ii) subsidize student transportation; and (iii) develop agriculture and industry. · Fuel price increases can have very detrimental effects on education, as transportation costs faced by students are high. Schools in Luanda are all located in the center of the city. Students often have to travel long distances to attend school. As fuel price increases generally cause increases in transportation fares, students may decide to quit attending school in order to avoid the larger transportation cost burden. · Current governmental policy provides subsidies to transportation companies. These subsidies are misguided, since they should be directed to consumers, not the firms. 100 · There is no current effort in coordinating the actions of federal, provincial and municipal levels of government. Some of the problems encountered in water provision stem from this lack of coordination, since EPAL is a branch of the federal government while the sewage system is responsibility of the provincial government. Prior to meeting with the donors, the team had the opportunity to meet with Mr. John Thompson, the British Ambassador. Mr. Thompson raised the following issues: · Prices of fuels have dramatically increased over the last nine months and yet no negative public reaction has been noticed. This is true even for kerosene, which is widely used in the countryside. · The study should not ignore the water sector; the poor are likely to be more affected by lack of access to water than to energy sources. The secondary water market makes the price of water prohibitively expensive. · Due to the logistic difficulties generated by the raining season in accessing the countryside, he suggested that Luanda should be the focus of the study. The problems facing the poor population in Luanda regarding lack of access to public services and hefty prices paid in secondary markets are exemplary of the rest of the country. At the meeting with donors, the following points were raised. Donors · Some thought that having Luanda as the focus of the study would be quite limiting, since the economic characteristics of the provinces are quite distinct from the ones in Luanda. Others argued that Luanda is in many ways representative of the whole country, especially if one considers the social aspects involving lack of access to water and energy sources. · As for the questionnaire to be used, some cautioned that in many of the households interviewed people cannot read or write. Some of the households also may not speak Portuguese. It may be necessary to have a team of Angolan interviewers that are able to communicate with the various ethnic groups residing in Luanda. · When conducting this study, one should consider human rights issues. It is commonly said that "education is free," but this is not really true. Hence, what are the effects of phasing out subsidies on the human rights situation and on the quality of services provided? These are questions that need to be addressed. 101 · How should the governmental resources saved be used? Policy recommendations are needed. · One should understand the managerial and operational issues facing the public utility companies. Are there alternative solutions for the provision of services? Should they be privatized? How can the services be made viable? These are issues that should be addressed in the study. · Is there a connection between the rationale for existence of subsidies and the prevailing salaries in the public sector? · Political questions must be addressed, since policy sustainability is highly dependent on political considerations. · Given that current demands for services have been conditioned on supplies available, one must be careful in drawing conclusions for future policy making. · The economic and social environments in peri-urban Luanda are the most important ones for this study. The main issues raised during the consultations with Professors Justino Pinto de Andrade and Manuel Alves da Rocha were as follows. · There is no economic justification for fuel price subsidies in Angola, since they do not spur economic growth or increase the competitiveness of the productive sectors in the economy. · Fuel prices should rise, but gradually rather than at once. Such price increases should be accompanied by economic stabilizers in order to reduce the social and economic negative impacts. · The government currently has a policy program to gradually adjust fuel prices. The target is to periodically increase the price of gasoline, for example, so as to reach 77 Kwanzas per liter in October 2005. · Politicians are more inclined to be motivated by ideological questions than by practical issues facing the society. · Bureaucrats are typically unmotivated to change the statuts quo. · Throughout Luanda one finds rich and poor people living side by side. After the Portuguese left, many households invaded the residential buildings that were previously occupied by the Portuguese families. The Angolan occupants were typically poor. Some of them, however, have been fortunate enough to get quite rich since then. Although there is now substantial difference in terms 102 of wealth among some of the originally poor households, their cultural backgrounds have remained the same. It is this common cultural background that enables them to coexist peacefully with each other. · Macroeconomic stability is of fundamental importance in Angola. As the government is not self-disciplined in its control of fiscal and monetary policies, a set of rules imposed by the IMF, which diminishes governmental policy discretion, is highly welcome. IV. Revised Scope The main lessons learned during the consultations are as follows: 1. There are serious obstacles in implementing a countrywide study within the timeframe for this project. The raining season makes access to the countryside extremely difficult. 2. The population currently living in urban and rural vicinities of Luanda is estimated to be around 4 million, which is likely to represent one third of the Angolan population. 3. Luanda's economic characteristics are not representative of the country's economic characteristics. However, the social indicators for Luanda are likely to be very representative of those for the country as a whole. 4. The inadequacies associated with water provision in Luanda appear to generate social problems of large magnitudes, since a small fraction of households have access to the publicly provided water services, and those who consume water at home may face severe disease problems associated with contamination from the sewage system. The informal water markets are highly lucrative. 5. Many households have electricity generators at home or share them with neighbors as a result of the unreliable electricity service provided by the electricity companies. 6. There is a high degree of public discontentment with the qualities of water and electricity services provided by the public utility companies. General opinion is that qualities would have to improve prior to the elimination of price subsidies present in water and electricity tariffs. 7. Public utility companies are criticized for being inefficient. Employees of such companies are thought of being unmotivated and to shirk their responsibilities. Some claim that the salaries paid by the public sector are low. However, the salaries paid by public enterprises are, on average, higher than those paid in the public sector as a whole. 103 8. There appears to be general agreement that fuel prices subsidies should be removed. However, some indicate that such removal may adversely affect the poor because of the implied increases in the cost of the basic basket of goods and services (i.e., transportation services, food, energy and water services). Palliative measures would be needed to counteract these negative effects. In light of these lessons, as well as of the other issues and concerns raised during consultations, the following modifications to the original scope of the study appear to be highly desirable: 1. The microeconomic and social portions of the study should be focused on Luanda, encompassing its urban and rural areas. 2. The study should investigate the impacts associated with the removal of price subsidies for fuels and utilities alike, paying close attention to the linkages there exist among all the goods and services (i.e., fuels, electricity and water) examined. 3. The study should also consider the inefficiencies associated with the current supplies of water and electricity. The managerial and operational deficiencies of the public enterprises should be examined and potential solutions proposed. The macroeconomic portion of the study should be maintained intact. V. Methodology There are two main parts to the study: (i) the macroeconomic portion; and (ii) the microeconomic and social portions. As for the macroeconomic portion, the study will mainly investigate the likely effects of subsidy removal on the fiscal finance and inflation. An analysis of official macroeconomic data will be conducted in order to complete this part of the study. The data analysis will also estimate the likely impacts of price increases for fuels and utilities on the basic basket of goods and services in Luanda. As for the second part of the study, both the demand and supply sides of energy and water markets will be considered. A theoretical model will be used to derive the testable hypothesis. The demands will be estimated with the data gathered through a survey of a sample of households located in Luanda's urban and rural areas. The estimation will also allow us to derive the welfare impacts associated with various hypothetical situations concerning the rates at which subsidies are removed. The questionnaire which will be used to gather household data is ready. 104 VI. Timeline for the Remainder of the Project Survey Questionnaire: Contains questions pertaining to urban and rural households' water and energy consumption, expenditure, access to water and energy sources and their perceptions about the quality of the products and services consumed. Dataset: Data collection will be carried out by an international firm in conjunction with an Angolan expert to assist it with sampling design and logistics as well as Angolan interviewers. The international firm will also be in charge of processing the data and generate the dataset in appropriate format for analysis. Collection efforts should start at the end of March 2005. Draft Report: Upon delivery of a satisfactory dataset, Emilson Silva will prepare a draft report that incorporates the qualitative information learned in the stakeholder analysis into the theoretical models, derives new testable hypotheses, estimates the welfare impacts of the reform under different policy scenarios, examines the likely implications of rising fuel prices on inflation and on the government accounts, analyses the results and provides policy recommendations. The analysis and recommendations will also address the likely effects of the reform on utility pricing and provision and be inclusive of the issues listed under the seven sub-titles listed in Section II above. The draft report should also be discussed in Luanda with the authorities and stakeholders by the end of May 2005. Final Report: Incorporates feedback from the second round of consultations with the Government and stakeholders. To be concluded by the end of June 2005. 105 APPENDIX V: RETAIL PRICES OF BASKETS OF GOODS AND SERVICES IN LUANDA 106 APPENDIX V: Retail Prices of Baskets of Goods and Services in Luanda Low and Medium Income Households Table A66: Retail Prices of Basket of Goods and Services - Low Income Households 2004 Product Quantity Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Water 3 37.50 37.50 37.50 37.50 60.00 60.00 60.00 60.00 60.00 60.00 60.00 117.00 Electricity 150 286.00 286.00 286.00 286.00 406.00 406.00 406.00 406.00 406.00 406.00 406.00 406.00 LPG, Kg 12 122.40 122.40 122.40 122.40 210.00 210.00 210.00 210.00 210.00 210.00 294.00 378.00 Transportation - Collective Taxi 60 1,500.00 1,500.00 1,500.00 1,500.00 1,800.00 1,800.00 1,800.00 1,800.00 1,800.00 1,800.00 1,800.00 2,310.00 Transportation ­ Bus 60 900.00 900.00 900.00 900.00 1,500.00 1,500.00 1,500.00 1,500.00 1,500.00 1,500.00 1,500.00 1,500.00 Cost of Fuel, Utilities & Transportation 2,845.90 2,845.90 2,845.90 2,845.90 3,976.00 3,976.00 3,976.00 3,976.00 3,976.00 3,976.00 4,060.00 4,711.00 Cost of Other Goods 18,151.94 18,089.18 18,148.83 18,232.11 18,726.38 19,113.16 19,105.03 19,024.72 18,850.80 18,908.50 19,218.11 19,614.32 Total Cost 20,997.84 20,935.08 20,994.73 21,078.01 22,702.38 23,089.16 23,081.03 23,000.72 22,826.80 22,884.50 23,278.11 24,325.32 Fuel, Utilities & Transportation/Total (%) 13.55 13.59 13.56 13.50 17.51 17.22 17.23 17.29 17.42 17.37 17.44 19.37 Monthly Cost Change 0.74% -0.30% 0.28% 0.40% 7.71% 1.70% -0.04% -0.35% -0.76% 0.25% 1.72% 4.50% Cost Index (Aug 2002=100) 210.42 209.79 210.39 211.23 227.50 231.38 231.30 230.49 228.75 229.33 233.27 243.77 Source: Ministry of Finance. 107 Table A67: Retail Prices of Basket of Goods and Services - Medium Income Households 2004 Product Quantity Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec Water 6 75.00 75.00 75.00 75.00 120.00 120.00 120.00 120.00 120.00 120.00 120.00 234.00 Electricity 500 1,180.00 1,180.00 1,180.00 1,180.00 1,675.00 1,675.00 1,675.00 1,675.00 1,675.00 1,675.00 1,675.00 1,675.00 Gasoline 100 1,200.00 1,200.00 1,200.00 1,200.00 2,000.00 2,000.00 2,000.00 2,000.00 2,000.00 2,000.00 2,700.00 3,400.00 LPG, Kg 12 122.40 122.40 122.40 122.40 210.00 210.00 210.00 210.00 210.00 210.00 294.00 378.00 Cost of Fuels & Utilities 2,577.40 2,577.40 2,577.40 2,577.40 4,005.00 4,005.00 4,005.00 4,005.00 4,005.00 4,005.00 4,789.00 5,687.00 Cost of Other Goods 25,018.03 24,881.97 25,098.93 25,178.95 26,011.14 26,561.99 26,575.56 26,327.36 26,151.32 26,165.16 26,688.35 27,219.03 Total Cost 27,595.43 27,459.37 27,676.33 27,756.35 30,016.14 30,566.99 30,580.56 30,332.36 30,156.32 30,170.16 31,477.35 32,906.03 Fuels & Utilities/Total (%) 9.34 9.39 9.31 9.29 13.34 13.10 13.10 13.20 13.28 13.27 15.21 17.28 Monthly Cost Change 1.64% -0.49% 0.79% 0.29% 8.14% 1.84% 0.04% -0.81% -0.58% 0.05% 4.33% 4.54% Cost Index (Aug 2002=100) 190.05 189.11 190.61 191.16 206.72 210.51 210.61 208.90 207.69 207.78 216.78 226.62 Source: Ministry of Finance. 108 wb04321 L:\WORD\ANGOLACEM\SUBSIDIES_APPENDIXFINAL.doc 12/1/2005 10:40:00 AM 109