Burundi - Second Coffee Improvement Project

[ocument of The World Bank FOR OFFICIAL USE ONLY Report No. 4576 BURUNDI SECOND COFFEE IMPROVEMENT PROJECT (CREDIT 593-BU) PROJECT COMPLETION REPORT June 23, 1983 Eastern Africa Projects Department Regional Mission in Eastern Africa This document has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization.  Currency Equivalents Currency Unit Franc Burundais (FBu) Appraisal Year Average: US$1.00=FBu 78.35 Intervening Years Average: US$1.00 = FBu 90.00 Completion Year Average: US$1.00 = FBu 90.00 Weights and Measures Decimal Metric = US/British 1 kilogram (kg) = 2.205 pounds (lb) 100 kilograms (kgs) = 1 quintal (qt) I metric ton (mt) = 2,205 pounds (lb) 1 hectare (ha) = 2.4711 acres (ac) 1 kilometer (km) 0 0.62 miles (m) Abbreviations BCC = Burundi Coffee Company CAMOFI = Caisse de Mobilisation Financiere ISABU = Institute des Sciences Agronomiques du Burundi MINAGRIC _ Ministry of Agriculture OCIBU = Office des Cultures industrielles du Burundi SOMEBU _ Societe Mixte d'Etudes du Burundi  FOR OFFICIAL USE ONLY BURUNDI SECOND COFFEE IMPROVEMENT PROJECT (CREDIT 593-BU) PROJECT COMIPLETION REPORT Table of Contents Page Preface ...., ...................... , I* Basic Data Sheet. . . ........ *........ ii Highlights ....... ... ........... ..... iv I BACKGROUND.... 1 II IDENTIFICATION, PREPARATION AND APPRAISAL . . 3 A. Identification and Preparation. 3 B. Appraisal. 4 C. Project Objectives and Description. 6 III IMPLEMENTATION. 7 A. Start-up of the Project. 7 B. Coffee Production Activities and Achievements. 8 C. Research 16 D. Coffee Processing .18 E. Foodcrop Production Activities and Achievements .22 F. Other Project Activities .30 G. Financial Aspects .32 IV INSTITUTIONAL PERFORMANCE .35 A. Organizational Aspect .35 B. Project Management .36 V PROJECT IMPACT ..39 A. Coffee Production and Yields .39 B. Production and FWA Coffee .42 C. Coffee Prices .44 D. Farmers' Income .45 VI ECONOMIC ANALYSIS ....46 VII IDA PERFORMANCE ....49 VIII CONCLUSION AND OUTLOOK ....50 Annex No. TABLES I Burundi Coffee Exports by Quantity and Quality II Burundi Coffee Exports: Totals III Burundi Coffee Exports - Five-Year Moving Average IV Pyrethrum Test - No. of Insects Per Tree This document has a restricted distribution and may be used by recipients only in the performance of their official duties. Its contents may not otherwise be disclosed without World Bank authorization.  V Inventory of Buildings Constructed During Second Phase VI Financial Analysis - Project Investments VII Financial Analysis - Project Results VIII Increase in Project Cost Ix Schedule of Disbursements X Expatriate Staffing Schedule Xr Production of Arabica Coffee in the Ngozi Province XII Coffee Tree Population in Ngozi Province with Projected Annual Yields XIII Production of Arabica Coffee in Ngozi Province xiv Production of Arabica Coffee in Ngozi Province - Five Year Moving Average XV Coffee Price Differentials XVI Economic Rate of Return Calculations XVII Economic Rate of Return Calculations  - i - Project Completion Report BURUNDI SECOND COFFEE IMPROVEMENT PROJECT Preface The project under review was a second phase in the Government of Burundi's program for improving the quantity and quality of national coffee production. Coffee exports are Burundi's largest single earner of foreign exchange. The Credit for the Second Coffee Improvement Project (Cr. 593-BU for US$5.2 million) was signed in December 1975 and closed fully disbursed in April 1982. It had been preceded by a first phase project (Cr. 147-BU for US$1.8 million) and was followed by the signing of a Credit for a third, (Cr. 1192-BU for US$16 million) in March 1982. This Project Completion Report (PCR) deals with the second phase of the coffee improvement program but reference is made to aspects of the other phases of the program since it is not meaningful to tret the separate phases as totally discrete operations. Indeed, the findings of the Project Performance Audit Report and the Impact Evaluation Report for the Ngozi project (phase one of the program) contributed significantly to the analysis of the findings in this present report. This PCR was prepared by staff of the Bank's Regional Mission in East Africa. No separate audit was carried out for the Project by OED staff. It is based on field visits to the Project area (Ngozi), on extensive discussion with staff of the Project, OCIBU, and of the Ministry of Agriculture and on review of the Project files, including the preparation, appraisal, supervision and annual reports.   - ii- BU'RUNDI SECOND COFFEE IMPROVEMENT PROJECT (CREDIT 593-BU) PROJECT COMlPLETION REPORT BASIC DATA SHEET KEY PROJECT DATA Appraisal Actual or Actual as . Estimate Estimated Actual of ApDraisal Estimates Project Costs (US$million) 7.5 10.9 a/ 145S Project Amount (US$uillion) '5.2 5.2 1007. Date Board Approval 11-18-75 Date Effectiveness 03-03-76 10-07-76 Date Physical Components Completed 06-30-80 07-30-80 Proportion then Completed 86* 1OOS Closing Date 03-31-81 04-30-82 Economic Rate of Return 217 b/ Institutional Performance Good Fair *Agronomic Performance Good Fair Number of Direct Beneficiaries (year...) Estimated 90-100,000 cumulative in 1982 CUMULATIVE DISBURSEMENTS FY76 FY77 FY78 FY79 FY80 FY81 FY82 Appraisal Estimate (USS million) 0.4 1.5 2.5 3.5 4.5 5.2 Actual (US$ million) 1.7 2.7 4.2 5.1 5.2 Actual as Percent of Estimate 68* 77S 93* 987 1007. Date of Final Disbursement June 25 1982 MISSION DATA No. of Mandays Speacializations Performance Types of Mission Date Persons in Field Represented 3/ Ratin& 4/ Trend 5/ Problems 6/ (mo./Yr.) Identification 08-73 Preparation -74 2 N.A. Appraisal 10/74 4 64 2 a, b Subtotal Supervision 1 4/76 1 5 Supervision 2 12/76 1 8 Supervision 3 7/77 2 24 Supervision 4 4/78 2 24 2 b, c 2 2 X,F,O Supervision 5 10/78 1 2 1 b 2 1 T,O Supervision 6 3/79 1 6 1 b 2 2 F,T,O Supervision 7 11/79 1 8 1 b 2 2 M,F,O Supervision 8 7/80 1 6 1 b 2 2 M,0 Supervision 9 3/81 1 6 1 b 2 3 X Supervision 10 Supervision 11 (COMP) 5/81 2 24 Supervision 12 Supervision 13 Supervision 14 Supervision 15 Subtotal 177 Total 177   - jiii - OTHER PROJECT DATA Borrower Government of Burundi Executing Agency OCIBU Fiscal Year Jan 1 - Dec. 31 Name of Currency (Abbreviation) (FBu) Currency Exchange Rate: Appraisal Year Average USS 1.00 = FBU 78.35 Intervening Years Average US$ 1.00 = FBU 90.00 Completion Year Average US$ 1.00 = FBU 90.00 Follow-on Project: Name Integrated Rural Development/NGOZI III Loan/Credit Number 1142-BU Loan/Credit Amount (US million) 16.0 Date Board Approval 12-08-81 1/ Summarize reasons for deviations in footnotes 2/ Percentage to third column to be calculated from Board approval date. 3/ a = agriculturist; b = agricultural economist; c = financial activist, etc. 4/ 1 = problem-free or minor problems; 2 = moderate problems; and 3 = major problems. 5/ 1 = improving; 2 = stationary; ancl 3 = deteriorative. 6/ F = financial; N = managerial; J = technical; P = political; and 0 = other. a/ The increase reflects an overdue income of 157 of the BUF in May 1976. b/ Coffee growing: 247 11 processing: 227.   - iv - Project Completion Report BURUNDI SECOND COFFEE IMPROVEMENT PROJECT Highlights This report reviews progress under the Second Coffee Improvement Project, Cr. 593-BU for US$5.2 million. The Project was a second phase of the Government of Burundi's continuing program to improve coffee production, the nation's largest single earner of foreign exchange. The Project's components were aimed at bringing about increases in coffee and foodcrop production and comprised agricultural services elements, an experimental rural development program, a program of management capacity building, coffee research and coffee processing activities, and rural infrastructure rehabilitation works. Overall, the Project was a qualified success. Start-up was delayed by frequent changes in management and implementation and was adversely affected by management difficulties throughout. Project achievements for foodcrop and livestock development fell well below targets set at the time of appraisal; in retrospect, these appear to have been too optimistic. The experimental rural development and economic and social infrastructure components were not implemented. In contrast, the results of the coffee development components were good and exceeded appraisal estimates both as to quantity and quality of incremental production. It can be stated with confidence that the coffee-washing stations constructed under the Ngozi II Project are economically viable. This has been the principal achievement of the Project. The following points are also of interest: - the realistic approach to the use of fertilizer and to financing (para. 7.01); - an over optimistic assessment of the Borrower's and the Project management's implementation capacities (para. 7.01); - the beneficial aspects of staff continuity in Bank supervision combined with appraisal of planned follow-up activities (para. 7.03); - the dramatic effects of high inflation on civil works costs (para. 3.68) and on the projected income gains for Project beneficiaries (para. 8.08); and - the need for a more reliable system of surveying and reporting coffee plantation and production (para. 5.08).   BURUNDI SECOND COFFEE IMPROVEMENT PROJECT - CREDIT 593-BU PROJECT COMPLETION REPORT I. BACKGROUND 1.01 Coffee is the principal export crop and source of income in Burundi. In fact, Arabica coffee is grown by about half of the estimated 900,000 farm families in the country, generating between 40 to 607. of their cash income. Moreover, coffee has been of primary importance for the development of the modern sector in Burundi, as coffee exports are the principal earner of foreign exchange (80-907) as well as the main source of tax and other Government income. The main smallhclder coffee growing areas are Ngozi, Gitega and Muyinga Provinces; Ngozi provides over one-third of total production. Coffee is grown in small stands, on average about 160 trees per farmer, covering about 0.10 ha, often on several plots. Coffee yields vary from one region to another (200-450 grams/tree) with Ngozi having the highest yields because of its better soils, climate and as a result of the two Coffee Improvement Projects. In fact, it is generally recognized that because of poor husbandry methods, lack of modern inputs and mulching material, inefficient extension services and increasing competition from foodcrops, yields in most coffee growing areas in Burundi have tended to decline during past years. 1.02 Total green coffee production and exports have fluctuated greatly during the 1970s. On the basis of five-year averages, production of Arabica coffee increased by some 3.257 annually until the early seventies, but since 1973/74 has remained in the region of about 20,000 tons a year (Annexes I through III). While the increased Arabica production until 1973/74 was mainly the result of a steady area expansion, the subsequent stagnation of coffee production reflects two divergent trends in Burundi's coffee economy. In fact, available statistical information indicates that while coffee production in most coffee growing areas in the country declined, this was offset by the increase of coffee production achieved in the Ngozi Province. This is also true for the declining trend of Burundi's coffee quality (Annex I) which to some extent has been reversed by the successful introduction of coffee washing stations in the Ngozi Project area. In fact, Burundi's fully wet processed coffee is now fetching a 157 premium over its traditional semi-washed coffees in world markets. Thus, seen in the overall coffee economy of Burundi, the main positive effect of the Ngozi Coffee Improvement Projects has been to arrest a declining trend in Burundi's coffee sector during past years. 1.03 The two Ngozi Coffee Improvement Projects have been Government's main vehicle to improve the quantity and quality of Burundi coffee. The first project (Credit 147-Bu) covering about one-half of the Ngozi Province was signed in April 1969 and closed, fully disbursed, in June 1976. Progress achieved under the first Ngozi Coffee Project has been limited, due to the internal disturbances of 1972, shortages of both expatriate and national staff and transport difficulties in neighbouring countries resulting in frequent delays in delivery of essential construction and other materials. As a result, experience with improved coffee processing techniques had been limited  - 2 - when the Project was completed in 1976. However, a satisfactory start was made in improving husbandry methods and the use of modern inputs, although their impact in coffee production could not be assessed since no yield data had been systematically collected. Experience under the first Ngozi Project was essential in drawing up the Second Coffee Improvement Project, which was signed in December 1975. The second Ngozi Coffee Project will substantially be completed by mid-1981 and is expected 1:o be fully disbursed by April 1982. While the progress achieved under the second phase will be discussed in detail in the following chapters of the PCR, it should already be mentioned at this stage that similar to the PCR of the first: Ngozi Coffee Project, no conclusive statements about the Project's impact can be made because of a lack of reliable quantitative information. In fact, the Second Ngozi Coffee Project did not succeed in setting up a data collection system as systematic surveys of coffee plantations and production started in 1977 came to a halt because of lack of fuel and vehicles at the beginning of 1979. The mission, nonetheless, has tried to assess from all available sources, the impact of the Ngozi II Project on coffee production and quality. 1.04 Government continues to foster coffee development through the third phase of the Ngozi Project as well as other rural development projects in other coffee growing areas (i.e. Kirimiro and Puyinga) which are expected to be financed with World Bank assistance. Whether these efforts will enable Government to accelerate coffee production and improve coffee quality in the future has still to be seen. In fact, because of farmers rapidly declining terms of trade, growing population pressure, increasing scarcity of arable land, etc., it is foreseeable that future coffee imtprovement programs will have to be carried out in a considerably more difficult environment than during past years. Burundi's coffee development prospects will be discussed further in the concluding paragraphs of the PCR. 1.05 To date, Bank Group involvement in the financing of the agricultural sector has been limited (US$ 17.3 million). In fact, although IDA financed a fisheries project (Cr. 629-BU) in 1976 and a forestry project (Cr. 918-BU) in 1979, the two Ngozi Coffee Improvement Projects have been so far the Bank Group's main contribution to the financing of agricultural development in Burundi. However, with the recently negotiated Ngozi III and Kirimiro integrated rural development schemes, Bank lending to Burundi agriculture is expected to increase in the near future. 1.06 This PCR was prepared by RMEA Bank staff. It is based upon field visits to the Ngozi Project area, extensive discussions with OCIBU/Project and Minagric staff, and upon review of Project files and the preparation, appraisal, supervision and annual progress reports. The PCR and the Impact Evaluation Report of the first Ngozi Coffee Project were also consulted. In fact, since a complete distinction of the two Ngozi Coffee Improvement Projects is difficult to make, the findings of the PCR for the Ngozi I Coffee Project were often used to complete the analysis of the PCR for its second phase.  -3- 11. ID XIFICATIOI, PREPARATION AND APPRAISAL A. Identification and Preparation 2.01 The desirability of continuing the activities of the First Igozi Coffee Improvement Project arose during discussions between RMA supervision missions and OCIBU as well as Goverment staff. The first steps to prepare a second phase of the Project were taken during an RNEA mission to Burundi in August 1973. A Governmont preparation team was set up, including the Manager of the on-going project, and a preparation report was completed with the Bank in mid-1974. Govereant's project proposal was in line with its development objectives which emphasized the continued Improvement of coffee production in the country. Thus, the Project principally- aimed at increasing the quantity and improving the quality of coffee produced by about 110,000 smallholders living in the Igozi Province. It specifically included: (a) further distribution of inputs for coffee growing and strengthening of extension services in all coffee growing districts in the Igozi Province; (b) construction of 38 eoffee washing stations in addition to the four established under the first project 1/ to fully wet process all coffee produced in the area covered by Credit 147-BU; (c) expansion of coffee research facilities to assist Project Management in solving any agricultural and technical problems arising in the Project area; (d) the development of fooderops as a necessary prerequisite for further coffee development; (e) feeder road maintenance and rural water supplies; as wall as (f) an Evaluation and Monitoring Unit to systematically assess the progress of the Project and to prepare a long-term national Coffee Development Program. Total Project cost, over a five-year project Implementation period, was estimated at US$ 13.3 million, to be financed by IDA (857) and Government (157). The Project was expected to be tarried out by the Project Department established within OCIBU which had already been responsible for the 1mplem6ntation of the firet Project. Key management positions were initially to be filled by expatriates, until sufficient Burundiens had been trained to take over Project management. 2.02 Except for the importance of the coffee processing component, there was consensus between the Government and the Bank preparation teams concerning the objectives, scope and content of the second phase Qf the Ngozi Coffee Project. Initially Government had proposed to construct 56 coffee washing stations during the second phase of the Project. However, the Bank preparation team convinced Goverment that the construction of so many washing stations would be far beyond OCIBU/Project's Implementation capacity. Moreover, the construction of hulling facilities in Ngozi as well as the improvement of water supply for Ngozi town were considered as possible Project components. These, however, were not included in the final preparation report as the remaining programs for the second phase Project were already regarded as rather ambitious, in particular in the light of the experience during the first phase of the Project. The Government also agreed to review cherry pricing policy for coffee cherries delivered to washing stations, in order to provide incentives to farmers. In fact, the poor performance of washing stations under the first Ngozi Project was considered to be due to the fact that cherry prices were lower than the equivalent in parchment coffee. Hence, I/ The Butegene washing station was built prior to the start of the first Nlgozi Project, so that five washing stations were in operation by the end of the first phase in 1975.  -4- farmers preferred to sell their coffee in private markets rather than deliver their cherries to washing stations. 2.03 Overall, the feasibility report for the Second Egozi Coffee Improvement Project was well prepared, except for its fooderop development component, t"hich was rather sketchy and vague, probably reflecting the lack of any experience in fooderop development in Burundi at that time. However, it was fully recognized that, in order to increase the interest of farmers in future coffee development, considerable efforts had to be made to develop their subsistence crops at the same time. B. Appraisal 2.04 During the appraisal mission, in October 1974, the mission made only a few changes in Government's initial project proposals. The mission recomended reductions in fertilizer application and price subsidies, further preparation of rural development and non-coffee agricultural activities and, more significant, a review of coffee cherry pricing policy and the construction of fewer coffee washing stations then the 28 under consideration. The reasons were as follows: a. Fertilizers: Since no clear results had emerged from ISABU's trials with which to reliably forecast yield response to fertilizers, the mission proposed that fertilizer applications be limited to a maintenance dose, only to replace nitrogen removed in thLe crop and to maintain farmer interest in fertilizer applications. Only farmers showing improved standards of coffee husbandry would be eligible to receive fertilizers. Moreover, the mission did not agree with the Government's proposal to continue the free distribution of fertilizers and, therefore, recommended that in the future farmers should pay a reasonable price for fertilizers, although Government would temporarily subsidize fertilizers and other input costs; b. Rural Development, food-crops and Livestock: In view of the inadequate preparation of Government proposals for a number of experimental rural development activities (a small livestock component, training of cattle for draught purposes, installation of grain mills, cultivation of vegetables in the marshes and establishment of farmers cooperatives) the mission recomwended that IDA disbursements be subject to detailed proposals, acceptable to IDA; c. Coffee Washing Stations: The mission first reduced the number of coffee washing stations from 38 to 20 anti, upon closer examination, expressed doubt as to their economic and financial viability. The reasons for the appraisal mission's reluctance were complex but plausible. One reason was that since only four washing stations had been built - not without great difficulty - during the first phase of the Project, the mission felt that the construction and handling of 38 washing stations over a five-year period would have been unrealistically ambitious. Concerning the economic viability of washing stations, the mission emphasized to Government that detailed economic analysis had to be done in Washington before it could conclusively support the Government's proposal for additional washing stations. d. Economic Viability of Washing Stations. Data with which to evaluate the merit of washing stations was very limited and, moreover, their  - 5-. economic viability has been a central issue from the beginning of the First ngozi Project until no-*. The Bank agreed to finance the construction of four washing stations under the first Project, it was hoped that upon project completion sufficient data would be available to assess their profitability. However, when the Second. Ngozi Coffee Project was prepared and appraised, only one washing station had been operational for two seasons. The results of these two seasons (1973 and 1974) were rather disappointing since not more than 40-45 tons of parchment coffee (35% of the designed capacity) were produced during each season. The main reasons were a lack of cooperation from local administrators, and unattractive coffee cherry prices (para. 2.02) as well as the inexperience of Project staff in managing washing stations. Moreover, the construction costs of the stations turned out to be much higher than estimated at appraisal. Finally, the export premium obtained for fully washed coffee turned out to be lower than projected during appraisal; only 757S of the total fully washed coffee was classified as FWA, and the balance was classified as either OCIBU 3A or OCIBU 3B. 2.05 Doubts about the washing stations were confirmed when, during the detailed preparation of the appraisal report, the mission concluded that a 67 estimated economic rate of return of washing stations was too low to justify the large investments proposed. The mission's negative conclusion then became subject of debate among Bank staff about the methodology of evaluating costs and benefits of washing stations. Although the Bank preparation team was aware of the uncertainties concerning the justification of washing stations, it had taken a more optimistic view of their prospects, in particular after a number of tests had proven the favorable liquoring quality of fully-washed Burudian coffee. The Bank preparation team also emphasized the importance of the washing stations as future centers for extension and other rural activities, as well as the extreme poverty of the country which considerably limits its development possibilities. Eventually the appraisal mission's view prevailed and the Burundi Government was informed that, with great regret, the Bank was unable to finance the processing component of the Second Ngozi Coffee Project. 2.06 A post-appraisal mission was sent to Burundi in May 1975 to explain the Bank's reluctance to finance washing stations. After protracted discussions, the Bank proposed providing for the construction of four additional stations to allow Government to continue the on-going coffee processing scheme in the Ngozi Province. This compromise was reluctantly accepted by the Government. In fact, Government had attached great importance to a continuation of the washing station program as it felt that it was in Burundi's long term interest to invest in quality improvement schemes. It also argued that the impact of coffee washing stations is greater than can be expressed in strict economic or financial term, since they were expected to serve as focal points for a large number of rural development activities, such a demonstration of better growing piractices for both coffee and foodcrops as well as sales of farm inputs and cconsumer goods. Government's disappointment at the Bank's conclusion probably explains why six months later it requested the Kuwait Bank to finance 10 washing stations instead of the four proposed by the Bank and formally agreed upon during negotiations.  6 - 2.07 Project negotiations were held in Washington in October 1975, during which no further major changes were made to the Project. During the Annual Meeting in September 1975, the Kuwait Fund enquired about the possibility of co-financing a Bank project in Burundi. The Bank proposed to the Kuwait Fund that it participate in the Ngozi II Coffee Project and it was then agreed that it would finance the construction of the four additional coffee washing stations, the coffee research component, water supply and road construction. Total Project costs including contingencies, were estimated at US$ 7.5 million, compared to US$ 13.2 million in the original Government proposal, mainly reflecting the considerable reduction in the number of washing stations. The Project, which would be financed by Government (US$ 1.1 million), Kuwait Fund (US$ 1.2 million) and IDA (US$ 5.2 million), was approved by the Board in November 1975 and signed one month later in December 1975. While the Second Ngozi Coffee Project thus seemed to be ready for a normal start in early 1976, a major hitch occurred, when in January 1976 the Kuwait Fund and the Burundl Government agreed to change the cofinancing program without prior consultation with IDA. In fact, during discussions with a Kuwait Fund delegation in Bujumbura, the Burundi Government decided to seek financing for the Project's research component from the Belgian Government and to request the Kuwait Fund to use the available savings for the financing of six washing stations in addition to the four agreed upon earlier, to which the delegation agreed. These changes differed substantially from prior agreements reached between the Burundi Government, the Kuwait Fund and IDA. Although the Bank requested clarification from the Burundi Government it did not apparently follow up on its request and tacitly accepted the agreement reached. Discussions, however, were held with the Belgian Government in Brussels to assure adequate financing and staffing for the Project's coffee research component. As the Credit Agreement had already been signed, neither the Credit Agreement (Cr. 593-BU) nor the appraisal report were amended to reflect these modifications to the Project. C. Project Objectives and Description 2.10 As described in the Appraisal Report, the Project objectives were to support the development of smallholder coffee and foodcrop production in Ngozi Province. It would also address a number of fundamental problems concerning the cultivation and processing of coffee, the expansion of food producticn, and the organization of extension services, and thus contribute to the future development of agricultural production in the country as a whole. Specifically, the Project would provide for: (a) inputs and strengthened extension services to sustain and extend the improvement of coffee cultivation started under the first Project; (b) construction of four additional coffee washing stations and technical assistance for the operation of the four existing stations; (c) staff and equipment for an expanded coffee research program; (d) inputs, field trials and extension advice for the cultivation of foodcrops;  -7- (e) construction of repair of potable water sources, bridges and culverts; (f) experimental rural development activities, including the improvement of small livestock, training of draught animals, and the establishment of farmer cooperatives; and (g) a resident managemenit team backed up by a small team of internationally recruited advisors, which would at regular intervals assist in formulating work programs and advise on specific Project problems. Apart from the coffee research program, which would be implemented by ISABU, the Project would be carried out by OCIBU. III. IMPLEMENTATION A. Start-Up of the Project 3.01 The second phase of the Nlgozi Coffee Improvement Project was originally expected to become effective in March 1976. However, due to delays in the finalization of the Kuwait Fund co-financing agreement, it did not become effective until October 1976. Project financing was arranged by the Government, not without some difficulty, to bridge the transitional period before the actual start of the second phase of the Project. Project start-up was adversely affected by frequent: changes in management, caused initially by the death of the Project Manager and the promotion of his local successor to Vice Minister of Agriculture during a change in Government in November 1976. After consultation with IDA, a newly recruited expatriate economist was appointed Project Manager and the former local Chief Extension Officer as Deputy Project Manager. The team of expatriate project officers were all in post by the effective commencement of the Project with the exception of the Financial Controller, who was recruited in February 1977. At start-up the Project absorbed all of the provincial technical extension staff of the Ministry of Agriculture, including former extension staff for the national anti-erosion control and reafforestation programs. 3.02 Coffee extension work started without major difficulties, although the performance of some of the Minagric extension staff absorbed by the Project appeared to be unsatisfactory (para 4.04). By the end of 1976, there were also indications that the new Government wanted to stop all fertilizer imports (para 3.14). Both the performance of extension staff and the use of fertilizers remained problems throughout the Project implementation period. The fooderop development program also made an encouraging start. Two seed multiplication stations and 180 demonstration plots were set up in the region. Small quantities of fertilizer were sold to selected farmers, but soon it was found that follow-up extension services were lacking. 3.03 All five coffee washing stations established in the first phase worked at full capacity during 1976 and produced 640 tons of parchment coffee (about 8.37 of the total parchment coffee produced in the Project area). Unfortunately, in the middle of 1976, it became apparent that the three most  -8- recently built washing stations were in need of major repairs, if not total reconstruction. This situation was reviewed by an IDA mission in light of the prepared construction program. It concluded that effective repairs could be carried out in time for the 1977 session and that the difficulty should not affect the construction of new washing stations. However, a critical shortage of cement from late 1976 stopped all construction activities in the Project for three months, a problem which continued to cause delays in Project implementation. B. Coffee Production Activities and Achievements Introduction 3.04 As set out at appraisal, the major coffee production objectives of the second Ngozi Project were to be achieved inter alia by improving farmers' husbandry methods and input use as follows: (a) better pest control through the use of newer insecticides and improved methods of application; (b) introduction of mulching to-improve the coffee soils and control erosion; (c) 'systematic pruning of mature coffee trees through use of appropriate pruning saws and shears according to recommendations by the Project extension service; (d) infilling and replacement of very old or dead coffee trees using seedlings provided by the Projects coffee nurseries; and (e) improving yields of coffee by the application of fertilizers, mainly in the high potential areas of the Project. An estimated 110,000 farmers were expected to benefit from the coffee development program at full development (Year 7) when incremental production would be about 1,000 tons of parchment coffee. Pest Control 3.05 The most serious coffee prest in Burundi at the start-up of the second phase of the Coffee Improvement Project was the antestia beetle, which had defied all attempts at control measures during the first phase of the Project. Antestia feeds preferentially on the large, green, expanding stage of coffee berry when the contents of the berry are still soft. The damage therefore caused to the coffee berry results in crop loss and is due to either "premature berry drop" (with a heavy infestation of antestia, this loss can be as high as 307. of the potential total crop); or, if the coffee tree retains the damaged berry until normal harvesting, it produces very low quality coffee beans (can be as high as 107 of the harvested crop). These combined crop losses can occur with population levels of antestia as low as eight to ten per tree and crop losses of up to 107 can occur with populations as low as three antestia per tree.  -9- 3.06 Attempts to control the antestia population during the first phase of the Project by dusting with DDT/Hlalathion were partially successful in that the antestia populations within the Project area were reduced below those outside the Project area. However, the residual level of antestia in the Project area remained above that required for effective control, is normally less than one beetle per tree. No. of Antesti.a per Tree - Pyrethrum Test Jan Sept Jan Har Sept Mar '72 '72 '73 '74 '74 '75 Project Area 3.5 2.5 3.1 3.5 1.5 2.7 Outside Project 7.5 9.6 7.1 6.2 4.1 6.1 Source: OCIBU/ISABU. See Annex Iv. 3.07 The deficiencies of these early campaigns were thought to be due in part to the creation of a DDT-resistant antestia population, and in part to the very inefficient method farmers used for application, viz. shaking a dusting bag. Entomological studies conducted by ISABU to find more effective insecticides conclusively revealed that the antestia beetle had not developed any marked resistance to DDT, and that the principal deficiency lay with the method of application, or an inferred absence of any insecticide application. These same studies also clearly showed that the organo-phosphorus insecticides Fenitrothion and Fenthion and the synthetic Pyrethrins offered a more efficient control system without the long term persistence problems associated with DDT. In 1978 field tests were carried out by ISABU, using Fenitrothion, and conclusively verified the earlier laboratory trials. A technical recommendation that these new insecticide formulations should be employed was prepared for Project management in 1978. 3.08 Initial full-scale anti-pest campaigns in 1978/79 and 1979/80 were not particularly encouraging, primarily because they were badly managed and beset by technical problems caused by the newly introduced 3,000 hand-sprayers. Unlike the previiously used dusting bags these devices suffered from mechanical problems and also required a complete re-orientation of the extension services program for implementing the insecticidal dust applications. Pyrethrum knock-down tests for 1978 and 1979 reflect these problems by showing that the campaigns had little impact upon the antestia beetle population. In contrast, the anti-pest control measures in 1980 were carried out by Project management in an orderly and timely fashion. The deficient "Tropic" hand sprayers were either replaced or repaired and the more efficient insecticide Fenthion was used in all districts of the Project. Pyrethrum tests carried out in November 1980 indicate that the antestia beetle has almost disappeared from the Project area. Any relaxation of the anti-pest control in the future, however, would result in a resurgence of the antestia beetle population and subsequent crop losses.  - 10 - Pvrethrum tests for 1978, 1979 and 1980 Number of Antestia per tree Agronomic October November November Sector 1978 1979 1980 Ngozi 2 3.1 0.2 Ruhororo - - 0.1 Mwumba 1.7 1.2 0.07 Gashikanwa 1.4 1.1 0.0 Tangara 1.8 7.6 0.6 Kiremba 2.0 1.3 0.03 Marangara 1.7 1.5 0.0 Kayanza 1.3 1.4 0.0 Jene 0.8 2.0 0.1 Matongo 4.2 4.4 0.04 Gatara 1.9 3.1 0.01 Gahombo 1.8 1.3 0.3 Banga - - 0.3 Busiga 2.1 1.3 0.1 Rango _ 0.04 Project Average 1.9 2.4 0.1 Mulching 3.09 Mulching of coffee is important for improving the texture and humus content of the soil, providing small amounts of balanced nutrients, to control both weed growth and soil erosion and to conserve moisture in the soil during dry weather. The first phase of the Project was clearly successful in encouraging better mulching standards and, during the early stages of the second phase of the Project, improvement in mulching standards continued. Banana leaves, maize and bean stover provide the major supply of mulch, but in many areas of the Project the supply of sufficient mulching material has become a serious problem. It is estimated that approximately 70. of the coffee plantations are now satisfactorily mulched compared to 40. before the second phase of the Project started. With growing population pressure and deforestation, crop residues are increasingly being used as a source of fuel and this has resulted in a growing shortage of organic material available for mulching and compost. Although project management continues to distribute trysacum and slips, farmers appear to be reluctant to use these as mulching material and further efforts will have to be made by extension staff to stress the importance of proper mulching to farmers. The Project's research unit continues to actively research for alternatives either for direct mulching material or for cover crops which do not compete with the nutritional and water requirements of the coffee. Pruning 3.10 In the first phase of the Project, approximately 807 of the pre-Project estimated tree population of seven million trees were brought into an average five-year pruning cycle during the last four years of the Project. A similar approach to pruning can be inferred to have occurred during the second phase (see Annex XII). Distribution of pruning saws and shears would  - 11 - appear to have been much more limited than in the first phase since, in 1979/80, only 10,000 pruning saws and 5,000 shears out of a total of 30,000 shears purchased by the Project were sold to the 150,000 farm families living in the Project area. The farmers reluctance to purchase these tools would appear to be related to the current Project sale price of FBu 200 compared to the heavily subsidized first phase price of FBu 50 per pair as well as their preference for using a conventional saw rather than shears. Use of Fertilizers 3.11 Coffee Fertilizer Trials. Based on fertilizer trials in farmers' fields in Ngozi Province, conducted under ISABU supervision from 1962-66, the 1968 appraisal mission for the first Project had concluded the importance of good husbandry practices and of the application of nitrogenous fertilizers for increasing yields of parchment coffee. Only a small effect of potash was observed in the 1962-66 trials and there were no indications that phosphate applications contributed to the yields. The 1974 appraisal mission for the second Project was less convinced by the results of the trials conducted in the 1960s and observed, after re-examining the data, that yields varied considerably both from year to year and in different locations, in respect of fertilizer application. Extraneous factors (location, altitude, type of soils, the standard of maintenance of the plantation, the stage in the pruning cycle and weather conditions) made it impossible to attach statistical significance to the trial results. This was confirmed by a second series of trials, conducted under the first phase of the Project from 1970-75, which revealed much larger and unexplained differences between the yields of individual plots which received the same treatment. 3.12 The Ngozi II appraisal mission concluded that the poor and uncertain results of fertilizer trials were due to: (a) faulty experimental design, i.e. the absence of control plots and application of fertilizer formulations; and (b) the implementation of the trials. The wide dispersal of the trial plots made it necessary to delegate! the supervision of fertilizer applications to inexperienced extension staff and there was no way of verifying that the prescribed types and doses of fertilizer were always applied at the right time to the appropriate trees in the prescribed manner. This system almost certainly jeopardized the systematic collection of the harvested coffee crop from the trial plots. Moreover, farmers' tendency to change other agronomic variables within their coffee plots probably influenced the yield levels more than the simple fertilizer variable. Because there were good reasons to doubt whether farm level trials, even when properly designed and consistently supervised, would produce definite answers as long as there was no guarantee that- the level of maintenance (mulching, pruning, insect control, etc.) remained consistent throughout the trial period, the appraisal mission suggested an ISABU research program geared specifically towards solving these problems. Distribution of Fertilizers 3.13 In spite of the lack of definite proof as to the contribution of fertilizers to coffee yields, Project staff - in accordance with ISABU recommendations - provided for the use of some fertilizers in order to compensate for the nitrogen deficiency in the Ngozi soils. However, fertilizer applications were to be restricted to about one half of the coffee stands in the high potential areas and distribution was limited to only  - 12 - farmers with properly maintained coffee plantations. Farmers and Government were expected to finance the recurrent fertiLizer cost, whereby Government's subsidy would gradually decline to 257 at the end of the Project period. Farmers' contributions and Governrent's subsidy would be placed in a revolving fund account to be managed by OCIBU. 3.14 At the beginning of 1977 the Project: received 500 tons of urea, part of which was used in the new Project areas anid distributed to farmers free of charge contrary to IDA recoimendations. Project management argued that during the first Project, fertilizer was distributed free of charge to farmers for the first three years as the fertilizer effects on coffee trees would only be apparent after a three year period. However, discussion on this issue was not pursued since, in February 1977, the Government stopped all fertilizer imports on the grounds that the use of chemical fert:ilizers was both uneconomical and ineffective. Government further argued that extension services in all agricultural projects should focus on improving husbandry methods and the production of compest, before using any chemical fertilizers. OCIBU, strongly supported by ISABU and Bank staff, requested Government to make an exception to the fertilizer ban for the Ngozi region, as husbandry methods had signficantly been improved since the beginning of the first Ngozi Coffee Project and that fertilizer experiments in the Project area would provide valuable information for coffee and foodcrops in other parts of the country. The debate ended only when, in 1979, an IFCC (Institut Francais du Cafe et du Cacao) mission to Burundi strongly recommended the use of urea on coffee plantations as a means of stopping rapid soiLd deterioration and, in particular, depletion of key nutrients, which had become evident. 3.15 After almost three years of uncertainty and hesitation, Government reversed its position and in mid-1979 officially authorized OCIBU/the Project to import 900 tons of urea. The first distribution of fertilizers took place in November 1980 for the beginning of the 1980/81 coffee season, during which 84 tons of urea were sold to about 12,000 farmers, representing an average of 7 kg per farmer or less than half the normal requirements for an average plantation of 173 trees. l/ Compared to initial expectations and projections, farmers' response to the fertilizer distribution program was d,isappointing. In fact, only about 8. of all farmers in the Project area purchased fertilizers in November 1980. This was all the more astonishing as fertilizers had been distributed - although with interruptions - in Ngozi Province since the beginning of the Project In 1969, and farmers were thus presumed to know the positive effects of fertilizer applications on their coffee plantations. 3.16 The apparent lack of interest by fairmers in applying fertilizers was explained as follows. First, according to Project management, the limited sale of fertilizers was partly due to the lack of motiviation of the Project's extension supervisors who were responsible for promoting and distributing fertilizers in their respective districts. More than half of the 84 tons distributed was sold in three districts where extension services were run by the more dynamic Project extension supervisors. Second, since fertilizer applications were recommended in November and March, Project management attributed the low fertilizer sales in November 1980 to lack of cash by farmers. Since this was three months after the end of the coffee harvest season, the farmers' cash situation was expected to be worse ln March 1981, when the second application was scheduled. 1/ An application of 100 g/tree was recommsended twice a year.  - 13 - 3.17 Project rmtanagement's explanation for the farmers' difficulty in paying cash for fertilizers was puzzling since, when the distribution and financing of inputs was discussed between IDA and OCIBU/Project Management, the latter had decided that all inputs wouid be distributed only for advance cash payment. This was in viewv of difficulties with credit sales of fertilizer during the first Project and the lack of Project staff and administrative facilities to handle credit recovery efficiently, since Burundi has no agricultural credit institution and no financial intermediaries such s cooperatives capable of handling input credit operations. Project management had, therefore, nroposed that farmers be given the opportunity to pay in advance for fertilizers at the end of the coffee season (July/August), for delivery the following November and March. Project management also decided to increase the use of washing statiorLs as centers for the sale and distribution of inputs. This seemed to be a reesonable decision, as farmer records are well kept at these stations. 3.18 So far only one year's experience has been gained from the new fertilizer distribution and financing system. At appraisal, it had been estimated that 600 tons of fertilizer would be used during the last year of the Project. IL 1980/81 only about 200 tons were actually distributed to farmers, but in 1981/82, volume sold under the system of advance payment for deliverv in November 1981 and Harch 1982 rose to 620 tons. 3.19 The apparent reluctance of farmers to purchase clearly subsidized fertilizers in the past is not necessarily indicative of their propensity to do so in the future if coffee yielct increases can be clearly demonstrated. With Government subsidies, farmers are presently paying only 50' of fertilizer costs and, with assured deliv;ery such as in 1981, they are indeed willing to pay the extra cost (FBu 850 per farmer on the average) to generate an additional FBu 3,000 - 5,0OO in gross income. 1/ If the fertilizer subsidies are phased out (gradually) as plannied, farmers will eventually have to bear their full cost, i.e. an average of FBu 1,900 each at present very high fertilizer prices (urea costs the equivalent of US$ 500/ton). Without subsidy, the lower increment in gross revenue (FBu 1,100) makes it relatively unattractive for farmers to risk using fertilizer, unless yield increases of 250 grams of parchment coffee per tree can be achieved. If this can be achieved farmers can then pay the full price of fertilizers, based on the conventional assumption that farmers will apply inputs in Burundi if the benefits are at least twice the costs of the imports. 1/ Without Government's subsidy the average farmer's contribution would amount to about FBu 1,900 (200g/tree x 173 trees - 34.6 kg x FBu 55/kg - FBu 1903). According to fertilizer trials; in Ngozi and other coffee growing regions (e.g. Kenya), the incremental yields from fertilizer use are expected to vary from 150 to 250 of parchment coffee per tree. In monetary terms this would mean that farmers will have to spend about FBu 1,900 to earn an additional income of between FBu 2,984 and FBu 4,973 (173 trees x 150 g/tree = 2,595 kg x 115 FBu = FBu 2,984, and 173 trees x 250 g/tree = 43.25 kg x 115 FBu/kg = FBu 4,973 respectively.  - 14 - 3.20 It must be concluded that the efficient use and distribution of fertilizers for coffee production have by no means been achieved by the end of the Igozi II Project. While ISABU's ongoing research seeks a technically and economically feasible fertilizer formula for increasing coffee yields - preferably a balanced compound fertilizer - technical experlts agree that a minisis amount of nitrogen must be applied to coffee in areas -uch as Ngozi. Economically, however, the case for applying urea at the equivalent price of US$500/ton is not yet convincing, unlecs average yields of parchment coffee on farmers' plots can be increased by at least 66bkg/ha (at present coffee and fertilizer prices). This would represent an approximate break-even point in terms of foreign exchange at prevent coffee and fertilizer prices. Taking into account Government subsidies of fertilizer costs (about US$100/ha) the tax revenue needed to at least offset the subsidy cost would require yield increases of roughly 130 kg/ha (at the 1980-81 rates of export taxes and duties indicated in the Ngozi III Appraisal Reporit). 1/ 3.21 The economic implications of fertilizer use ar>e sulbsidies will continue to be central issues in the success of coffee and foodcrop improvement scheses in Burundi, as they affect to farmers incentives and real income, Government revenues, and foreign exchange costs and benefits. With the deterioration in the price relationship between coffee and fertilizer since the appraisal of the Project on the one hand, and the overall effects of inflation on farmers' real income on the other, some justification exists to maintain a level of fertilizer subsidies in order to at leaat keep coffee yields from falling. The dilemma, however, is that the Government may not be able to continue its present subsidy of fertilizer -- estimated at the equivalent of US$150,000 for Ngozi II alone in 1981/82 -- and other imports for coffee products as coffee prices decline in real terms. The Government's share of total coffee export earnings has considerably declined in recent years and may decline further in future years as world markit prices are expected to continue to fall, and the Burundi Coffee Stabilization fund had no more funds in 1981 to support the coffee price under the sliding scale pricing mechanism (para. 5.14). New Coffee Plantings 3.22 In 1981, it is estimated that the number of coffee trees has reached 28.7 million, or an increase of about 5.5 million trees since 1976. At full development from 1983 onward, there would be 29.6 million trees (Annex III). Both the total and the increase in the numwber of trees are significantly greater than the appraisal estimates because: (a) the coffee tree population, 17.6 million trees in the Project area in 1976 given in the appraisal report, was re-estimated to have been 22.2 million trees through an auditead tree count done by Project Management in 1978/79; and (b) plantings of seedlings needed to replace low-yielding varieties, or as infilling for old and diseased trees, and for new coffee plot extensions, exceeded the rate of 700,000 seedlings per year estimated in the appraisal report as shown below. I/ Because of the uncertain lag between the applLcation of fertilizers and the achievement of yield increases, financial and economic income streams should be discounted, thereby further increasing the minimum desired yield increase above these levels.  - 15 - Seedlings Planted for Extension and ReDlacement Year 1976 1977 1978 1979 1980 1981 Purpose -'000 seedlings - Extension 872 706 889 871 833 833 Infill and Replacement 705 871 216 65 65 65 Total 1.577 1.577 1.105 936 898 898 3.23 The initial difference of 5.7 million trees between the appraisal estimates (17.6 million) and the 1978/79 tree count (23.2 million) is explained by the unreliable data base avilable to the'appraisal mission. According to 1978 OCIBU estimates, it was estimated by the appraisal mission that Ngozi Province had about 35S of Burundi's 50 million coffee trees. The only fairly reliable figure available to the mission in 1975, however, was 8.3 million trees in the area of the first project, which was thought to represent roughly one half of the coffee trees in Ngozi Province. No other data were available for tree planting for replacement and expansion in the Province as a whole. The revised figure of 23.2 million trees for 1976 was obtained, as whown in Annex XII, by matching the tree audit data shown below, with working data on transplanted seedlings during 1976/78. Category Number of Trees V/ 1978/79 Trees 1 year old 196,760 2 year old 791,939 3 year old 1,639,785 Nature 18,666,605 Pruned 1978 2,599,979 Pruned 1979 3,172.147 Sub-total (1) 27,067,215 Seedlings Extension 870,990 Infills, Replacement 64,910 Sub-total (2) 935,900 TOTAL (1 - 2) 28,033,115 1/ The difference in the total number of trees between the tree audit data and the 1979 Annex XII data is insignificant for reasons given in Annex XII. The differences pertain to the age distribution of trees. Although it was an important milestone in the Project, the tree audit was static count requiring over 12 months to complete, which did not take into account the transplanting of seedlings and the aging of trees in each category by one year during the audit period. The reconciliation is explained in detail in Annex XII.  - 16 3.24 As a result of the increased rate of planting and seedling production since 1976 (4% annually vs 2% during Phase I), and extrapolation of the 1978/79 tree audit by age and trees pruned in 1978 and 1979, it is estimated that the final coffee tree population at full developing beginning in 1983 would be about 29.6 million trees. This would be a net increase of 6.4 million treeo as a result of the second phase of this Project, with concommitant increases in coffee production (para. 5.04). C. Research 3.25 At appraisal it was agreed that the project would support a strengthened ISABU coffee research program which would focus on insect and disease problems and on husbandry techniques applicable to smallholder farmers, including further trials with chemical fertilizers. Unfortunately, the start of the research component was delayed until the latter part of 1978, by protracted negotiations between the Burundi and Belgian Governments concerning the funding of the various components of the research program. As previously mentioned (para. 2.07) the Kuwait Fund was initially expected to finance the Project's research program. However, the Burundi Government then decided to seek financing from the Belgian Government and use the Kuwait Fund resources to finance the additional washing stationss. 3.26 Implementation finally started when it was agreed that the Belgian Government would finance the costs of two expatriate research experts -- one agronomist and one entomologist/phytopathologist -- while other investment costs as well as operating costs, including the salaries for local staff, would be funded from the OCIBU budget. The financial involvement of OCIBU enabled it to participate with ISABU in formulating a detailed five-year coffee research program which would be reviewed on an annual basis. Thus, OCIBU retained some control over coffee research activities within the Project area. This control was important, since ISABU management would have preferred this program to have a more national scope whereas OCIBU insisted on concentrating research efforts on specific problems in the Project area. Most Project research work is presently carried out at Kayanza, where a coffee research sub-station operates within the agronomic zone of the Project. Experts believe that whatever research results will be obtained in the Project area will, to a large extent, also be applicable to other coffee growing areas in Burundi. 3.27 The late start-up of the coffee research component of the Project meant that, within the time remaining for completion of the second phase, priority was assigned to research on urgent problems requiring immediate attention. Three of these were: (a) the on-going entomological problem of antestia control; (b) the agronomic problems of fertilizer response trials; and (c) the availability of sufficient quantities of suitable mulching materials.  - 17 - 3.28 As set out at appraisal the entomological trials involved pyrethrum knockdown tests on plots throughout. Ngozi province to ascertain the identity and population incidence of insects found in the immediate vicinity of the Project. It was confirmed that antestia was the major insect pest and that other coffee pests were present in only insignificant proportions. Control tests, using caged trees, and a range of potential insecticides clearly demonstrated that the antestia beetle had not developed any marked resistance to DDT, but the organo phosphorus insecticides Fenitrothion and Fenthion and a range of synthetic pyrethroids gave a more rapid initial kill potential and a longer effective control than the previously applied DDT/Malathion dusting material. Field trials fully vindicated the optimism shown in the laboratory tests and full scale trials using E'enitrothion were initiated in 1978/79. The initial tests were disappointing, mainly due to bad management and wrong timing, but full scale field trials in 1980 with Fenthion appear to have solved some of the problems of formtulation of the insecticide dust and the method of application. Antestia appears to have been virtually eliminated from the Project zone (para 3.06 - 3.08). 3.29 The late acquisition of the Kanyanza sub-station land (1979) meant that the planting of the fertilizer trial plots was delayed until 1979/80 and results from these carefully designed trial plots are not expected to give any definitive answers until the latter half of the 1983. The ISABU coffee research plots at Kayanza, Gisha and Mwumba included replicates of variety trials (approximately 23 different arabica hybrids) provided by Rwanda, Ethiopia, Kenya and Tanzania and a number of trials on potential mulching materials and cover crops. Optimum planting density trials are being carried out at the larger land area provided by the Kanyanza research station, where tree densities from 1,000 to 5,000 trees/ha are being studied under standardized conditions of fertilizer application, mulching, pruning and insect control. Results may be expected from these density trials during the latter part of 1983. 3.30 In conclusion, even though the results so far obtained from the Project's coffee research program are less than expected at appraisal, they are very promising in view of the short duration period of full-scale implementation. It is hoped that the progress will continue now that ISABU has access to its own coffee research plots and that neither farmers nor the extension staff of the Project can be blamed for subsequent research management deficiencies. Many of the outstanding research components identified at appraisal can now be studied, not least of which is the long overdue pedological survey of the Project area. However, one major cautionary note should be sounded, namely the question of the dwarf arabica varieties being tested at the ISABU research plots - Caturra and Catuai. While these dwarf varieties give a potentially very high density of planting (5,000 to 10,000 per hectare), and hence a potential high yield of crop per hectare, they have poor quality characteristics and are unlikely to be suitable for the Project's fully wet processing program. The Project's coffee research program should place equal emphasis on both quality and quantity in all of its fertilizer, variety and planting density trials.  - 1 8 - D. Coffee Processing 3.31 After a shaky start under Ngozi 1 and additional management difficulties -during the first three years oef Ngozi II, the w-rshing stations have finally fulfilled their promise of efficiently producing nigh quality fully weashed coffee. The targets forecast at appraisal have Seen achieved; throughput and yields of the fifteen workin,g stations reachnd optimum production efficiency in 1981. 3.32 The four coffee washing stations built under the Ngozi I project and the four new ones agreed during Ngozi TI appraisal and negot:iations, were specifically intended to increase Burundi's coffee export earnings based mainly on the quality old premium, price paid for fully-washed arabica coffee without necessarily increasing volume. The premium price ad&antages of fully washed (FWA) over traditionally processed semi-washed c.- fee in world markets had been proven elsewhere (e.g. Kenya) but, for lack of experience in Burundi, only four stations were included as a trial in Ngozi I. Failure to fully attain the benefits during Ngozi I was also the reason for the Bank's initial reluctance to finance a major expansion of the washing station in Ngozi II. Consequently, it was proposed by the Bank that, under Ngozi II, additional technical asslstance would be provided for the operation of the four original washing stations and would support only the construction of four more stations as an extended trial, in order to obtain more reliable lnformation on the construction and operating costs of washing stations, as well as the price differential to be obtained for FWA coffee in world markets, 3.33 Since the Kuwait Fund agreed to finance the construction of ten new washing stations instead of the four proposed by the Bank, it effectively terminated the experimental nature of the processing component proposed by the Bank. Subsequent results showed that this decision to terminnate the experimental nature of the Project's processing component was sound. 3.34 Construction of the first washing station (Mushonge' started in December 1971 (two years after the start of the Project) but, because of late arrival of equipment, the station did not 'become fully operational until the 1973 season, and had disappointing results during the 1973 and 1974 seasons when only 40-45 tons of parchment coffee were processed, An evaluation of Mushonge was done in 1973. This delayed construction of the remaining three washing stations until October 1973, but they were finally fully operational for the 1975 season, a full six years after the commencement of the Project's first phase. By the end of the 1976 processing season it was apparent that three out of the four new washing stations suffered from faulty design and construction. While a major portion of the defects were attributed to the local contractor who built the washing stations, a share of the blame was placed on the Project's processing engineer who had not adequately supervised construction. Major repairs and even reconstruction were proposed for these three washing stations but, so far, these have functioned throughout the second phase of the Project. by simply making minor repairs before each processing season.  - 19 - 3.35 The problems encountered by relying on a local contractor to build under Ngozi I were avoided in Ngozi II by utilizing the Project's own construction group to build the ten new washing stations under force account. This arrangement has so far proved to be very satisfactory and, despite the recurrent shortages of cement and other essential building materials, the ten washing stations were to be fully operational at the end of the 1981 season. 3.36 A major problem arose for the management of the washing stations with the departure of the expatriate processing engineer in mid-1977. it appeared at a time as if the whole of the washing station component of the Project would come to a half. However, with the arrival of a new expatriate processing engineer at the end of 1L978, and with the assistance of Project management, the trend was decisively reversed. The washing stations are now well managed and a systematic training program for washing stations personnel has been established. After a week of formal training at Project headquarters on all aspects of running washing stations, candidates are appointed on a probationary basis as deputy washing station managers for one coffee season to assess their capabilities. This system of on-the-job training has proved satisfactory and regular supervision by the processing engineer and his deputy has ensured that technical and organizational management of washing stations is sound, and that fraud is curbed. Farmers' Use of Washing Stations 3.37 A crucial factor in the successful operation of the washing stations has been the farmers' willingness to deliver sufficient volume of coffee cherries to the washing stations to enable them to work at an economically justifiable throughput capacity. During the early stages of the first phase of the Project, farmers were not motivated to supply sufficient amounts of coffee because of the low prices paid for the cherry and the delay in payments, while the local traders -- despite weight cheatlng -- paid farmers a slightly higher pro-rata rate for home processed semi-washed parchment coffee and, moreover, paid immediately. OCIBU's pricing policy was rectified, during the latter half of Ngozi I and payments to farmers were made more quickly by the washing stations to limit the appeal of sales to local traders. 3.38 Farmers now have two major incentives to provide coffee cherry to the washing station. First, they have a financial incentive because the price paid for coffee cherries based upon the technical coefficient of conversion to parchment coffee obtained by the washing station during the previous harvesting season. In 1979/80, the technical coefficient for washing stations was 4.9 kgs cherry per lkg of coffee parchment. Since it is estimated that about 5.5 to 6 kgs of coffee cherries are required to produce 1 kg of parchment coffee by the traditional semi-washed process, farmers have an approximately 171 higher financial incentive to provide cherry to the washing stations, despite the much stricter control on acceptability of cherry quality. 1/ Secondly, farmers are eager to sell their coffee cherry to the washing stations, despite the strict grading, as this enables them to save time and labor on coffee pulping and drying during the main harvesting season, and to utilize this time on production and harvesting of food crops. No pricing system has yet been evolved to give the farmer the added price bonus 1/ Only one standard of cherry quality has been accepted since the 1979 season; all lower qualities are rejected.  - 20 - obtained for the fully washed coffees over the semi-washed. If this were possible, the farmer would have a third incentive for providing cherry to the washing stations and, moreover, would have an even greater incentive to concentrate on quality aspects of production which play no part in the traditional marketing system for the semi-washed coffees. Washing Station Production 3.39 The low annual production of parchment coffee noted for the Ngozi I washing stations was not, as assumed at appraisal. Of Ngozi II, due to a low design capacity but caused simply by a lack of initial cherry inputs. The standard design of washing stations constructed in Ngozi I and II is capable of an indefinite throughput of 20,000 kg of coffee cherry per day. At a conversion coefficient of 4.9 kg of cherry to 1 kg of parchment coffee, the design capazity of these washing stations is for 4.0 tons of parchment coffee per day with eight hours being allowed for pulping. With good management, the design capacity allows for approximately a 50% increase in daily cherry intake (12 hours pulping) and hence a maximum daily production of 6.0 tons of parchment coffee. With arabica coffees, it is normally assumed that 90. of the total crop will be harvested within a 50 day period in the area surrounding a single washing station. Hence, with virtually no strain on management, the washing stations should each be able to process an average 200 tons parchment coffee per annum (50 days x 4.0 tons - 200 tons parchment coffee), while absolute maximum production, given good drying conditions throughout the harvesting seasons, will be between 250 to 300 tons of parchment coffee. 3.40 The validity of these assumptions is demonstrated by the parchment production figures for all fiftenn l/ of the washing stations within the Project area; the supposed design capacity of 125 tons parchment coffee calculated at appraisal has been exceeded many times, especially in 1981 when the overall average exceeded 200 tons. 3.41 With the exception of the pre-project washing station of Butegana (constructed around 1952), all of the Phase two washing stations have increased their throughput production to within 80% of the optimum 200 tons parchment coffee capacity within three years of commissioning. Fluctuations from year to year are an indication of climatically induced cherry Input fluctuation to the washing stations and are a reflection of variations in the total coffee produced in Ngozi Province, not of deviations in the standard of washing station management. 1/ Including Butegana.  -21- Parchmnt Coffee Production Statistics: 1973 - 1981 (tons) Washinag Stations 1973 1974 1975 1976 1977 1978 1979 1980 1981 1. Butegana 212.0 170.5 118.1 62.5 3.6 8.9 70.6 72.6 73.2 2. Wushonge 46.8 40.7 141.4 177.4 136.4 160.2 178.1 102.8 257.0 3. Burenza - 0.5 110.3 147.7 151.9 136.0 169.8 113.9 213.5 4. Kagoma - - 88.4 120.3 144.0 133.6 174.2 119.2 247.7 S. tuba&& - - 75.4 102.3 89.8 99.4 214.2 154.0 266.2 6. Rugonera - - - - 73.4 113.0 258.0 98.3 332.5 . Rukurazo - - - - 4.0 117.2 232.6 127.8 279.0 8. Murabi -- - - 45.3 38.1 85.9 91.6 341.6 9. Rugori - - - - - - 153.8 165.5 401.4 L0 :kA - - - - - - 82.4 107.2 270.2 ' 1. iilram - - - - - - 15.6 61.6 182.1 .12 . 2ubuga - - - - - - - 128.1 266.4 13. Kibuye - - 123.6 14. lirango - - - - - - - - 232.8 15. Karinzi - - - - - - - - 101.5 TOTAL 258.8 211.7 533.6 610.5 653.4 806.4 1,635.2 1,339.8 3,588.8 Perf'ormance Indicators 1973 1974 1975 1976 1977 1978 1979 1980 1981 Total production (tons parchment) 259 212 534 610 653 806 19635 1,339 3,589 Yields (cherry/ parchment in Kg) 5.08 5.88 5.16 5.32 5.18 5.18 4.89 4.86 4.56 Yields (cherry/ green coffee in Kg) 7.38 6.61 7.56 6.76 6.52 6.51 6.05 6.16 5.7 Percentage A3 parchment 1/ NA NA NA NA. 18.8 19.0 10.8 4.6 4.7 Percentage mill-, ing losses 21.16 21.52 32.55 21.41 20.61 20.63 19.25 21.11 19.3 Percentage AA/A Total FW.A 86.50 93.08 73.60 56.29 78.44 73.00 94.10 90.90 85.0 1/ Floating coffee beans.  - 22 - 3.42 The Table indicates that average yields of 1 kg of parchment began to improve during the second phase of the Project. A similar improvement is noted in the yields of green coffee from a given weight of cherry, with a gradual decrease in the percentage of milling losses when converting parchment to clean coffee. Both of these indicators reflect not only the increasing efficiency of washing station management but also the increasing quality of cherry being provided as inputs to the washing station. Further indicators of yield and quality improvements in the fully-washed coffee produced by the Project can be seen from the rising trend in the percentage of higher quality coffee (AA and A) of the total fully-washed produced, and in the decrease in the percentage of "floaters" (A.3) present in the parchment coffee. The latter indicator (A.3) reflects control of cherry quality accepted by the washing stations and improvements in the control of insect pests which are normally responsible for creating floaters.. E. Foodcrop and Livestock Production Activities and Achievements Foodcrop Production 3.43 The Ngozi II Appraisal Report strongly emphasized the importance of foodcrop development, as follows: "Because of its high population density, Burundi is obliged to preserve and improve its land resources and to develop food production capacity through intensification of cultivation techniques in order to meet the growing requirements for food. This is particularly true for Ngozi Province, which has a higher population density (270 per km2) than any of the other six provinces. Efforts to increase coffee production should therefore be accompanied by attempts to raise the productivity of foodcrops. Fairly detailed information is available on the crops which are grown in Ngozi and on the level of nutrition of the population. There are many sound ideas as to how production can be improved and soil conservation practiced. But thus far, experience in these fields is very limited and agricultural extension work, for all practical purposes, is restricted to coffee growing only." 3.44 On the basis of studies carried out by the French consulting firm SEDES in 1967 and 1973, out of 0.85 ha of :Land cultivated on average per family, food crops - mainly intercropped - were grown on about 0.75 ha and coffee on the other 0.10 ha. The same surveys indicate that food crops not only provided virtually the totality household diet but that they also contributed substantially to farmers' cash income, some 43% of the total, of which, however, about two-thirds is for banana and other beers. While the share of coffee in cash revenues has no doubt increased during the Project period, due to a substantial rise of producer prices since the coffee boom, there is little doubt that the production of sufficient food to satisfy subsistency needs remains largely the farmers' primary concern, and this will be even more true in the future (para. 3.61). 3.45 Based on these considerations, one of the Project's objectives was to increase the level of food production in general, with special emphasis on cereals, oilseeds and fruit production. Since little prior experience with these activities was available, the Project would however concentrate on further exploration of promising courses of action through large scale experiments at the farm level. Therefore, projections were necessarily of an  - 23 - indicative nature and detailed annual programs would have been prepared. Nevertheless, the following targets had been set tentatively at appraisal: (a) Establishment of two seed multiplication farms, covering 30 ha, at full development, and producing annually 30 tons of selected maize seed, plus improved seed and planting material of a number of other crops such as, inter alia, beans, sweet potatoes, cassava, groundnuts and soybeans; (b) Distribution of the selected seed thereby produced to a number of farmers, increasing gradually over the years, with 6,000 farmers cultivating 1,500 ha of improved maize in the last year of the Project. 1/ (c) Distribution of fertilizer to a gradually increasing number of selected maize growers; in the last Project year, 2,000 farmers would be using 75 toins of DAP and 40 tons of urea to fertilize 500 ha of maize; and (d) Study of the response to fertilizer of mixed stands of beans and maize, the most common crop association in the Region, and follow-up on fertilizer trials in high altitude crops (potatoes, peas and wheat) which had been established by FAO and showed promising results. 3.46 Selected seed would have been sold at 10' above the ongoing market price for the crop produce, to prevent domestic consumption, while terms of sale for fertilizers to be used on food crops would be the same as for coffee and, to allow recovery of the cost initially, only coffee growers would participate in the program. Improved seeds and fertilizer were expected to increase maize yields from 1,100 to 1,300 kg/ha and from 1,300 to 2,200 kg/ha respectively. On the basis of these figures, production of maize would have increased by 750 tons in the fifth year of the Project, with 300 tons coming from use of selected seed and 450 tons from the effects of fertilizer. 3.47 Other components of the projected food crop development program consisted of (a) rehabilitating and, if needed, expanding through multiplication, distribution and planting of Setaria grass, terraces which were developed prior to independence to protect soils on steep slopes from water run-off; (b) encouraging farmers to conserve animal manure to improve soil productivity to compensate for the low organic content of soils, and (c) encouraging proper drainage of the marshy bottom lands, some of which are cultivated in the dry season with, however, low productivity due to unsatisfactory control of the water. 3.48 Erosion control and the use of compost or manure were part of the technical package which extension services have tried to persuade farmers to adopt, at least towards the end of the Project. Unfortunately, although no figures are available, it appears likely that little has been achieved. Farmers in particular showed little interest in erosion control works and, furthermore, Project management largely ignored this important aspect of extension activities. 1/ The appraisal report mentions 1,750 ha but this is clearly in error.  - 24 - 3.49 Drainage of marshes in the Nyamuvaga valley, which has reportedly good production potential, started on a limited scale in 1976, with the objective of making 50 ha of land available for cultivation. In the first quarter of 1978, Project management reported that indeed 230 ha had been drained, of which about 60 ha had been made available for cultivation by neighbouring farmers. However, a field investigation by a Bank irrigation specialist indicated that the wording "drain" was totally inappropriate. The 230 ha were located in the middle of a 8,000 ha of marsh surrounded by a large (about 70,000 ha) catchment area. The main and secondary canals which had been excavated provided no control of the river, nor of the water flowing down the catchment basin, and the water level continued to fluctuate with the water table. It seemed evident that drainage of the marshes would have to be considered in a context for the development of the valley and, therefore, would require detailed technical studies 1/ and subsequently the construction of dams and the design of a well laid-out drainage network. Project management was not equipped, nor were funds available from the Credit, for such a sophisticated program. Upon the recommendation of the Bank irrigation specialist, drainage works were stopped immediately and were not resumed until the end of the Project, with the understanding that further developments might possibly be considered under a third phase project. 3.50 According to data provided by Project management, the two main components of the foodcrop development program - the distribution of selected seed and fertilizer - would appear to have fluctuated over the years. Information supplied by the Project and complemented by estimates of the area cultivated with fertilizer and improved seed is summarized below up to 1980: 1/ A study on Drainage of the Nyamuyaga Valley had been prepared in 1960 by the consulting firm SOCINCO but would have needed review and updating.  - 25 - 1976 1977 1978 1977 1980 Appr. Actual Appr. Actual Appr. Actual Appr. Actual Appr. Actual Ent. Eat. Eat. Est. Est. Seed multiplication farms 1 1 2 2 2 3 2 1 2 1 Selected maize seed production (tons) 5 n/a 15 n/a 25 n/a 30 n/a 30 n/a Distribution of: - selected maize seed (tons) - - 5 1.2 15 4.45 2/ 25 2.2 30 0.S - fertilizer - 1.8 17 4 40 0.7 69 10.2 2/ 115 3.5 3/ Cultivated with selected atize seed (ha) - - 250 60 1/ & 750 355 1/ & 1250 110 1/ & 1500 23 1/ 4/ 44 4/ 4/ Area fertilizer 1 7 l/ 75 17 l/ 175 3 1/ 300 44 1/ 500 1S 1/ 2/ Farmers receiving selected seed (no). - 336 300 736 700 1U4 1200 221 2/ 2000 579 Incroeental maize production (tons) - n/a 50 n/a 150 n/a 250 n/a 300 n/a 1/ Estimates based on volume of selected seed and fertilizer reportedly distributed. 2/ Figure probably inconsistent (para. 3.51). 3/ Estimate, based on volume distributed in first planting season. 4/ Maize and other crops, in particular beans and soybeans.  - 26 - 3.51 Although there is a degree of uncertainty in the above figures (para. 3.52) and virtually no data are available for 1981, it is clear that project achievements have been far below appraisal targets, except possibly during the first Project year. Initial Bank supervision missions reported good progress with both the fertilizer distribution and the seed multiplication program. An expatriate agronomist joined the Project in the second half of 1976 and, in addition, a technical advisor from the Rwandese research institute (ISAR) visited the Project in June 1976 and again in the last quarter of 1977 to help management prepare the fooderop development program. A first seed multiplication center established on a former colonial farm became operational in the first Project year. A second and a third center were opened in 1977 and in 1978 respectively, thus exceeding the appraisal objective which called for two centers. Farmers appeared to show genuine interest in using fertilizers on their subsistence crops. Indeed, a program of fertilizer trial plots both on hills and bottom land was set up with the assistance of FAO specialists and many more farmers than anticipated volunteered to participate, even though fertilizers were sold for cash. 1/ Preliminary results of the demonstration tests indicated a clearly positive effect of fertilizer on hillsides 2/; the effects of fertilizer on bottom land with unknown and varied soils was inconclusive. The initial favorable reaction of the farmers led management to believe that a considerable expansion of the foodcrop development program would take place rapidly and that more than 3,000 farmers would be likely to participate as early as 1977, i.e. a year earlier than assumed at appraisal. 3.52 The expectations, however, were much too optimistic, even though project records indicate a substantial increase in 1977 in the number of farmers applying fertilizers, from 336 to 736, while improved seed was distributed to 143 farmers. 3/ In view of inconsistencies in data provided by project management, however, the reliability of the foregoing data is very doubtful. 4/ 1/ At FBu 100 kg NPK, at the beginning of the planting season for which farmers reportedly paid in cash in a period when they are allegedly short of money. 2/ Particularly on beans, the main staple crop. 3/ It is not known if the 143 farmers are in addition to, or part of, the 736 using fertilizers. 4/ According to the table in para. 3.49, in 1979 some 50 kg of fertilizer were distributed to participating farmers, which would be about ten times the average volume per farmer in any other given year. There is also considerable unexplained fluctuation in the reported distribution of selected seeds per participating farmer, from 1 kg in 1980 to more than 30 kg in 1978, enough to cultivate close to one hectare (the average area of foodcrops per holding is only about 0.75 ha).  - 27 - 3.53 An in-depth supervision carried out in April 1978 confirmed the positive effect of fertilizer on yields in demonstration plots, but reported poor management of the three seed multiplication farms. Neither the layout nor the administration of the stations were following any comprehensive production plan and coordination between the three centers was extremely poor. The Bank urged Project management to take a number of measures in order to remedy the situation, which would have consisted of: (a) drawing a map/plan for each center, indicating the various plots, the location cf buildings and existing control works; (b) limiting seed multiplication to major foodcrops grown in the area, in view of the limited capacity of both extension services and farmers to adapt to new crops and varieties; (c) establishing close coordination between the three centers; (d) preparing a three-year program in close cooperation with extension officers and farmers, taking into consideration the farmers' needs and the potential number of participants; (e) illustrating the pro(luction program by simple drawings; and (f) systematically recording all yields and production. In addition, the supervision mission requested Project management to prepare for mid-78 a detailed report on the foodcrop development program, providing detailed information on technical results obtained since the start of the Project and on costs involved, to show whether all elements were available to prepare a more comprehensive scheme within the Project, how extension services would transmit new technology to farmers and how implementation of the program would be maintained. 3.54 Not only was the report requested in early 1978 never delivered, but an early 1979 supervision mission found that recommendations for improving seed multiplication centers' operations had been totally ignored and that the whole foodcrop development program was collapsing. Only one of the three seed farms was still in operation, without any work program. In spite of a labor force largely in excess of normal requirements, production of the farm appeared to be well below potential. Moreover, there were rumors that part of the crop was being diverted for the benefit of the staff in charge of the farm, which would explain why no production data were being kept. Still more disturbing, it appeared that, because not enough selected seed was available from research, or because it had not been ordered, Project management was simply using unselected material for demonstration plots which, in a number of cases, yielded less than the seeds used by the farmers on the rest of their holdings. Even fertilizer trials showed disappointing results, presumably because inexperienced and/or unmotivated extension staff failed to prop9erly monitor fertilizer application on the demonstration plots which, moreover, were often located far from any road and out of sight of other farmers, thus considerably limiting their educating role (if they had been successful).  - 28 - 3.55 The Bank reported to Government its concern with the poor performance of Project management and cautioned that it would be difficult to include a large foodcrop component in a third Ngozi Project if no convincing results were achieved. Corrective measures were taken by a new expatriate technician who replaced the former, unreliable adviser in December 1978. In agreement with the BanK, the foodcrop component was redefined -- startling virtually anew -- and substantially scaled down in order to ensure better control. Seed multiplication was limited to only one center, in view of the shortage of qualified staff. For the first time, a well-prepared production plan was established for the seed multiplication farm, based on an appropriate rotation, with a work program over four years. Operating costs and yields of major crops were to be systematically recorded to analyse the cost of seed production and yield potentials. In the first systematic attempt to distribute improved maize seed and fertilizer, a total of 221 farmers were selected in three separate areas, to participate in the Sept:ember 1979 crop season. Out of the 221 farmers, 75 were to be closely supervised and assisted by qualified extension agents who would also systematically assess the results and productivity of resources used (i.e. husbandry, fertilizer application and plant protection, etc.) 3.56 The test demonstrated that, with selected seed, adequate fertilizing and proper husbandry, maize yields could be substantially increased, reportedly from about 900 kg to more than 2,000 kg/ha. The test also showed that the "Bambou" maize variety was well adapted to the region and that its taste was well accepted by the population. In view of the encouraging results from a systematically conducted test and distribution of selected seed (maize and beans) it was expanded to 332 farmers in June 1980, when marshy bottoms are being planted. However, this time no data could be collected because farmers apparently prematurely harvested their maize to avoid theft and bean production suffered from sustained insect damage. This seems to have discouraged farmers from taking part in the September 1980 program, as only 247 participated, hardly more than in the September 1979. 3.57 After the sudden departure of the expatriate agronomist in November 1980, the whole program deteriorated once more. A May 1981 supervision mission noted that the work plan which had been prepared one year earlier for the seed farm was not being followed, and that no costs and production records were being kept. Growth on large patches of land was poor for reasons which could have been easily detected and corrected in time. The labor force was even larger than in early 1979 and was mainly employed on 8 pineappled field, not doubt of more interest to management than to farmers. Demonstration plots at far level had also deteriorated as well. Fertilizer and selected seed had been distributed, but no records were available and, in the absence of adequate supervision by senior staff, monitoring of the trials was clearly unsatisfactory. 3.58 The food crop development component was not a total failure. Indeed, it appears that whenever farmers felt that trials would be adequately monitored they responded reasonably well, demonstrating that under favorable circumstances they would be receptive to improved husbandry and to new varieties. The project has also tested a number of crops and at least found some varieties of maize and beans to be suitable to the region, in particular  - 29 - the "Bambou" maize. 1/ Finally, it was shown as anticipated, that crop yields can be substantially increased with appropriate doses of fertilizers, although the tests should be confirmed in a statistically reliable way, with more trial plots and over a longer period of time, in order to be convincing. 3.59 Physical achievements in foodcrop development have, nevertheless, been much less than the targets set at appraisal. Incremental maize production has been negligible at best and overall, little has been contributed to saving farmers' income from foodcrops. The objectives set out at appraisal were obviously over optimistic in expecting farmers to adapt to both new varieties and new techniques in a short (i.e. five year) period of time, particularly under difficult conditions with respect to expert advisers, extension, and availability of imports. The lack of prior foodcrops research on which to base field trials, demonstrations and seed distribution also inhibited progress, as well as the lack of permanent cooperation between research and extension during execution of this component. In Burundi, foodcrop development has traditionally been subordinated to export crops with the result that Project management naturally -- perhaps instinctively -- concentrated mainly on the coffee objectives of the Project. The foregoing did not contribute to the motivation and dedication of the staff, except during the short period (1979-80) when a second technical adviser was in charge of foodcrops. 3.60 Notwithstanding the over-optimistic nature of the goals at appraisal, the Project could and should have achieved substantially more in foodcrop development. Bank supervision missions repeatedly pointed out to Government and to Project management that the situation was deteriorating and recommended corrective measures. This resulted in the encouraging, though belated, improvements of 1979-80 which, unfortunately, were not sustained. A 1978 Bank supervision mission noted that, due to demographic pressures in various parts of the Province, coffee plantations were being reduced to make room for subsistence crops, while new selected coffee trees were being planted in marginal soils, as no new fertile land remained available. These alarming signs indicated that considerable efforts were needed to increase foodcrop production, as a prerequisite of future coffee development. The opportunity offered by the Project to start a balanced development of both coffee and foodcrops has, unfortunately, been largely missed. 3.61 As a consequence, since foodcrop development is expected to be the keystone of the Ngozi III Project, this component will begin with little basic information with which to get a quick start. The third Ngozi Project will be confronted with the important issue! of defining an appropriate policy regarding the application and pricing of fertilizers. Tests conducted in 1979/80 would indicate that selected seed, combined with adequate husbandy and fertilization could increase maize yields from 900 kg/ha to an estimated 2,000 kg/ha (para. 3.56). With a largely expanded program, the increase in yields would probably be lower, of course. Nevertheless, with the present cost of fertilizer at about FBu 55/kg and the recommended application for optimal yield increase of 150 kg/ha of DAP and 80 kg/ha urea, additional production 1/ In neighbouring Rwanda, "Bambou" maize is also performing well but the taste is disliked and farmers show no interest in developing it for human consumption.  - 30 - would amount to about FBu 12,600/ha. Even with an increase if 1,100 kg/ha in maize yields, the incremental revenue from maize would be about FBu 13,000/ha at present prices of FBu 12/kg. If farmers were to bear the full cost of fertilizer, the market price of maize would have to go up to FBu 23-35 kg in order to provide incentives to use fertilizer, based on a ratio of two to three times the costs of the input. In addition, part of the incremental maize production would probably be used for subsistence, therefore generating no cash income. Consequently, since there is no alternative to the application of fertilizer, decisions on the relation between fertilizer subsidies and maize prices need to be made soon and frequently reassed to meet both economic and financial necessities. Livestock Production 3.62 It has been assumed at appraisal that the Project would assist in improving the production of livestock, mainly by distributing improved breeding stock of sheep, goats, rabbits and possibly poultry, and would also provide for training of draft animals. The establishment of two modest breeding farms was being considered, which would raise and issue improved stock to interested farmers. These were only ideas, however, and disbursements for this component were conditional on the preparation by Project management of detailed proposals acceptable to the Bank. 3.63 Because Government has not been able to define a national livestock development policy, and within that framework a comprehensive livestock scheme for the Ngozi Province, no pilot livestock program was ever prepared by Project management. In 1980, this component was abandoned altogether and available funds in the Credit were reallocated to other categories. It now appears that given the high population pressure and relative scarcity of land, there is little potential for cattle development in Ngozi Province. Only the development of small stock appears to be justified, with a view in particular of improving the diet of the rural population. F. Other Project Activities Civil Works 3.64 After the high construction costs experienced during Ngozi I, mainly because of the remoteness of the Project area and limited competition among private contractors, it was decided for Ngozi II that all construction under the Project would be carried out by force account. A Project construction unit was established for this purpose and was hired to direct it. Throughout the entire Project implementation period (1976-1981), construction activities were seriously affected by sporadic shortages of cement and other construction materials due largely to circumstances beyond the control of the Project 1/ While farmers initially were reported to have fertilizer for foodcrops against cash, subsequent fertilizer was free of charge, as well as selected seed; this is contrary to assumptions made at appraisal.  - 31 - management. 1/ Hence, on several occasions, construction works had to be interrupted for several months. The ensuring delays in the building program also added to building costs which more than doubled. As a result, since the start of the Project in 1976, funds originally earmarked for the construction program were not sufficient to cover actual expenditures, which have been a major factor in the overall cost overrun of the Project. 3.65 Considering the difficulties which the construction unit had to face since the beginning of the Project, it has accomplished a most satisfactory job. It has built ten washing stations, an office complex with a central warehouse, district input stores as well as houses for senior and medium level Project staff (Annex V). Most of the original construction program of the Ngozi II project was completed only about a year behind schedule. These achievements highlight the efficiency of the Project construction unit and confirm that its approach was the most effective way of carrying out the Project's building program. Water Supply, Bridges and Culverts 3.66 The Project provided for the construction and/or rehabilitation of about 300 potable water resources as well as 300 bridges and culverts. This component was part of the Kuwait Fund co-financing program and was expected to be carried with the help of communal labor. None of these works were done, as all funds earmarked for this purpose were used to cover cost-overrun of the ten washing stations. However, it must be stressed that even without any financial constraints, Project management would have had great difficulties in reconstructing or rehabilitating 300 wells as well as a similar number of bridges and culverts, not to mention the problem of mobilizing communal labor for these works. A rural water development program, as part of a national water development scheme, is presently being implemented by UNICEF in the Ngozi Province. Concerning the reconstruction of bridges and culverts in the Project was less pressing than initially assumed by the appraisal mission. The Project has not been handicapped by any major transport problems, due to poor roads within the Project area. Anyhow, the implementation of this component would have considerably overtaxed the capacity of the Project's construction unit. Under the Ngozi III project, funds will nevertheless be included for rural economic and social infrastructure facilities needed in Ngozi Province. 1/ The main bottlenecks in the supply of building materials were the recurrent congestion at the port of Dar-es-Salaam, the break down of the Zairian cement factory, the handling of all cement imports by Government, and political events such as the Uganda civil war.  - 32 - G. Financial Aspects Project Costs 3.67 Final Project costs are estimated at FBu 983 million, or US$ 10.9 million, as against an appraisal estimate of FBu 675 million 1/, or US$ 7.5 million. This implies a cost overrun of FBu 308 million or 45.61. Total project costs have been calculated on the basis of financial statements for 1976-1980 and estimates of investment and operating costs for January-June 1981, when the Ngozi II Coffee Project is expected to be completed. Total costs exclude outlays for coffee research, (FBu 71.5 million) financed by the Burundi and Belgian Governments. It is not possible to fully reconcile the appraisal estimates with actual Project costs 2/, as Project cost accounting and its breakdown of actual expenditures does not correspond to the categories used in the appraisal report. Moreover, a meaningful comparison is further complicated by the late start of some programs, the deletion of components (e.g. livestock) and reduction in others (i.e. fooderops), thereby somewhat reducing Project scope while total costs nevertheless exceeded appraisal estimates. 3.68 The effect of inflation is best illustrated by the soaring costs for washing stations, which increased from the appraisal estimte of FBu 5.6 million to an average amount of FBu 11.5 million for each of the ten washing stations. The cost of the last washing station, completed in early 1981, was estimated at FBu 17 million, or three times higher than the original cost estimate (Annex VITi). While inflation in Burundi was mainly externally induced, it must also be noted that salaries for national staff and laborers were regularly adjusted to inflation and thus also contributed to spiraling Project costs. In fact, salaries of most employees and skilled laborers more than doubled during the Project implementation period. In calculating the project's price contingencies, the appraisaL report, assumed declining inflation rates from about 137 in 1975/76 to about 107 in 1979/80. In actual fact, inflation rates during the Project implementation period went up to about 271. in 1979/80. Hence, price contingencies were inadequate, particularly in view of the sharp rise in inflation at the end of the Project development period and the one-year slippage in completing t:he Project. Project Financing 3.69 A comparison of appraisal and actual financing is shown below. The Table indicates that actual percentage of IDA and Kuwait Fund financing was substantially reduced compared to original appraisal estimates, whereas Government had to finance a considerably larger share of the Project's cost than initially envisaged. 1/ The original appraisal estimate of total Project costs was FBu 591 million. The increase reflects the devaluation of 15% of the FBu in May 1976. Thus the cost over-run would be higher than in Fi3u terms, if the original FBu cost figure had been taken, in consideration. 2/ Actual Project costs, as reflected in Project official accounts for the years 1976-1980 are given in Annexes VI and VIII.  - 33 - Appraisal Estimate Actual FBu million US$ FBu million US$ IDA 468 69 468 47.6 Kuwait Fund 108 16 108 11.0 Government 99 15 407 41.4 TOTATL 675 100 983 100 The fourfold increase in Government's contribution reflects the fact that it had to bear the burden of financing all over-runs of the Project. Research, whose total cost at completion of the Project is estimated at FBu 71.5 million was financed by the Belgian Government (FBu 40 million) and by Government (FBu 31.5 million) through the OCIBU budget. 3.70 Government financed its share of Project costs entirely through a special tax on all coffee exports. This particular project financing system -- whereby counterpart funds were generated by a tax on coffee exports and not by regular budgetary contributions through a Project account -- caused cash problems for Project management, in particular during the first half of each year. In fact, annual Government contributions were set by OCIBU/Project management at the beginning of each year when an estimate of next season's coffee exports was made. The tax rate was then calculated to match Project financial requirements. However, as exports are difficult to project at the beginning of the year, the Project had sometimes more funds available than required to finance Government's contribution and less at other times, when export projections did not materialize. Moreover, export tax receipts were received two or three months following the coffee season and not regularly over the whole year. The uncertainty and irregularity of the financing system was a source of constant financial problems for Project management, which could have been solved by setting up an adequate pre-financing system for all Project expenditures. However, no provision had been made. Appraisal Report did not for a pre-financing mechanism and during Project implementation Government failed to remedy this constraint in any permanent or efficient manner. Hence, Project expenditures were sometimes pre-financed by OCIBU funds and the Coffee Stabilization Fund, sometimes by CAMOFI or Bank credit. 3.71 Overall, Project implementation has not been severely handicapped by a lack of funds. However, a less cumbersome and uncertain financing mechanism would have left Project management with more time to concentrate on technical matters, instead of spending scarce management time on how to pre-finance Project expenditures at the beginning of each year. Government contributions for all future Bank-financed rural development schemes will be financed through Government's investment budlget and pre-financing will be arranged through CAMOFT. Disbursements 3.72 Because of the delays in Plroject implementation, disbursements were slower than anticipated at appraisal. By the original closing date, 867 of the Credit has been disbursed. The closing date was postponed until April 30, 1982, when it is certain that all remaining IDA funds will have been disbursed. In mid-1979 it became evident that because of continued shortages of fuel and cement in the country, the Project could not be completed by mid-1980 as originally envisaged. It was also evident that the Ngozi III  - 34 - Project was unlikely to start before the original closing date because of delays in Project preparation. Thus in order to ensure continuity in Project management before the start of Ngozi III it was decided that all remaining IDA funds would be used for the financing of priority investments, in particular technical assistance, whereas Government would cover all cperating costs from 1980 onward. The allocation of IDA Credit wras accordingly altersd and the Credit Agreoent amnded ao that unallocated funds were shitted to technical assistance (Category 6) and construction (CaLtegories 3 and 4). Category 4 for livestock was deleted (Annex IX). Accounts and Audits 3.73 Since the beginning of 1976, administration and accounts have been handled in Ngozi and financial information processed according to the procedures of the "OCAN" accounting system. -The previous accounting system had been very unsatisfactory and a thorough review of it at the beginning of 1977, when a Belgian expert accountant was recruited, revealed serious shortcomings and irregularities. Since then, Project's accounting department, bookkeeping procedures, particularly for Project's stores and inventories have bee reorganized and strengthened. In fact, a number of well-qualified Burundian accountants were hired, modern accounting equipment purchased and responsibilities for administration and accounting staff redefined and strictly separated to ensure efficient financial control and management. The accounts of washing stations have been kept separte as their more comercially oriented activities are quite different from the other development and extension activities of the Project. A first balance sheet wlas prepared in mid-1977, covering the period from 1969 up to December 1976: since then, Project's financial statements have been issued at the beginning of each year and all accounts have been up-to-date and of a high standard. Strict and frequent controls of Project's accounts and financial management were carried out by the expatriate financial controller. However, Project. accounts were not audited externally, but regular IDA dupervision missions were impressed by the high professional standard of accounting and efficient internal control, and did not press for an external audit. Howev-r, this position was changed when, during 1980, it became certain that a new Government entity would become responsible for the implementation of the third phase of the Ngozi Project. All Project accounts covering the period 1976-1980 were audited in May 1981 and the auditor's report was submitted. The auditors declined to certify the 1980 accounts for reasons which have since been clarified and/or corrected. Althoigh the reasons for withdrawing the certification of accounts were justifiable, discussions with Project management before issuing the audit report would have clarified most substantive points and prompted corrective actions, which were taken, on methodological issues (valuation of assets, payment for land, accountability for work in progress and past/due receivables, etc.). The accounts for Ngozi II will be closed as of December 31, 1981 and new accounts will be opened for Ngozi III effecltive January 1, 1982. Reporting 3.74 In general, OCIBU/Project reportinig was satisfactory and in accordance with Bank requirements. However, the quality of progress reports varied and reflected the frequent changes in Project management. The quality of reporting improved somewhat when a new F'roject Director was appointed in mid-1980. However, further improvements will be required to make progress reports a more meaningful and management-oriented instrument.  - 35 - IV. INSTITUTIONAL PERFORMANCE A. Project Organization 4.01 The second Ngozi Coffee Iimprovement Project was implemented by OCIBU, a financially autonomous parastatal institution, which had already executed the first Project through a project department whose head reported to the OCIBU General Director. During the second Project more independence was progressively given to Project management in Ngozi, in particular with regard to administrative and accounting matters. However, decisions concerning the recruitment and firing of personnel, international procurement and relationship with Government and external financing remained with the OCIBU General Director. During the early years of Project implementation, working relations were strained between the OCIBU General Director and Project management. Communications were poor and considerable delays occurred when OCIBU assistance was required to recruit personnel and to provide iniputs, equipment and materials. However, with the appointment of a new OCIBU General Director, in early 1978, working atmosphere and coimmunications between the Project and OCIBU markedly improved and, with some exceptions, remained satisfactory until the end of the Project. When the management situation worsened in 1978/79, the new OCIBU General Director decided to entrust the management of coffee factories to a three member management committee (OCIBU General Director, Project Director and the newly recruited expatriate processing officer). This arrangement proved very satisfactory, as it entrusted the de facto management of coffee factories to the expatriate processing officer. 4.02 When the Ngozi Coffee Improvement Projects were conceived, it was certainly reasonable to have placed the management of the Ngozi Coffee Improvement Projects under OCIBU, whose experience and professional competence were recognized. However, with hindsight one must recognize that the effective split in responsibility for Project management was often a source of tensions which tended to aggravate the difficult internal management situation of the Project in Ngozi. This situation has now been alleviated with the creation of an autonomous Regional Development Authority (Societe Regionale de Development) which will be responsible for implementing the third Ngozi Project. 4.03 The appraisal report provided for the establishment of a committee composed of the Minister of Agriculture, the Minister of Finance, the Minister of Planning, the Director General of OCIBU and ISABU, the President of the Central Bank, the Governor of the Ngozi Province, and the Project manager to coordinate and supervise all Project activities. This committee was never established, as it was felt that the OCIBU Board, on which most of the above ministries and institutions were represented, would in practice perform the same functions as the proposed Coimmittee. Central Government control of the Project was rather limited and carried out mainly through the annual OCIBU/Project budget discussions. Project implementation was often handicapped by slow public budgetary procedures and obligation to plan and adjust Project expenditures to strict budgetary allocations. In fact, while OCIBU/Project budgets were prepared on time (October year year) for Board discussion and Government approval (Ministry of Agriculture), final approval was often delayed until March-April of the following year. These delays were  - 36 - frustrating for efficient Project implementation, since forward planning, ordering of inputs, equipment, etc., had to be done early in the year, usually with uncertainty as to budgetary authorizations. Moreover, frequent severe cuts were made in proposed Project expenditures, not because of any shortage of Government funds, but because of a lack of understanding of the nature of development expenditures compared to those incurred by Government administrative departments. B. Project Hanagement 4.04 Helped by a bountiful coffee crop and completion of most project works, Project management began to function smoothly and efficiently in 1981, with the assistance of only four expatriate. This was achieved only after Project management had been frequently disrupted by one crisis or another almost every year since the Project began. These crises were due to frequent turnover of both managers and field staff for a variety of reasons: incompetence, negligence, poor work discipline and low motivation and, on occasions, outright fraud. The results were frequently low morale of field staff and, occasionally, tension between expatriates and Burundians, and ultimately the failure to make the Project independent of expatriate technicians and managers. For example, by mid-1980, nearly all national staff that had been trained during both Ngozi I and II had either been dismissed or transferred to other Government posts. As a resutt, an almost completely new team had to be recruited, except for the four expatriates (coffee processing officer, construction manager, financial controller and coffee extension officer) who had been with the Project only since 1978. 4.05 The Project's first manager, a Burundian who had been deputy Project manager during Ngozi I, served less than one year in that capacity, as he was appointed to a higher post in November 1976 (para 3.01). He was replaced by an expatriate early in 1977 who, although technically competent, was young and inexperienced in the management of rural development schemes. He was not well in touch with the field work of the Project and found it increasingly difficult to get along well with the General Director of OCIBU. He left the Project in 1978 and was replaced by a Burundian who, since 1969, had been associated with both the Ngozi I and Ngozi II projects as Director of Extension Services. He was dismissed in 1980 on grounds of negligence after a major management crisis following the discovery of a widespread embezzlement scheme concerning the sale of pruning tools to farmers (para. 4.07). 4.06 Until 1978 there had also been considerable instability and turnover of expatriate staff. This too hampered effective project management. Three internationally recruited experts were already in post when the Project started at the end of 1975: a coffee processing officer, a construction officer, and a coffee extension officer. The internationally recruited Project Manager, a second Extension Officer for fooderops and the Financial Controller joined the Project during the second half of 1976. The new team, however, had a difficult start in view of the gap created in Project management at the end of 1976, which led to a general breakdown of discipline and morale among all Project staff. The appointment of an expatriate Project Manager coincided with a period of tensions between expatriate and local staff in mid-1977, during which the expatriate coffee processing officer was fired. Moreover, at that time, Government restrictions on the use of official  - 37 - vehicles and the elimination of travel hotel allowances for expatriates working under Government contracts made working conditions difficult. The contract of the first Financial Controller had been terminated only after three months for reasons of incompetence and bad health. The Project Manager resigned in mid-1978, and the extension officers also left at about the same time. Up to that time, experience with internationally recruited management and technical staff had not achieved the intended purposes. The difficulties stemmed in part from OCIBU's recruitment procedures. Hiring decisions were only based on the candidate's C.V. and OCIBU generally chose from a very short list of candidates --- sometimes one -- since very little publicity had been given to job openings. 4.07 Lack of discipline, poor motivation and, on at least four occasions since 1978, discoveries of fraud armong staff responsible for key Project operations (e.g. washing stations and extension), contributed to the disruption and weakness of Project management throughout the execution of the Project. For example, in 1978, Project management had to dismiss two senior staff members who had become responsibile for directing the coffee washing stations after the departure of the expatriate manager in early 1977. Both had been trained for five years to assume this responsibility; however, under their management it was discovered that the operations of the washing stations had come to a virtual halt because they had falsified the payment lists. Subsequently, four young Barundi university graduates who had been hired in early 1979 to eventually join Project management with responsibilities for extension services and washing stations, were fired for lack of work discipline and poor motivation. Project management was shaken at its foundations when, in early 1980, a widespread embezzlement scheme based on malversation of proceeds from the sales of pruning tools to farmers, was discovered. It involved eleven out of fifteen of the Project's agronomes (extension supervisors) who had retained for their own personal use about one-half of approximately FBu 2 million. A similar embezzlement scheme, on a smaller scale, had already been discovered in mid 1978. Since dismissal of the alleged culprits would have left the Project with only four of the fifteen agronomes, all were retained. Apparently the leniency was misinterpreted, as the practice of diverting for personal use funds derived from the sale of pruning tools continued in 1980; in 1980/81 it extended to funds collected from farmers from the sale of fertilizers. Finally, more than half of the fifteen agronomes were fired in 1981. But proceeds from the sale of fertilizers continued to be unaccounted for at the end of 1981, indicating that the problem still had not been brought under control. 4.08 The difficulty of management in imposing sanctions on the perpetrators of fraud illustrated above made it doubly difficult to impose work discipline and motivate field staff responsible for extension services. The agronomes, who are generally responsible for rural development at the district level and were thus expected to play a key role in the implementation of the Project, were apparently ill-prepared for their tasks. Their four-year training at the Burundi Training Institute (ITAB) is rather academic and little field work is included in their program of studies, so that upon graduation they are inexperienced in the realities of agriculture and lack completely in experience for super-vising the moniteurs (field agents) who work directly with the farmers. Once in the field, the agronomes seemed to be rather frustrated and showed their lack of motivation by doing little or nothing, an attitude which was then reflected in the work of their subordinates. The seed multiplication farms (for the foodcrop development  - 38 - program) failed partly because of a lack of diligence by those managing them, in failing to maintain appropriate records for trials, and tolerating the diversion of the farms' output for personal use by the staff (para. 3.54). The apparent lack of motivation and work discipline of extension staff has been all the more disappointing as those working for the Project were better paid, equipped and housed than their peers working for the Plinistry of Agriculture in other parts of the country. Moreover the Project has a well defined extension program, based on the training visit (T&V) system within a well established hierarchy, prescribing simple technical packages and techniques to farmers. 4.09 There is no immediate remedy to the ineffectiveness or lack of sanctions imposed for inefficiency or fraud. That is, in fact, one of the most disappointing aspects of the Project since it had been expected that upon Project completion, it could continue independently of expatriate technicians and managers, one of whose important tasks has been to maintain a degree of accountability for funds and work performance. After more than twelve years of project implementation with expatriate technicians and managers, therefore, the third phase of the Ngozi Project will start with a new technical assistance team of seven internationally recruited experts.  - 39 - V. PROJECT IMPACT 5.01 The impact of the two IDA-financed Projects on coffee production in Ngozi Province is as difficult to assess upon completion of Ngozi II as it was during preparation and appraisal. As in 1974, the only reliable figures are those of relaized FOB prices, and/or exported quantities, which are usually equated with production. Even though systematic surveys of coffee plantations and production started in 1977, they stopped in early 1979, because of lack of fuel and vehicles at that time. The need for more reliable information is further discussed in para 5.08. 5.02 The annual coffee production figures for the country as a whole are quite accurately known, since virtually all of it is exported. Only insignificant amounts are internally consumed in urban areas, and coffee growers themselves do not drink coffee. The official export figures may be understated because of unrecorded sales in border areas with Zaire and Rwanda, but the quantities involved are probably insignificant. Production figures adjusted to compensate for annual fluctuation indicate that the production of Arabica coffee has been increasing at an average rate of 2.47 per year during the last 18 years and that of robusta coffee (grown on a few private estates) by 9.7% per year 1/. However, during the last five years the average annual rate of arabica production has been declining by -0.1%. These figures compare unfavorably with those given in the Ngozi I PCR, which showed that the production of arabica coffee had been increasing at an average of 3.47 per year during the previous 13 years, and that of robusta by 12.5% per year. During the implementation period of Ngozi I, it appears (paras 1.01 and 1.02) that coffee production is most coffee growing areas has tended to decline since 1974/75. This decline, however, has to a large extent been offset by the increase of coffee production in the Ngozi Province. 5.03 Coffee production is estimated to have increased at an average annual rate of 101 annually during 1976/77 - 1980/81. This is based on adjusted production estimates for 1976-78 and OCIBU marketing data since 1977/78. Except for the production of the coffee washing stations, which at present process about 67 of the national coffee production, all coffee is pulped and dried by the farmers themselves and sold to local traders, who in turn arrange for transport to one of the hulling factors in Bujumbura. Since 1977/78, coffee marketing has been organizied by OCIBU and all coffee traders are 1/ The most striking feature in annual production series is the large fluctuations from one year to another. A record production of 26,000 tons green coffee in 1974, for exaimple, was followed by a 7-year low of 15,000 tons in 1975, and again in 1979 and 1980 production figures were respectively 26,000 tons and 17,000 tons. Weather conditions are believed to be the major determinant factor for the volume of the harvest and a year of unusual high production is invariably followed by one of low yields, during which the trees recuperate from the previous year's heavy bearing. It is generally believed that consistently good maintenance of the plantations reduces the annual fluctuations to some extent, since the trees are then in better shape to withstand the demands of heavy bearing. In order to eliminate the effect of the yearly fluctuations, therefore, 5-year moving averages of annual production have been calculated (Annex XIV).  - 40 - required to indicate the origin of their coffee purchases upon delivery to one of the hulling stations in Bujumbura. Coffee traders are selected on a regional basis by OCIBU and Provincial Administrators who, in turn, control coffee marketing in primary markets to ensure that only authorized traders purchase coffee in their districts. OCIBU believes that this system, after initial difficulties, has provided relatively reliable information about the geographical distribution of coffee production in Burundi during the two coffee seasons, 1979/80 and 1980/81. Prior to 1977/78 regional distribution of coffee production was based mainly on agricultural statistics collected by Provincial agricultural staff. These statistics usually underestimated coffee production by about 10 to 20%. It is, therefore, likely that coffee production in the Ngozi Province was underestimated during the 1976/77 and 1978/79 seasons. According to these estimates, coffee production of the Ngozi Province averaged between 35 and 457. of national production during the last decade, with a definite upward trend during the last five years. In fact, the three moving average production for Ngozi indicates an average annual growth rate of about 107 during the years 1976/77 - 1980/81 (Annex XI). 5.04 On the basis of available information, its is difficult to assess to what extent the growth in coffee production was caused by higher productivity of existing trees or by an increase in the number of trees (paras 3.21 - 3.23) and Annex XII. Nevertheless, if the assumptions made about the number of trees are correct, it can be concluded that in the Ngozi Province coffee production has been growing by about 2% per year because of a corresponding increase in new plantations and by a further 87 because of increased productivity of the trees, as discussed below. 5.05 Projected yield at full development of Ngozi II (1991-1995), including the effects of Ngozi I, is expected to be 425 g oi parchment coffee per tree annually (XII and XIII). This is admittedly lower than the average yearly production of SOOg/tons in the country as a whole estimated by the appraisal mission. It is now apparent that the yield estimate made during appraisal was too high because the total number of coffee trees had been underestimated by a comparable factor to that in the Ngozi Province (327). On the assumption that the underestimate is comparable to that found in Ngozi, a revised estimate of the countrywide tree population at 1975 would be 65.9 million, producing 19,000 tons green coffee, which would be equivalent to an average of 380g parchment per tree per annum rather than 500 g. The lower yield figure was confirmed during the preparation of the Kirimiro Rural Development Project in 1978 1/. The revised countrywide yield estimate is in line with the observed yield from the audited tree population of the Ngozi Province of 1978/79, which was 380g parchment per tree (tree population 27.9 million and production of 10,765 tops of parchment coffee). The average incremental yield between pre-Project and with Project can be expressed as a 72g parchment, which is in line with the 75g parchment assumed in the PCR for Ngozi I. 1/ Surveys conducted in the Kirimiro region - the second most important coffee producing area in Burundi - indicated that average yield figures varied between 250-350g parchment coffee/tree.  - 41 - 5.06 As indicated in the study on tree population and potential yields (Annexes XII, XIrl and XIV), maximum production is not reached until 1985, four years after the presumed Project completion. However, because of the eight year period required for young trees to reach the standard five year pruning cycle, combined with the five year pruning cycle adopted for mature trees at the commencement of the Project, full development is not attained until 1991 (even without a third phase, i.e., Ngozi Irr). It is equally obvious from this study that considerable variations in production can be expected each year, which reflect the number of trees at any given point of the five year pruning cycle. These variations are completely divorced from fluctuations caused by climatic conditions or latency following a bumper harvest. 5.07 The long term impact of the Ngozi I and Ngozi II projects is indicated below, assuming that production of parchment coffee increased at an annual rate of about 107. and yields per tree by about 97g parchment coffee during the implementation of Ngozi II. ('000) (tons parch) (gm parch) Project Period Year No. of trees Production 1/ Yield Per tree Pre-project (1976) 22,2213.20 8,333 359 Project Completion (1981) 28,71L.00 13,099 456 Difference 5,482.56 4,766 97 Full Development 29,569.56 13,378 2/ 452 Difference 3/ 6,341.36 5,045 93 1/ Based on official production statistics (Annex XI). 2/ Mission estimates (Annex XII). 3/ Full project impact. If, as previously estimated (5.04), 2. of this increased production is due to extension planting prior to Ngozi II, then 77g of parchment coffee (87) represents the increased productivity due to better husbandry and inputs, i.e., mulching, pruning and minimumal fertilizer application. This figure is comparable to the assumptions of the PCF for Ngozi I (75g) and agrees with the mid-Project year (1979) projection discussed at Para 5.05 (72g). The result of the Project's husbandry activities (pruning, infilling, replacement and extension) does not reach equilibrium until 1991 when the average yield per tree over a representative five year pruning cycle (1991 to 1995) is estimted at 452g of parchment coffee. The average incremental yield per tree of 93 g of parchment coffee is then the outcome of all aspects of husbandry inputs, including tree plantings, carried out during the five years of implementation of Ngozi II. It should be noted that large fluctuations in production from 1991 onwards in any given five year period (high of 13,820 tons to a low of 12,700 tons parchment coffee) are a result of the distribution of trees in any given phase of the five year pruning cycle.  - 42 - 5.08 There is clearly an urgent need for better information on tree population and the productivity of the coffee trees at various levels of coffee husbandry, not only in the Project area but at a national level as well. In the absence of reliable yield figures for the Project as a whole, it is almost impossible to disentangle the effects that improved husbandry practices (rpuning, mulching, pest control and fertilizer application) may have had on the yields. The proposed third phase of the Project has included a formal monitoring unit to systematically collect and evaluate the required data; this, combined with ISABU research findings, should provide more conclusions answers upon completion; hopefully most of the third project. B. Production of FWA Coffee Price Differentials for Fully Washed Coffee 5.09 A basic aim of the first and second Ngozi Coffee Improvement Projects was to achieve an export premium for the fully washed coffee produced by washing stations over the traditional semi-washed coffees produced by farmers. By extrapolation with Kenyan coffees, which are processed in the same way as by the Project washing stations, at appraisal of both the first and second phases, it was presumed that the premium of fully washed high altitude arabica coffee over semi-washed arabica coffee (OCIBU 3A/3B) would be between 10 and 157. 5.10 Upon completion of the Ngozi I Project (1975) the actual premium was only 57, probably because the small quantity of fully washed coffee produced by the Project's five washing stations had not established a reliable "identity" on world markets. In the succeeding years of the second phase of the Project, as the quantity of fully washed coffee increased, the average premium has steadily increased to 157 in 1980 (Annex XV, Tables 1 to 16) and appears to have remained at this level. 5.11 The following Table summarizes the performance of washing stations and indicates the rapid increase in total production of FWA coffee and the favorable price differential progressively obtained for Burundi FWAA/A coffee in world markets. Other technical performance indicators (para 3.46) underline the success of the Project's washing station component.  - 43 - 1976 1977 1978 1979 1980 Total Production Clean Fully Washed Coffee kg ('000 FWAA/A) 265 355 450 1,242 899 3,191 Price Differential: Fully 1/ Washed minus OCIBU 3A/3B (BuFr per kg) 26.01 -245.52 0.78 50.31 40.86 Price Differential in 7 30 - 12 16 15 Total Differential Benefit (BuFr '000) 6,892 -82,249 13,851 62,485 36,733 37,712 (119,961) 2/ Source: Annex XX Table 2 1/ In actual prices. 2/ Excluding 1977. The total differential value of FWA coffee production during the 1976-1980 Project implementation period amounted to only FBu 37.7 million because, in 1977, the fully washed coffee was inexplicably sold by BCC to the USA at approximately 457 below the unit value for Burundi semi-washed coffee. If total benefits are thus compared with total FWA coffee production, the price differential obtained is only about F Bit 11/kg or much less than its additional processing costs. This grace loss illustrates that all Project efforts to produce FWA coffee can be nullified, if other institutions in Burundi's coffee economy are not geared to the marketing of a quality product. 5.12 The present post-washing station infrastructure in Burundi is unlikely to improve upon this 157 premium, since the sophisticated milling and electronic sorting equipment required for the highest premium mild arabica coffees are not available in Burundi. The small milling units run by OCIBU are quite satisfactory for the lower quality semi-washed arabicas but are far too rudimentary for the stringent milling requirements of the high premium milds. Further, the BCC marketing system currently employed for the fully-washed coffees is identical to that used for the semi-washed coffees. To achieve the highest possible premiums, it is essential that the marketing system for the fully washed coffees be converted to an open bidding auction system (including liquoring facilities for potential buyers to taste the coffee) and not be the longer term closed contract bidding, currently employed for the semi-washed coffees. Finally, the highest premiums are awarded for the highest tasting quality of coffee. Improvements in premium must necessarily come from further improvements in quality starting from the coffee hybrids planted, through improved husbandry practices, processing, milling, sorting, marketing, storage and transportation. Each phase of the coffee industry has contributions to make in maintaining or improving the quality of the end product for the gustatory benefit of the final consumer and the economic benefit for Burundi. It is essential that each participating entity contribute to the whole in an integrated manner,  - 44 - preferably through the unifying directives of an integrated management system. Thus, with Government's decision to set up more washing stations in the Ngozi Province and in other coffee growing areas of the country (e.g., Kirimiro), a complete structural change in the present Burundi coffee production, processing and marketing system will be required, whereby the separate institutions presently involved in the management of the coffee sector should be brought under a unified management. C. Coffee Prices 5.13 The average export prices for Burundi's Arabica coffee and official producer prices during the Project period were as follows (at current prices): 1975 1976 1977 1978 1979 1980 1981 /76 /77 /78 /79 /80 /81 /82 Average Exp. Prices BuF/kg clean coffee 103 235 479 241 328 278 195 1/ Appraisal 2/ Estimates BuF/kg clean coffee 86 109 136 169 194 219 237 Producer Price BuF/kf parchment coffee 39 65 112 112 115 115 115 Producer Price BuF/kg equivalent clean coffee 54 91 154 152 153 154 154 1/ Shares of farmers 7 52 38 32 53 48 55 79 1/ 1/ Estimates. 2/ Appraisal report's coffee price projections in 1975 constant terms were converted into current terms by applying the International Price Index. World Bank November 12, 1980. The above Table indicates the exceptional world market conditions for Burundi coffee at the beginning of the Project implementation period, when coffee prices more than quadrupled because of the Brazilian frost in 1975. The exceptional character of the coffee boom becomes all the more evident when actual world market prices are compared with those projected by the appraisal mission in 1975. To pass on part of its higher coffee export earnings, the Government raised producer prices from FBu 39/kg parchment coffee in 1975/76 to FBu 112/kg in 1977/78, thereby almost a tripling of farmgate prices over a two-year period. Despite the substantial rise in producer prices, Government reaped the major profits from the coffee boom through export taxes and contributions to the Coffee Stabilization Fund, which in 1977/78 had accumulated more than FBu 4.0 billion. In fact, Government's share of total coffee export earnings rose to about 507 during the 1976 and 1977 coffee seasons, from 20-307 during previous years.  - 45 - 5.14 This very favorable world market situation changed rapidly when, in 1978/79, world market prices for coffee started to decline. If the decline continues, major adjustments will be required in the near future to adapt Burundi's economy to deteriorating terms of trade. In fact, while world coffee prices declined, producer prices have remained stable at FBu 115/kg of parchment coffee since 1978/79, because a reduction in producer prices in line with world Market prices would have been very difficult for Government to implement for political and social reasons. Thus, while farmers' relative share of Burundi's rapidly falling coffee export earnings increased, the Government's share - essential for the financing of its current and investment budgets- plunged and it could become negative during the present 1981/82 coffee season. In fact, indications are that, because of a further decline of world market prices in 1981/82, Burundi's coffee sliding scale can be financed only with the held of foreign financial assistance (i.e. Stabex) as Burundi's Coffee Stabilization Fund has no more funds available to cover the foreseeable financial gap. In fact, the considerable resources of the Stabilization Fund were mainly used to finance public investment expenditures during past years. Unless world prices for coffee recover soon, and since foreign aid is unlikely to continue to subsidize Burundi's coffee economy, indications are that only a change in the exchange rate would provide the needed revenue. D. Farmers Income 5.15 For reasons explained earlier (para 5.03) it is difficult to determine accurately the net income of the farmers who benefitted from the Project. Based on official production statistics, it can be estimated that the average income increased from FBu 4,036 in 1976 to about FBu 9,072 in 1981, representing an increase of 124%. Most of the increase (78*) is due to the substantial rise in nominal farmegate prices for parchment coffee, while the remaining 227 derives from higher yields. It can be safely assumed that incremental incomes varied substantially, depending on the extent to which farmers adopted improved husbandry methods and on whether or not they sold their cherries to washing stations (para 3.43). Income gains also depended on the importance of coffee plantations, which are not evenly distributed throughout the Ngozi Province. In fact, 447 of farmers have fewer than 150 coffee trees, 357 have between 150 and 300, and 217 own more than 300 trees 1/ Average income figures are based on an average of 173 trees per farmer. 5.16 Whatever the uncertainties concerning the rise in farm income, it is certain that farmers have reaped considerable income gains, due in part to the Project, but mostly due to the exceptional world coffee prices at the beginning of the Project. Since 1978, as the producer prices remained unchanged (para 5.13). It is estimated that in real terms, farmers' 1980 coffee income was more or less at the same level as in 1976, when the Project started. Unfortunately, because of unfavorable worked market conditions the decline in the farmers' terms of trade is expected to continue in the 1/ These figures are based on survey of coffee plantations in Ngozi Province, conducted during the 1977-1979 period by Project Management and Minplan.  - 46 - foreseeable future, which may lead to a growing dis-interest in coffee production and thus seriously hinder future development efforts in this field. Thus, despite five years of Project efforts in coffee development in the Ngozi Province, farmers are not better off now then they were at the beginning of the Project. Admittedly, their present situation would have been worse without the Project. VI. Economic Analysis 6.01 The primary benefits of the Ngozi II Project were envisaged, at appraisal, to be the increased production of coffee and foodcrops and an improvement in the quality of the coffee prqcessed in the washing stations. The economic rate of return for the Project as a whole was estimated at 217; the rate of return of the agricultural services component was estimated at 24*; and that of the coffee processing component at 97.. Foreign exchange was valued at 207 above the official exchange rate and costs for the coffee research program, experimental rural development activities and social infrastructure, were not included in the economic analysis. 6.02 The Project's economic rate of return has been re--estimated for its two main components, namely, agricultural services and coffee processing. The economic rate of return for the agricultural services was found to range between 247 and 1057., depending on the statistical information and estimates used, whereas, the rate of return for washing stations would be between 12 and 227 (para 6.08). These favorable rates reflect the fact that Project's impact on coffee production and throughout of washing stations has been much greater than originally anticipated. The substantial rise in coffee export prices at the very beginning of the Project implementation period and higher than expected price differentials for Project's fully washed coffees have also contributed to these favorable results. 6.03 In recalculating the rates of returns of the agricultural services and coffee processing components of the Project, the following assumptions were made: (a) costs and benefits originally estimated in 1975 prices were updated to 1980 constant terms; (b) in contrast to the appraisal report, foreign exchange earnings and costs were not shadow-priced during the 1976-1980 Project implementation period, because of Burundi's favorable balance of payment position during that time; from 1981 onwards foreign exchange has been valued at 207 more than the off'icial rate to reflect its scarcity; (c) as in the appraisal report, the cost of the research program has not been included in the analysis, as its benefits would accrue only in future projects; (d) in line with the appraisal report, unskilled labor has not been shadow-priced, the net effect of doing so would improve the rate of return;  - 47 - (e) in calculating the economic rate of return, incremental coffee production has been valued at F.O.B. Dar-es-Salaam prices (actual until 1980, estimated thereafter in accordance with Bank projections of November 12, 1980), less processing, marketing and transport costs; the same information was used to calculate the actual and projected differentials of FWA coffee; (f) the life of the Project is estimated to be 25 years instead of the 20 years assumed at appraisal, mainly to reflect the benefits of washing stations, built during the last year of Project implementation. Agricultural Services 6.04 The main objective of the Project was to increase both coffee and fooderop production and agricultural services, therefore, was the largest component of the Project. The annual increase in coffee production brought about the Project at full development (Year 7) was estimated at 1,000 tons of parchment coffee. It was also estimated that about 6,000 farmers would participate in Project's foodcrop development programs and at full development (year 5) would produce an additional 750 tons of maize. Since benefits from the foodcrop development programs have been negligible, they are excluded from the economic ana:lysis. However, development and operating costs of these programs, mainly the running costs of the Gisha see multiplication center, have been included In the cost stream, as these costs will have to be recovered by Project's coffee production benefits. 6.05 Actually achieved incremental coffee production was estimated as varying between 3,000 and 5,000 tons of parchment coffee per year after full development of the Project, depending on the statistical information and estimates used (Annexes XI and XII). Despite the great uncertainties in accurately assessing the impact of the Project on coffeee production (para 5.03), it is clear that that the appraisal target of 1,000 tons of parchment coffee at full development has largely been exceeded. Benefits were taken as the incremental yields of coffee, including the maintenance of yield increases already achieved under Ngozi I, as well as the production of new coffee trees planted during the first phase of the Project. As the impact of Ngozi I on incremental production achieved during Ngozi II is difficult to estimate, all operating costs of the Second project have been included in the cost stream. In fact, the maintenance of the achievements of the first Project would have required the continuation of extension services in the Project area of the first phase. However, all projects costs of Ngozi I have been treated as sunk costs. Actual export prices for coffee F.O.B. Dar-es-Salaam far exceeded appraisal estimates at the beginning of the Project implementation period, as a result of a quadrupling of world coffee prices in 1976/77. However, since then, world market prices have started to decline and are not expected to recover before the middle of the 1980s. The different world market situation as compared to appraisal projections is best illustrated by the fact that, in 1975 constant prices, the appraisal report estimated world coffee prices to increase from USc5O/lb in 1975/76 to USc 76/lb in 1980/81. At present (1981) world market prices for Burundi's semi-washed coffee have dropped to USi 73/lb in current terms. 6.06 Based on the preceding assumptions and considerations, the rate of return of the agricultural services component was first calculated by using  - 48 - the official coffee production figures for the Ngozi Province (see Annex XI and Annex XVI). The rate of return was found to be higher than 100% resulting mainly from the favorable impact of the Project on the quantity of the coffee produced in the Project area. This high rate of return is explained by a number of factors, i.e., (a) the treatment of all Project costs for the first phase as sunk costs; (b) the shadow pricing of all foreign exchange earnings by 20%; (c) the high world coffee prices at the beginning of the Project period; and most importantly, (d) the likelihood (para 5.03) that total coffee production at the beginning of the Project had been underestimated, implying that incremental coffee production throughout the projection period has systematically been overestimated. A second rate of return was therefore calculated (Annex XVI) based on ProJect management's coffee tree population count in 1978/79 and mission yield estimates (Annex XII), which assume a much higher baseline coffee production for 1976/77 than official OCIBU figures. The rate of return for this calculation, 247, is a more likely, though prudent, estimate of Project's economic performance. 6.07 In order to take into account the windfall from the 1977-1978 coffee price increase, coffee prices were averaged over a three year period throughout the life of the Project. By this method the 1977 price would be reduced from FBu 493/kilo to FBu 306/kilo and those for 1978 and 1979 would rise somewhat. The IRR, however, would remain at 24%. If only 1977 prices are reduced to attenuate the effects of the windfall, the economic rate of return remains at a respectable 19.6% Consequently, whwate:ier the rate of return of the Project may be, within the range, it is evident that favorable results were obtained by the Ngozi Coffee Improvement Projects. Coffee Processing 6.08 Even though the appraisal mission regarded the washing stations under Ngozi II mainly as an extended trial period during which the assumptions with regard to price premiums and throughput capacity were expected to be tested, their economic viability with sound management has now been conclusively demonstrated. With efficient meLnagement, washing stations can produce high quality fully washed coffee which in European markets, commands a premium of about 15% over the price of Burundi traditional semi-washed coffee. The appraisal mission had assumed a maximum l010 price differential, reflecting the poor results achieved by washing stations during the first phase of the Project, and had also underestimated the throughput capacity of a three disc-pulper washing station. In fact, while a production capacity of between 125 and 150 tons of parchment coffee was used in the appraisal rate of return forecasts, the actual amount of coffee parchment produced varied between 200 to 250 tons. The higher than expected price differential and throughput capacity had obviously a decisi-ve effect on the economic rate of return of washing stations, which was re-estimated at 22% compared to 9% in the appraisal report (Annex VIII). As in the appraisal report, the mission has treated the investment costs of the four washing stations built during the first phase of the Ngozi Project as sunk costs. 6.09 For the sake of consistency, the rate of return on investments in washing stations was recalculated to account for the lower value of fully washed coffee in 1977, (Annex XVII). The loss, in financial terms, amounted to about BuF 124 million. This one time loss reduces the IRR for washing stations from 22 to 11.75% (say, 12%).  - 49 - 6.10 In addition to the major elements of price differential and throughput capacity in the economic analysis of washing stations, there are number of other factors which woulcl further improve the economics of washing stations. However, because of their altogether secondary importance and difficulties in quantifying them, the mission, has excluded them from the economic rate of return calculations. Firstly, it must be noted that one washing station would replace eight hand-pulpers which ould have to be set-up by Government in the "without" case. Hence, the investment and operating costs of eight hand-pulpers should be deducted from the cost stream of washing stations in the economic arLalysis. Secondly, under the fully-wet processing method, it can be assumed that the ratio of cherries to parchment would improve from 5.5 to 6 in hancLpulping centers to about 4.5 to 4.9 in washing stations. Moreover, the hulling losses for coffee produced by washing stations are about 207 compared to about 257 for Burundi's semi-washed coffee. In other words, in addition to the price differential obtained for FWA coffee, the greater efficiency of the fully-wet process allows for a greater ouput of clean coffee compared to that obtained by the traditional semi-washed process. Thirdly, under the present marketing and hulling system for fully-washed cof'fee in Burundi, a price differential of about 15% is considered a maximum. However, with better hulling facilities (a new hulling station is expected to be built in Ngozi in the near future) and a more integrated marketing system, the price differential may inrease from 157 to 18-20S over the long term, in particular in an over-supply world market situation in which demand for high quality coffee is immune (i.e. demand is inelastic with respect to price) to price changes. If these various factors are taken into consideration, the economic rate of return of washing stations would increase and probably ranger between 257 and 307. Finally, the expectations during the preparation of Ngozi II, that the washing stations are to become focal points for a large number of rural development activities have materialized. There are unquantifiable economic benefits. In fact, the use of washing stations as centers for the distribution of inputs and consumer goods as well as catalysts for the setting up of cooperatives in the Ngozi Province has already been considered. VIT. IDA PERFORMANCE 7.02 The Bank played a major role in Project identification and preparation. During this stage, IDA was urging Government to be more restrained in its objectives, especially the construction of new washing stations. The changes in Project design suggested by the appraisal mission were also reasonable, in particular the further reduction in the number of washing stations to be built and the realistic approach to fertilizer use and financing. Thus, while Bank staff has tried to reduce the Second Ngozi Project gradually to a size which was considered more compatible with Government's execution capacity it turned out that Bank assessment of the borrower's implementation capacity had still been too optimistic. With the benefit of hindsight, it now appears that the program of building the 20 washing stations initially accepted by the appraisal mission (the further reduction to four additional coffee factories was based on ecomic considerations only) would have been much too ambitious, as it took Project management more than five years to construct the 10 washing stations which were finally included in the Project. As pointed out earlier, while the appraisal mission had overestimated the construction capacity of the Project,  - 50 - the successful management of washing stations, in particular during the last three years proved the appraisal mission to be too conservative with respect to the profitability and the economic justification of these stations. In contrast, the appraisal was again too optimistic concerning capacity to carry out a number of other production and infrastructure programs. In fact, neither the experimental rural development activities, including economic and social infrastructure programs, were carried out during the second phase of the Ngozi Project, nor were the pilot foodcrop development programs achieved. Admittedly, when the Ngozi II appraisal mission had to judge the profitability of coffee factories six years ago, data on the merits of these factories was very limited and not too encouraging. 7.02 IDA supervision, both in quality and frequency, was good; staff continuity was maintained in all (except the first) of nine supervision missions, which were carried out by the same Bank staff member during the five-year implementation period. Moreover, as the same staff member was also responsible for the preparation of new rural development schemes and therefore visited Burundi almost every third month, he was able to establish a more frequent dialogue with OCIBU/Project management than is normally possible. The combination of supervision and project preparation proved, indeed, to be beneficial as the experience gained from the supervision of the on-going project was very helpful in the preparation of new rural development schemes (Ngozi III and Kirimiro). Supervision missions were frequently able to highlight the major issues and obstacles to Project implementation, to alert Government officials to these problems, and to suggest appropriate corrective measures. For instance, with the assistance of consultants, supervision missions identified the major problems the Project was facing in its pilot drainage scheme, foodcrop development programs, Government fertilizer policy and coffee research program. Finally, supervision missions acted as project coordinator, if not mediator, to reduce tensions which frequently existed between Project and OCIBU management, as well as between OCIBU/Project management, ISABU and Government, not to mention the strained relations between expatriate and local staff. Despite these often delicate tasks, relationships between IDA and Government, on the one hand, and OCIBU/Project management on the other remained very good throughout Project implementation. VIII. CONCLUSION AND OUTLOOK 8.01 The Project was intended to improve the quality and increase the production of Burundi's most important export crop by more efficient processing of parchment coffee, together with the application of an improved technical package by farmers. A start was also expected to be made to improve farmrs' foodcrops as a necessary complement to future coffee development in the Project area. At the end of the second Ngozi Coffee Improvement Project, it can be concluded with some degree of confidence that measured against appraisal objectives, coffee development results in quantitative as well as qualitative terms have significantly exceeded appraisal expectations. In contrast, the Project failed to meet the appraisal objectives of increased foodcrop production and other social and economic infrastructure programs. This failure stemmed mostly from difficulties in Project management and, in particular, inefficiencies in its extension services. In fact, after twelve years of Project implementation  - 51 - with expatriate technical assistance, the Project did not succeed in training and setting up a national management team nor did it succeed in establishing a well-functioning extension corps. Despite these difficulties, in economic terms the Project has been of benefit to Burundi and substantially contributed to arrest a declining trend in the country's coffee economy. On balance, the Project is perhaps best described as a qualified success. 8.02 Without any doubt, the washing station component of the Project has successfully demonstrated its ability to produce the required product. It has been the Project's most commendable achievement. It now remains for the research program to produce arabica hubrids with the highest possible inherent quality potential (combined with high yield potential and disease resistance), and for the milling and marketing components of the Burundi coffee industry to capitalize on the high quality coffee demonstrably produced from the Project's washing stations. Thus, after the factors essential to efficient and profitab:Le processing of quality coffee have been successfully identified and implemented under the Ngozi II Project, future progress towards improving Burundi coffee to fully-washed standards will depend mainly on continued efficient management of washing stations and Government's willingness to establish a system whereby all entities presently involved in the management of the coffee economy will be geared to the production of quality coffee; this is an objective which can only be achieved under an integrated management. 8.03 On the coffee production side, the Project has partly accomplished its aims. After a protracted struggle, spanning a full decade, the Project has successfully demonstrated its ability to control the major insect predator of the Burundi coffee industry - antestia. This pest should no longer significantly affect either the quantity or quality of the farmer's coffee production and, hopefully, the success of future pest control campaigns will not be impaired by a lack of diligence in management or timing of insecticide applications. 8.04 A further positive aspects for the Project was the adoption of sound mulching practices by most farmers in the Project area. In fact, the majority of farmers (707) recognized the obvious differences in the physical well-being of coffee trees provided with adequate mulch and those with none. The major problem for the future is to provide sufficient available mulching material for the farmer to use. In the long term, the ideal solution will be provided by the introduction of a suitable cover crop, but for the short term all available sources of suitable material should be tapped. In this respect it is unfortunate that the extensioni services have not been entirely successful in convincing farmers about the importance and correct use of mulching grasses like trypsacum and pennisetum. 8.05 The Project has still not proved unequivocally that any advantage of either yield or quality is to be gained by the application of inorganic fertilizers. The Project management: and the Projects' research group still feel that the application of urea to the nitrogen deficient soils of the Ngozi Province must have a beneficial effect. Hopefully, the newly designed trials, which commenced in 1979/80, will give the desired answer during the early part of the proposed third phase of the Project (Ngozi III). However, the result will not only depend on t:he result of the fertilizer research trials, but also on the economics of fertilizer use at that time.  - 52 - 8.06 The Project has only achieved a partial success in persuading farmers to adopt correct primary and secondary pruning routines. This is pernaps most easily seen from the smaller number of prunino tools sold in the second phase of the Project than in the first, despite the greater area and the much larger number of trees requiring rehabilitation. While the increased cost of pruning tools may have played some part in the lack of enthusiasm on the part of the farmer, the major problems seems to stem from the observed lack of motivation and dynamism of Project extension services, which has overshadowed all extension work of the Project. 8.07 None of the experimental rural development activities have been carried out during the second phase of the Project and the results of the foodcrop development programs have been most disappointing. This failure, in our view, is partly due to a lack of a proven technical package and farmers' responsivEness, but mainly to inefficient management extension services. Thus, an important opportunity was missed to start a more balanced development of both coffee and foodcrops in Ngozi Province, which would have been particularly relevant because of the high population density in the Project area and farmers' deteriorating financial situation, which both are likely to overshadow the future development of the Ngozi Province. 8.08 The Project illustrates the dramatic effects of high inflation and rapid deterioration of Burundi's terms of trade on the inconne gains achieved under the second phase of the Ngozi Project. In fact, becatuse of these factors farmers' real income from coffee production has considerably declined since 1978 and can be expected to continue to drop in the medium term future. Hence, with coffee production becoming less and less profitable for farmers, their incentive to properly use modern costly inputs on coffee plantations is likely to diminish in the future. More important, future coffee development must be seen in the dynamic situation of the Ngozi Province, where population will continue to increase and soil fertility and land availability continues to decline. Under these circumstances, farmers will require more time and land to grow foodcrops for their survival, and to feed a rapidly growing urban population. Hence, declining world market prices for coffee and growing population pressure must be considered the main risks, which in the long run, may jeopardize the favorable achievement of the Ngozi Coffee Improvement Projects. 8.09 To help maintain these achievements and to further assist farmers to improve their income and welfare, a third Ngozi Project was prepared by Government with Bank assistance. It is expected to start in early 1982 without interruption from the completion of Ngozi II. While development programs for coffee production and the construction of washing stations will continue, a new start will be made on foodcrop development, albeit limited to six districts in the Project area. Moreover, another objective would be to improve rural living conditions by building or maintaining water sources, equipping social centers and developing wood resources. Finally, by putting strong emphasis on training and providing technical assistance, the Project would enhance the management capability of the Regional Dlevelopment Authority, which will carry out the third phase of the Ngozi Project. Because of the above mentioned constraints it is likely that the third Ngozi Project will have to be carried out in a considerably more difficult environment than prevailed during the implementation period of the second phase. Besides these constraining factors, the success of the continued  - 53 - development efforts will to a large extent depend on the efficiency and motivation of extension services and other national staff, as well as Government's willingness to gear its policies and institutions to the development of the rural sector.  I  - 55 - ANNEX I BURUNDI Second Coffee Improvement Project: Credit 593-BU Project Completion Report Burundi Coffee Exports by Quantity and Quality OCIBU 3B, 4B, FWAA and FWA OCIBU 3A B & HT Years Total Exports Superior Quality Standard Quality Inferior Quality tons tons % tons % tons x 1970 21,218 458 2.16 15,756 74.26 5,002 23.58 1971 23,271 377 1.62 17,611 75.68 5,283 22.70 1972 17,901 70 0.39 10,657 65.12 7,174 34.49 1973 19,600 260 1.33 11,207 57.18 8,223 41.49 1974 26,446 571 2.16 18,636 74.26 7,239 23.58 1975 15,016 228 1.52 7,418 49.40 7,370 49.08 1976 19,982 265 1.33, 8,462 42.35 11,255 56.32 1977 15,127 335 2.22 7,383 48.81 8,409 48.97 1978 20,878 450 2.16 9,045 43.32 11,383 54.52 1979 25,533 1,242 4.86 12,450 48.76 11,841 46.38 1980 16,781 899 5.76 7,211 42.9 8,671 51.6 Source: OCIBU.  - 56 - ANNEX II BURUNDI SECOND COFFEE IMPROVEMENT PROJECT: CREDIT 593-BU PROJECT COMPLETION REPORT BURUNDI COFFEE EXPORTS: TOTALS (Tons) Year /a Arabica Robusta Total 1958/59 9.640 300 9.940 1959/60 20.727 900 21.627 1960/61 9.991 278 10.269 1961/62 12.937 196 13.133 1962/63 13.972 204 13.176 1963/64 5.856 197 6.053 1964/65 17.689 600 18.289 1965/66 12.826 731 13.557 1966/67 14.094 917 15.011 1967/68 17.929 753 18.682 1968/69 15.517 1,023 16.540 1969/70 13.299 1,251 14.550 1970/71 21.218 900 22.118 1971/72 23.271 1,780 25.051 1972/73 17.901 1,502 19.403 1973/74 19.600 1,795 21.395 1974/75 26.446 1,694 28.140 1975/76 15.016 1,908 16.924 1976/77 19.982 1,446 21.428 1977/78 15.174 1,856 17.030 1978/79 21.007 1,776 22. 83 1979/80 25.866 1,862 27.728 1980/81 16.781 1,694 18.475 Sources: Banque de la Republique du Burundi. OCIBU /a From June 1st to May 31st of the folLowinv year.  - 57 - ANNEX III BURUNDI SECOND COFFEE IMPRO'VEMENT PROJECT: CREDIT 593-BU PROJECT COMPLETION REPORT BURUNDI COFFEE EXPORTS: TOTALS (Tons) Year /a Arabica Robusta Total 1958/59-1962/63 13.453 376 13.829 1959/60-1963/64 12.697 355 13.052 1960/61-1964/65 12.089 295 12.384 1961/62-1965/66 12.656 386 13.042 1962/63-1966/67 12.887 530 13.417 1963/64-1967/68 13.679 640 14.319 1964/65-1968/69 15.611 805 16.416 1965/66-1969/70 14.733 935 15.668 1966/67-1970/71 16.411 969 17.380 1967/68-1971/72 18.247 1.141 19.388 1968/69-1972/73 18.241 1.291 19.532 1969/70-1973/74 19.058 1.446 20.504 1970/71-1974/75 21.687 1.534 23.221 1971/72-1975/76 20.447 1.736 22.183 1972/73-1976/77 19.789 1.689 21.458 1973/74-1977/78 19.244 L.740 20.984 1974/75-1978/79 19.525 1.736 21.261 1975/76-1979/80 19.409 1.770 21.179 1976/77-1980/81 19.762 1.727 21.489 Average growth rate on 18 years 2.4% 9.7% 2.7% On the last 5 years -0.1% -0.1% -0.06% /a From June 1st to May 31st of the following year.  - 58 - ANNEX IV BURUNDI SECOND COFFEE IMPROVEMENT PROJECT: CREDIT 593-BU PROJECT COMPLETION REPORT Pyrethrum Test - No. of Insects per Tree 01/72 09/72 01/73 03/74 09/74 03/75 06/78 11/79 11/80 Project area 3.5 2.52 3.11 3.5 1.46 2.7 2.1 2.4 0.1 Outside project area 7.5 9.56 7.06 6.22 4.1 6.1 3.6 n.a. n.a. 100-Cherry Test - % of Damaged Cherries 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 Ngozi 14.1 27.2 32 19.9 33.5 26.7 n.a. 28.0 n.a. 19.1 National Average 15.5 27.5 32.5 20.4 30.8 29 29.2 20.3 n.a. n.a. Source: OCIBU/ISABU  - 59 - ANNEX V BURUNDI SECOND COFFEE IMPROVEMENT PROJECT: CREDIT 593-BU PROJECT COMPLETION REPORT Inventory of Buildings Constructed During Second Phase Location A. Agricultural Storage Facilities Matongo Ruhororo Gisha Marangara Giheta Tangara Mubanga Muramvya Kabamba Mihigo Mparami rundi Jene Buraniro Rango TOTAL = 15 B. Houses for "Agronomes" Rango Maramvya Gashikanwa Mubuga Buraniro Kayanza Matongo Marangara Gatara Jene Gatsinda Ruhororo Mihigo Tangara Kizemba TOTAL 15 C. Residential Buildings 5 houses for Expatriate Staff 4 houses for National Staff 5 houses for medium-level local Staff TOTAL = 14  - 60 - ANNEX V Page 2 LDcation D. Administrative Buildings Office Building at Ngozi Central Warehouse Workshop Water Reservoir Petrol Station E. Washing Stations Mugomera Rukurazo Murambi Murama Rugori Nkaka Mubuga Kibuye mirango Karinzi TOTAL = 10 Source: OCIBU/Project.  BURUNDI Deuxiemie Project d'Amelioration du Cafe Rapport d'Achevement du Projet Analyse Financiere Investissement du Projet 1976 1977 1978 1979 1980 1981 INVESTISSEMENTS 1976-1980 VARIATIONS Projet Usines Projet Usines Projet Usines Projet Usines Projet Usines Projet Usines comptes libell