Tunisia - Third Urban Development Project

Document of The World Bank FOR OFFICIAL USE ONLY Report No. 15713 PERFORMANCE AUDIT REPORT TUNISIA THIRD URBAN DEVELOPMENT PROJECT (LOAN 2223-TUN) June 10, 1996 Operations Evaluation Department 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 = Tunisian Dinar (TD/US$) Appraisal Estimate (1982) $1.00 = .5907 TD Actual: Average 1984 = .7768 TD Average 1985 .8345 Average 1986 .7940 Average 1987 .8287 Average 1988 .8578 Average 1989 .9493 Average 1990 .8783 Average 1991 .9246 Average 1992 .8844 Average 1993 1.0037 Acronyms and Abbreviations AFH Agence Fonci&re d'Habitation (Land Development Agency) ARRU Agence pour la Rhabilitation et la R6novation Urbaine (Urban Upgrading and Renewal Agency) ASM Association pour la Sauvegarde de la M6dina (Association for the Safeguard of Medina) BDET Banque pour le D6veloppement Economique de Tunisie HB Housing Bank CNEL Caisse Nationale d'Epargne-Logement (National Housing and Savings Fund) CPSCL Caisse de Pr6ts et de Soutien des Collectivit6s Locales (Local Communities Support Fund) FNAH Fonds National pour l'Am6lioration de l'Habitat (Housing Improvement Fund) FOPROLOS Fonds pour la Promotion des Logements Sociaux (Social Housing Fund) LA Loan Agreement LCB Local Competitive Bidding MEH Ministere de l'Equipement et de l'Habitat (Ministry of Equipment and Housing) ONAS National Sewerage Authority PAR Performance Audit Report PCR Project Completion Report SAR Staff Appraisal Report SNIT Societ6 Nationale Immobili&re de Tunisie TA Technical Assistance VAT Value Added Tax Fiscal Year: January 1 - December 31 FOR OFFICIAL USE ONLY The World Bank Washington, D.C. 20433 U.S.A. Office of the Director-General Operations Evaluation June 10, 1996 MEMORANDUM TO THE EXECUTIVE DIRECTORS AND THE PRESIDENT SUBJECT: Performance Audit Report on Tunisia Third Urban Development Project (Loan 2223-TUN) Attached is the Performance Audit Report (PAR) on the Tunisia Third Urban Development Project (Loan 2223-TUN, approved in FY82 ), prepared by the Operations Evaluation Department (OED). This was the third Bank-financed urban project in Tunisia. The first dealt mostly with urban transport; the second with sites-and-services development and housing rehabilitation. Once the Government decided to shift its housing policy away from demolishing the existing housing stock to renovating and rehabilitating valuable assets, a third project was prepared to implement the new policy on a major scale. The Agence pour la R6habilitation et la Rdnovation Urbaine (ARRU) was especially created for the same purpose. Among the project sites, there was also a pilot experience to rehabilitate an area (Hafsia) of the old Medina of Tunis while preserving its cultural heritage. The project did help to expand the supply of housing and infrastructure for lower income urban households. Except for the Hafsia component, which attracted some middle-class professionals, all the other sites essentially housed the targeted population (below the median urban income). The appreciation of the US dollar in the mid-1980s allowed to extend the scope with additional sites and a special program to bring sewerage to low-income areas in many secondary cities. This second phase was innovative because the Bank accepted to refinance the additional subprojects under a more flexible credit line approach. This helped the project to achieve most objectives, both physical and institutional, except for the strengthening of the municipal finance agency, Caisse de Pr8ts et de Soutien des Collectivit6s Locales (CPSCL). The recovery performance was disappointing since the frontage tax was introduced late and in only one municipality. The collection on mortgage loans by Caisse Nationale d'Epargne- Logement (CNEL)/Housing Bank (HB) was also excessively low (30 percent, except in Hafsia where it was 50 percent). This project confirmed on a large scale the benefits of the renovation and rehabilitation approach to the low-income housing problem in cities. The Hafsia rehabilitation pilot was so successful that it was extended to a larger area of the Medina. The project also showed that design complexity is not a constraint if the implementing agency is adequate and given sufficient autonomy to carry out its mission. The Audit rates the project outcome as satisfactory, its institutional development as moderate, and its sustainability as uncertain. Bank performance is rated as satisfactory. Attachment 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.  FOR OFFICIAL USE ONLY Contents P reface . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3 B asic D ata Sheet . .................................................. 5 Evaluation Sum m ary . .............................................. 9 1. Background ................................................... 15 2. Project Description .............................................. 17 Objectives .................................................. 17 Components and Financing Arrangements ............................. 17 Implementing Agencies ......................................... 18 Project Design Issues ........................................... 19 3. Project Implementation and Physical Results .......................... 21 Time Overruns and Use of Savings ................................. 21 Procurement and Disbursements ................................... 21 Beneficiary Targeting ........................................... 22 Rehabilitation of the Tunis Medina (Hafsia Component) ................... 23 Economic Rate of Return ........................................ 24 4. Institutional Development ......................................... 25 Technical Assistance and Training .................................. 25 Cost Recovery ................................................ 25 5. Conclusions and Lessons Learned ...... ........................... 29 Borrower Performance .......................................... 29 Bank Performance ................ ......................... 29 Ratings . .................................................... 30 Lessons Learned ................................................ 30 Annex Impact of the Hafsia Component ................................... 31 This report was prepared by Mr. Jean-Francois Landeau, Senior Evaluation Officer, who audited the project in June 1995. Mrs. Maryvonne Mauprivez provided administrative assistance. The report was issued by the Infrastructure and Energy Division (Mr. Yves Albouy, Chief) of the Operations Evaluation Department (Mr. Francisco Aguirre-Sacasa, Director). 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.  3 Preface 1. This is a Performance Audit Report (PAR) on the third World Bank loan to the urban sector in Tunisia. The loan for US$25 million equivalent was approved on December 21, 1982. After three amendments, the unused balance of US$0.45 million equivalent was cancelled and the loan closed on January 28, 1994. 2. The Audit is based on the Project Completion Report (PCR, Report No. 13514, dated September 12, 1994) prepared by the Middle East and North Africa Regional Office, the Staff Appraisal Report (SAR) and President's Report (PR), the loan documents, and a study of the project files. An Operations Evaluation Department (OED) mission visited Tunisia in June 1995. The excellent cooperation and valuable assistance provided by the authorities in the preparation of this report are gratefully acknowledged. 3. The PCR provides a comprehensive account of the project experience. The PAR reassesses the Project's achievements, in particular the Hafsia component which rehabilitated an area of old Medina in Tunis and the cost recovery performance. A social impact survey of beneficiaries in the Hafsia area carried out by the Association de la Sauvegarde de la Medina (ASM) is summarized in the Annex to give this preliminary impact evaluation a broader dissemination. 4. Following standard OED procedures, copies of the draft PAR were sent to the Borrower for comments. However, no comments were received.  5 Basic Data Sheet THIRD URBAN DEVELOPMENT (LOAN 2223-TU) Key Project Data (amounts in US$ million) Appraisal Actual or Actual as % of estimate current estimate appraisal estimate Total project costs 60,485 45,239 Loan amount 25 24.55 Cancellation .45 Date physical components completed 12/31/1989 12/31/1992 Economic rate of return 18-24% n.a. Cumulative Estimated and Actual Disbursements (end of calendar year) FY83 FY84 FY85 FY86 FY87 FY88 FY89 FY90 FY91 FY92 FY93 Appraisal estimate (US$M) .94 5.40 10.92 15.68 19.36 22.35 24.00 25.00 - - - Actual (US$M) .37 .70 2.38 4.22 5.80 8.21 10.31 16.95 20.03 22.24 24.55 Actual as % of appraisal 39.4 13.0 21.8 38.6 30.0 36.7 43.0 67.8 - - - Date of final disbursement: October 30, 1993 Project Dates Original Actual Initiating memorandum 06/1981 Negotiations 05/07/1982 10/21/1982 Board approval 06/29/1982 12/21/1982 Signing 01/11/1983 Effectiveness 04/1983 09/08/1983 Closing date 06/30/1990 06/30/1993 6 Staff Inputs (staff-weeks) 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 Preappraisal 45.7 Appraisal 59.4 13.6 Negotiations 7.2 Supervision 5.3 14.7 9.0 10.8 6.9 7.1 10.3 10.0 8.0 5.6 9.0 Other 8.0 Mission Data Date No. of Staff days infield Specializations Types of problems (month/year) persons represented Identification/Preparation 11/1981 3 Arch. Appraisal 10/1981 1 Arch. Supervision 1 04/1983 1 20 Arch Supervision 2 08/1983 3 Arch., FA Supervision 3 11/1983 2 30 Arch., Eng. Supervision 4 03/1985 1 Arch. Supervision 5 11/1985 2 20 Arch. Supervision 6 10/1986 1 11 Eng. Supervision 7 05/1987 2 14 Arch., Eng. Supervision 8 09/1987 3 10 Urban., Eng. Supervision 9 03/1988 2 15 Urban., Eng. Project Management Supervision 10 12/1988 2 20 Urban., Eng. Supervision 11 02/1989 1 18 Eng. Supervision 12 96/1989 I 10 Eng. Procurement Status Supervision 13 10/1989 I 13 Eng. Supervision 14 12/1989 I 5 Eng. Supervisionl5 02/1990 2 10 Eng., FA Procurement Status Supervision 16 07/1990 1 10 Eng. Supevision 17 01/1991 1 17 Eng. Supervision 18 03/1991 1 Transpr. Eng. Technical Supervision 19 08/1992 1 Eng. Assistance status Technical Supervision 20 12/1992 2 15 Eng.. FA Assistance status Supervision 21 04/1993 2 Eng., FA Project Management Completion 12/31/1992 7 Other Project Data Borrower: Government of Tunisia Executing Agency: Agence pour la R6habilitation et la R6novation Urbaine FOLLOW-ON OPERATIONS Operation Loan Nos. Amount Board Date (US$ million) Fourth Urban Development Project 2736 30.2 07/03/1986 Fifth Urban Devlopment Project 3064 58.0 05/18/1986 Municipal Development Project 3507 75.0 07/02/1992  9 Evaluation Summary Project Description 1. This PAR covers the Third Urban Project (Urban III) carried out by the Bank in Tunisia, which in effect was the first project attempting to introduce a new approach to low- income housing on a large scale. The urbanization pace had been rapid with the urban population reaching 55 percent of the 6.6 million inhabitants by 1980. The rural-urban migration was flowing to Tunis and the large coastal cities such as Sfax, the two cities where pilot upgrading components were implemented under the Second Urban Project (Urban II). The high rate of urban population growth resulted in a great demand for urban housing, densification, over-crowding, scattering, and generally led to the deterioration of the urban environment in the outskirts of cities and within their crowded and cramped traditional quarters (Medinas) (PAR, para. 1.1). 2. This situation provided the rationale for expanding Bank assistance to large scale sites and services/rehabilitation projects in the cities most affected by the migrations. Partly as a result of Urban II then being implemented, the Government had accepted the concept of upgrading instead of demolishing squatter settlements, of developing low-cost sites and services projects, and of recovering infrastructure costs. A public sector agency, the Agence pour la Rehabilitation et la R6novation Urbaine (ARRU), had been created in August 1981 to implement such policies. 3. Project Objective and Components. The project was to satisfy multiple objectives, as it was the practice with many urban projects designed in that timeframe. The SAR characterized the main goal as essentially to increase the supply of low-cost housing affordable to low-income urban dwellers. This broad goal was divided into four complex objectives vs. the provision of housing, the redressing of inequities in the housing policy, the strengthening of the sector's agencies, and the housing policy formulation (PAR, paras. 2.1-2.2). Five sites were chosen, three in the greater Tunis and two in the Northwest region that were representative of the three varieties of sub-housing conditions in Tunisia (PAR, para. 2.3). 4. There were three physical components (PAR, para. 2.4). The upgrading component on four sites consisted of improvement and extension of infrastructure networks, provision of community facilities, and self-construction loans on about 400 infill plots. The sites and services component consisted of subloans to purchase about 1,600 developed plots and to construct dwellings on them. The revitalization of Hafsia was a pilot program which comprised the improvement of infrastructure (street pavement, water supply, sewerage, electricity and gas network), construction of commercial buildings and housing on part of vacant municipal land, and sale of the remaining serviced land to private developers (including hotels) with the profits funding the upgrading of about 47,000 m2 of housing in low-income areas. A technical assistance (TA) component was added to strengthen the project institutions, mainly ministries, since bilateral aid was used to assist the implementing agencies. 10 5. Project Cost and Financing Plan. The project cost was estimated at US$60.1 million, one third greater than Urban II. Although the foreign exchange component was estimated at 42 percent, the Bank's share in the total funding was 25 percent, still making it the largest single financier, followed by the Government (22.7 percent). The beneficiaries and private developers shared the remainder through cost recovery. The Bank loan was onlent through the Government (the Borrower) to: (i) various ministries for further onlendingand to public agencies (ARRU and utilities: 54 percent); (ii) the municipal fund, Caisse de PrEts et de Soutien des Collectivites Locales (CPSCL: 25.5 percent); and (iii) the housing finance institution, Caisse Nationale d'Epargne-Logement (CNEL: 19 percent) (PAR, para. 2.4). 6. On-Lending Rates. The on-lending interest rates to the various intermediaries used in the project were lower than the Bank rate (11.6 percent at loan signature) because the Government used a weighted average between its funds and the Bank's to lower the cost to beneficiaries, either households or municipalities. Of concern to the Bank at the outset was that lending rates were negative in real terms (PAR, para. 2.11). Implementing Agencies 7. ARRU was the main project executing agency and its statutes gave it broad powers to intervene in subprojects rapidly, including the right to expropriate the necessary land (PAR, para. 2. 7). The municipalities, including Tunis for the Hafsia component, delegated responsibility for implementing the project to ARRU. 8. The residential land development agency, Agence Foncibre d'Habitation (AFH), was responsible for the physical implementation of the sites-and-services components. A division was created in 1986 to implement the project, but the organization was not ready to adapt its procedures to the new type of subprojects promoted by ARRU, resulting in delays (PAR, para. 2.8). 9. CPSCL was merely an account managed by the oldest industrial development bank and it had no staff. It was chosen as the intermediary for a quarter of the Bank's funds because it was the official lending agency for the municipalities (PAR, para. 2.9). 10. CNEL was the major housing finance institution and although it did not need external funding since it relied mainly on contractual savers, its cooperation was needed to bundle the purchase of a serviced plot and the financing of construction into a single package, which was a new concept in Tunisia (PAR, para. 2.10). CNEL received limited incentives for its contribution as financial intermediary and this was reflected in the poor mortgage loan recovery. 11. Project Design and Risks. The main design problem was to finalize the Hafsia scheme which, by Board presentation, was not fully funded. This was solved with a dated covenant to establish a Special Account budget of the in the Municipality of Tunis (PAR, para. 2.11). The negative on-lending interest rates were seen as impairing project sustainability and efforts were multiplied to resolve the issue early on. Two implementation risks identified were the newness of the Housing Ministry and ARRU (PAR, para. 2.12). Securing the large tracks of land to start the project was also a concern. No mention was made, however, of possible II slippage in beneficiary targeting, of difficulties with cost recovery through the yet-to-be enacted frontage tax, and of the resistance to the new housing policy by the experienced agencies of the sector (AFH and CNEL). Project Implementation 12. It took ten years to implement the project, three more than estimated. Most original components were slow to start and to execute (two years delay on average). But the delay was mainly to accommodate three amendments aimed to use the surplus created by the US dollar appreciation around 1985. The first amendment allowed three additional sites covering 74 hectares to be serviced. Then the ARRU-led rehabilitation experience was successfully extended through a credit line approach to two governorates. Last, a US$7 million component allowed National Sewerage Authority (ONAS) to provide sewerage service to 19,000 dwellings on 72 sites which had failed to get priority before under the "Projet Sp6cial d'Assainissement des Quartiers Populaires" (PAR, para. 3.2). This second phase of the project showed a step-up in the degree of Bank delegation in the implementation process. This increased "wholeselling" reflected the Bank's confidence in the agencies' technical implementation capabilities. 13. The PCR noted a few limited procurement irregularities, but there were no major difficulties (PAR, para. 3.3). Although disbursements lagged the (optimistic) projections by 31 percent every year, disbursements picked up significantly in 1990 with 25 percent of the loan disbursed compared to 40 percent for the six previous years. The eventual disbursement profile was close to the standard for Tunisia (PAR, para. 3.4). 14. Because the purchase of land for sites-and-services components is often a cause of delay, the authorities thought of circumventing the difficulty by giving ARRU the power to expropriate the land necessary to carry out its mission (PAR, para. 3.5). To avoid long court battles over compensation, however, ARRU tried to purchase the land through direct negotiation with the owners. It was not successful in all cases and delays ensued, especially at Ettadhamen where the area involved was very large. 15. External Audits. Except for Agence Fonciere d'Habitation (AFH), whose 1983-85 accounts were heavily qualified, the external audits of the various participants, including each municipality, did not raise special issues except for the usual delays (PAR, para. 4.3). 16. Targeting. The project was designed to satisfy the housing needs of the urban poor. For serviced sites, priority was given to residents whose houses were demolished by the project. A 20 percent quota was reserved to CNEL savers and to members of a small Social Housing Fund (FOPROLOS). Targeted non-priority households, which were the majority of beneficiaries, had to have a monthly income of TD50-1 10 in 1981 Dinars (indexed on the guaranteed minimum wage), which was around the Bank-defined absolute urban poverty threshold (TD60). Beneficiaries were required to start construction within one year of plot delivery and resale, after 5 years, required municipality approval (PAR, para. 3.6). At appraisal, tests showed both upgrading and sites-and-services components to be affordable to between 85 and 92 percent of 39,500 targeted beneficiaries which made the loan mortgage recovery so much more likely (PAR, para. 3. 7). The PCR does not provide data on the results of this targeting, but ARRU is completing an evaluation which will shed light on the profile of beneficiaries. 12 Results 17. Hafsia Component. The Hafsia component had an uncertain start because the local cost funding was not identified until over a year after loan effectiveness. The Hafsia Special Account was eventually cofinanced by the Fonds National pour I'Am6lioration de l'Habitat (FNAH), the municipality of Tunis, the sale of land to developers, and cost recovery from the beneficiaries. Implementation was not without difficulties, however, mainly because civil work firms were initially reluctant to bid for contracts given the conditions (impossibility to bring heavy equipment into the narrow streets). Yet, five years after the project had been launched, the Hafsia component was extended to the entire Medina of Tunis based on the renovation success. Bank staff described the improvements to the area as having "completely transformed the social life in what was before a decayed place" (PAR, para. 3.12). While the renovation of the old city quarter was intended to provide better living conditions to the existing (poor) population, it attracted a non-negligible number of middle-class residents. Less than 250 tenants, however, were forced to resettle on other project sites. 18. Technical Assistance and Training. Not crucial to the project's success, the TA component was revised several times to meet the changing needs over the long implementation period. Assistance was provided to the Ministry of Housing for a study of urban planning in mid-size cities, for preparation studies for the Urban IV project, for a future transport project, and for the Municipal Development Project (Urban V), which aimed, inter alia, to give CPSCL greater administrative autonomy. Overall, training benefitted a large number of public servants (PAR, para. 4. 1) 19. The economic rate of return (ERR) is likely to be satisfactory. The estimated ERR ranged from 1 8 percent for the sites-and-services and the Hafsia components and 24 percent for the less costly upgrading component. The PCR did not provide the recomputed ERR. OED estimates that the actual rates are in this vicinity or better based on the rental values for housing, which are visibly higher than projected at the outset. Cost Recovery Performance 20. The cost recovery record of this project is not exemplary. The stated objective was to recover about 73 percent of total costs directly from beneficiaries with the balance also partially from them but indirectly through existing utility tariffs and local taxes that were to be introduced. While there was no problem with the utilities charging fees (notably for connections), taxing the beneficiaries for infrastructure improvements remains essentially to be done. This is one of the few (but key) project features which did not elicit "ownership" by the Borrower. The target date for introducing the "taxe de contribution des riverains" (herein the frontage tax), a property tax based on the street side length of each lot, was June 1986. The decrees allowing municipalities to levy this tax were not promulgated except for one of the six sites. The Bank did not push the issue for several years partly because it was aware of the political resistance to tax recovery from the beneficiaries. The Bank took a firm stance for the first time in 1991 when it asked both the Ministry of Interior to take measures to collect the frontage tax from the beneficiaries and the Housing Bank to start enforcing recovery procedures on default subloans, but with little success (PAR, paras. 4.4-4.6). 13 21. As of December 1992, none of the municipalities had started to reimburse CPSCL mainly because the latter had not prepared the loan repayment schedules. This hindered CPSCL's necessary move toward autonomy. On the mortgage lending side, the CNEL/Housing Bank (BH) eventually improved its loan collection, but it was still no more than 25 percent in 1990. At the time of this Audit, it was only 30 percent overall, but 50 percent for loans in Hafsia (PAR, para. 4.7). 22. Sustainability. The PCR listed four reasons in favor of the project's sustainability: (i) many components had been duplicated in the follow-up projects, with evidence of continuity; (ii) ARRU's experience also ensured the likely continuity of the rehabilitation programs; (iii) AFH and BH have started working together; and (iv) private developers could replicate some of the components. This is not disputable. Yet, project sustainability means that the project is capable of generating the current flow of benefits and to recover the recurrent expenditures and the cost of servicing the debt so that it does not affect public finances, and if desirable, funds can be recycled to finance other similar projects; in that sense, for Urban III, it is unlikely as long as cost recovery through loan repayments and frontage tax is low or non- existent (PAR, para. 4.10). The PCR acknowledged that reality. Conclusions and Lessons Borrower Performance 23. The Government took a series of decisions which provided a conducive environment to implement the project (PAR, para. 5.1). The Ministry of Housing was created to give more prominence to the new housing policy orientation. A specialized agency (ARRU) was created within one year of project identification which showed that the Government was serious about implementing the new policy of housing rehabilitation, demonstrated under Urban II. The appointment of a very competent staff to head ARRU was also a key decision in the success of the project. This gave an invaluable impetus to the launching of a large scale upgrading component. 24. The performance of other public agencies was not impressive (PAR, para. 5.2). AFH did not give priority to the low-cost plot development for a long time because it implied a radical change of its strategy. CNEL was also not keen to support an exception to mortgage application with prior savings. Its loan collection performance was, and still is, much worse than for its own portfolio. CPSCL was merely an account in a local financial institution without ties to the housing sector during the implementation period. Bank Performance 25. The Bank read correctly the Government's interest in promoting housing for the low- income population as a way to mitigate the adverse effect of urbanization. The Bank took the correct decision to go from a pilot program still under implementation (Urban II) to a full scale country-wide program to be carried out by a dedicated institution to be set up. The complexity of a project involving: two ministries,Ministry of Transport (MOT), Ministry of the Equipment (MOE); two financial institutions (CPSCL and Housing Bank); three agencies,(ARRU, AFH, ONAS); and indirectly the Ministry of Interior, Regional Planning, 14 and the District of Tunis should not be forgotten. It was a challenging task in terms of coordination and the delays that ensued are hardly surprising (PAR, paras. 5.4-5.5). The project took an innovative turn when the Bank accepted the Government's request to modify the upgrading component into a line of credit. Although originally a minor component, the solution to deal with the intricate problems of the old Medina of Tunis was also innovative. 26. A clear failure of the Bank was in enforcing the recovery through the frontage tax. Since this mode of recovery accounted for almost 30 percent of the project cost, an early firmer stance was warranted. The Bank was also unsuccessful in strengthening CPSCL. As the lending window to the municipalities, CPSCL remained what it was throughout the project, a lending window (PAR, para. 5.6). Project Ratings 27. The Audit assesses the outcome of the project as satisfactory. It delivered more serviced plots, more rehabilitation housing and more sewerage connections than planned, and the Medina renovation was an unexpected success. The sustainability of the projet is rated, however, as uncertain because of the poor recovery record, and the institutional development is rated as moderate because of the slow pace of reform at CPSCL. Bank and Borrower performance is rated as satisfactory (PAR, para. 5. 7). The PCR-based ratings were the same except for sustainability,which was rated as likely. Lessons Learned 28. An important lesson is that complex projects can work if the implementation agency's management and staff are given sufficient autonomy to take the actions needed to launch a new activity. Tackling many urban issues in parallel is tempting. On balance, however, the transaction cost and risks of multicomponent projects (delays, supervision) suggest that simpler projects have merits. Even more so that urban projects are generally processed in series and a component with a lower chance of success can be postponed to a follow-up project. OED concurs with the PCR and the Borrower's recommendation of making a single agency responsible for coordinating all the operators (PCR, para. 5.04 (a) and PAR, para. 5.9). 29. When a great number of civil work contracts are expected in a project, administrative ways to alleviate the procurement management burden on both the Borrower and the Bank should be found. This could be achieved by bundling small contracts into coherent sets and delegating the process on the model of the SOEs on the disbursement side (PAR, para. 5.10). 30. The PCR recommended that to avoid bottlenecks at project launch, the necessary land be purchased prior to start. This is also a sensible recommendation although the Government's commitment to the project is often firmed up not before, but at about the same time the Bank loan is submitted to the Board. Rather than making the process more rigid, it is more realistic to require a better risk analysis to identify the likely causes of implementation delays (PAR, para. 5.11). 15 1. Background 1.1 Despite its small size in terms of total population (6.6 million according to the 1980 census prior to project appraisal), Tunisia had to deal with the same problems as larger countries faced with rapid urbanization. With 55 percent of urban population in 1980, the urbanization rate was high and expected to increase. The rural-to-urban migratory flows put "great pressures on urban housing and led to a deterioration of the urban environment in the outskirts of cities and within the crowded traditional quarters (Medinas) of many cities." (SAR, para. 1.01). Tunis had the most serious housing and upgrading problems, both in terms of quantity and importance. Although the total urban housing stock had increased by 30 percent between 1975 and 1980 to reach 605,000 units, the demand out-paced the public sector's ability to control urbanization. This situation provided the rationale for expanding Bank assistance in the sector from the pilot Urban II project to a large scale sites and services project in and around Tunis and in several cities. 1.2 The urban policy formulation and implementation responsibility was shared between the Ministries of Housing, Equipment, Interior, and National Economy (the first two were later merged). The Government's response to the shelter needs was to create its own real estate agency, Societe Nationale Immobili&e de Tunisie (SNIT), whose programs where heavily subsidized followed by a land development agency. Agence Fonci&re d'Habitation (AFH), and a housing savings fund, Caisse Nationale d'Epargne-Logement (CNEL). More recently, partly as a result of the Urban 11 project, the Government had accepted the concepts of (i) upgrading instead of demolishing squatter settlements, (ii) developing low-cost sites and services projects, and (iii) recovering infrastructure costs although the latter consensus proved less solid. Another agency, Agence pour la Rehabilitation et la Renovation Urbaine (ARRU), was created in August 1981 to be the spearhead in implementing the new policy.  17 2. Project Description Objectives 2.1 The objectives were complex and interpretations varied over time. According to the SAR, the project objective was essentially to increase the supply of low-cost housing affordable to low-income urban dwellers. The "Objectives" section in the SAR stated four distinct, though connected, objectives, however. They were: (i) "to design and implement projects providing better shelter and improved urban services for low-income populations; (ii) to redress inequities in housing policy by reaching low-income families which currently cannot afford housing with adequate urban services as provided by the formal sector; (iii) to strengthen the institutional framework of agencies involved in the urban and housing sector, and to ensure better coordination between them; and (iv) to encourage a national approach to housing planning and policy formulation." 2.2 As an alternative to these broad objectives, the PCR suggested a list of practical sub- objectives for the housing component (against which the project's performance was evaluated (PCR, para. 3.1): "(i) To concentrate low-income housing activities of local authorities on the supply of serviced land; (ii) to introduce private sector financing for low-income housing through existing private sector financial intermediaries: (iii) to establish new procedures for financing and construction of low-income housing; (iv) to maintain standards for low-income housing to ensure continuity of the reformed process; and (v) to strengthen urban management capacities of local governments." 2.3 Five sites were initially retained, three in greater Tunis and two in the Northwest region. In Tunis. they were Hafsia in the Medina, Kram Ouest and Ettadhamen (the later two sites being in the then distant suburbs); outside Tunis they were Zghadia in Jendouba and Chrichri in Le Kef (both West of Tunis). They were representative of the variety of sub- housing conditions met in Tunisia: low density, illegal land tenure and inadequate infrastructure (Ettadhamen), squatter settlements with deteriorated living conditions (Kram Ouest); over-crowding in an old Arab city (Hafsia). The choice was dictated by the intent both to cover the country better and to further test the efficiency of the upgrading approach launched under Urban II, but in more varied conditions. Components and Financing Arrangements 2.4 There were as many physical components as Bank approaches to low-income settlements (e.g., rehabilitation of existing but deteriorating housing stock, sites and services on scattered land, and upgrading of infrastructure in fast-growing urban areas). The upgrading component on four sites consisted of improvement and extension of infrastructure networks, of providing community facilities, and of self-construction loans on about 400 infill plots. The sites and services component consisted of subloans to purchase about 1,600 developed plots and to construct dwellings on them. The revitalization of Hafsia was a pilot program which comprised the improvement of infrastructure (street pavement, water supply, sewerage, electricity and gas network), construction of commercial buildings and housing on part of vacant municipal land. sale of the remaining serviced land to private developers (including hotels) with the profits funding the upgrading of about 47,000 n2 of housing in the low- 18 income areas.' A technical assistance component was added to strengthen the project institutions, mainly ministries since, for cost reasons, bilateral aid was used to assist the implementing agencies. 2.5 The project cost was originally estimated at US$60.1 million with a foreign exchange component estimated at 42 percent and local taxes accounting for 20 percent. Physical contingencies were set at 7.6 percent and price contingencies at a substantial 20.5 per cent to cover the 7 years of planned implementation. The financing plan gave the Bank the largest share with 25 percent, followed by the Government (22.7 percent); the beneficiaries and private developers shared the remainder, mostly through cost recovery. 2.6 The Bank loan was to be onlent through the Government to the municipal fund, Caisse de Prts et de Soutien des Collectivites Locales (CPSCL) (25.5 percent), to the housing finance institution, Caisse Nationale d'Epargne-Logement (CNEL) (19 percent), and 54 percent to various ministries for further onlending to subborrowers and to public agencies (ARRU and utilities). The onlending interest rate to the various intermediaries used in the project was lower than the Bank rate (1 1.6 percent at loan signature) because the Government, in an effort to lower costs to beneficiaries, used a weighted average cost between the Bank rate and the agencies' rates. The resulting rates were 7 percent through CNEL and 7.9 percent through CPSCL. These rates were negative in real terms when compared to inflation (8.9 percent in 1981). Still they were much greater than the rates normally charged by the two institutions (4.5 and 2.0 percent, respectively) although well below rates for medium-term lending by commercial banks. The Government had agreed at negotiations to review, soon after loan effectiveness, the rate structure in the housing sector with a view to make the rates positive by the end of 1987 (SAR, para. 3.10). However, this did not happen mainly because the authorities did not want too much of a gap between the normal activities and the Bank- supported subprojects. Implementing Agencies 2.7 ARRU was the main project executing agency and its statutes gave it broad powers to intervene in subprojects rapidly. It was created in 1981 under private law, with the attributes of a land bank (i.e., right to expropriate and to develop land).2 The new head of ARRU was very effective in creating a favorable momentum right at project launch. Its financial autonomy was helped in great part by its access to commercial bank credits. Four months before loan effectiveness, ARRU was fully staffed, partly due to its ability to pay competitive salaries, which is a rare enough occurrence among project units for it to be mentioned. ARRU was given broad responsibilities for a new institution. All five municipalities, including Tunis for the Hafsia component, delegated responsibility for implementing the project at each site to ARRU. This greatly simplified the administration of this complex project. Coordination with the various other agencies was achieved through a Technical Committee, but this proved to be insufficient to ensure total cooperation. 1. A detailed description of the physical components is available in the SAR, Annex 1. 2. Respectively Unite Centrale de Projet and Socit6 d'Economie Mixte pour I'Am6nagement de Tunis. It absorbed the project unit set up to implement Urban II and an agency for developing Tunis 19 2.8 Agence Fonci&re d'Habitation (AFH), the residential land development agency, was responsible for the physical implementation for the sites and services components. A division was created in 1986 to implement the project, but the organization was not ready to adapt its procedures to the new type of subprojects promoted by ARRU. Treasury-type difficulties forced AFH to focus on its core business rather than its contribution to the Project, and the lack of coordination between AFH and CNEL more than eight years into project implementation was a constant impediment.' Some improvements were obtained when a new head was appointed in 1989, the same who launched ARRU at the beginning of the project. It is significant that one of the first changes implemented was the introduction of cost accounting for each subproject. This was made necessary by the changing operation conditions from a quasi-subsidized situation' to more complex upgrading projects. 2.9 CPSCL, the municipal fund, was to be strengthened under Urban II, but little progress had been achieved. It was still merely an account managed by the most experienced development bank, Banque pour le D6veloppement Economique de Tunisie (BDET), and it had no staff. There was little progress during project preparation on the municipal issues (local taxation, financing municipal operations, and strengthening municipal administration). Still CPSCL was chosen as the intermediary to onlend a quarter of the Bank funds. 2.10 CNELI, with its attractive contractual savings scheme which guaranteed a low interest mortgage loan after a saving period, CNEL became the major housing finance institution in the space of a few years. This gave CNEL a monopoly position in the sector and little incentive to try an approach where mortgage loans could be obtained without prior savings. Yet, it was a key agency of the Government and the recent evolution of the housing policy required a diversification of the financial instruments offered to dwelling buyers. The bundling of the purchase of a serviced plot and the financing of construction into a single package was new in Tunisia and it required the cooperation of AFH and CNEL. The incentives for CNEL were to receive one percent on disbursements and two percent on recoveries which was substantial for the sole function of transferring funds. The penalty was to absorb five percent of the unrecovered loan amounts (SAR, para. 3.08 (iii)), which was an adequate incentive, if applied. Project Design Issues 2.11 The main design problem was to finalize the Hafsia scheme which, by Board presentation, was not fully funded because all the counterpart funds had not been identified (SAR, para. 4.05). The establishment of a Special Account in the Municipality of Tunis for the upgrading of Hafsia was made into a dated covenant. The fact that the on-lending interest rates were negative in real terms was viewed by the Bank as impairing project sustainability and efforts were multiplied to resolve the issue early on. 2.12 Two implementation risks identified were the newness of both the Housing Ministry and of ARRU (SAR, paras. 6.06-6.07). Trusting most of the project execution to an unproven 3. Back-to-Office Report, August 20, 1991. 4. For most of its land development operations. AFH had been relying on medium-term advances from its buyers. 5. CNEL was transformed into the Housing Bank in 1989 and the two names are used interchangeably in the report. 20 agency (ARRU) was a substantial risk to take even with the quality of sector dialogue found in Tunisia. The concerns were not founded, however. The securing of the large tracks of land involved in the project was also a concern during preparation (para. 3.9). No mention was made, however, of possible slippages in beneficiary targeting, of difficulties with cost recovery, and of the half-hearted ownership by the experienced agencies (AFH and CNEL) of the sector, which proved to be real risks. 21 3. Project Implementation and Physical Results Time Overruns and Use of Savings 3.1 For a time the new Ministry of Housing did not have the administrative capacity to implement adequately its plans. Given the great number of ministries and agencies involved, coordination was a practical issue. While ARRU was the lead implementing agency, the coordination among ministries was achieved through the same Inter-Ministerial Committee set up under Urban II. For some agencies, in particular AFH and CNEL, the project was a marginal activity which elicited limited "ownership". Monitoring relied on semi-annual progress reports prepared by ARRU and on frequent supervision missions. In the end, the physical targets were met, some of them even exceeded, because of loan savings, but the project required three more years to be completed than scheduled (43 percent time overrun). Time overruns were partly due to the use of savings as explained below but the PCR acknowledged that most original components were slow to start and to execute, by two years on average. 3.2 Three amendments aimed at using the surplus created by the appreciation of the US dollar around 1985 instead of facing the likely cancellation of US$8.6 million. Three additional sites covering 74 hectares were serviced. Taking note of the earlier success of ARRU, the rehabilitation experience was extended through a credit line approach to the Bja and Gafsa Governorates. Last, a new US$7 million component implemented by ONAS (the "Projet Sp6cial d'Assainissement des Quartiers Populaires"), provided sewerage system in 72 sites (which had low priority before), with 190 kms of pipes to service 19,000 dwellings (or about 100,000 persons). The Audit agrees that this use of savings was consistent with project objectives. Procurement and Disbursements 3.3 The Project attracted a large number of contracts because upgrading and rehabilitation involved many small tasks, which justified procurement on the basis of local competitive bidding (LCB) or force account (Hafsia component). The PCR noted a few limited procurement irregularities, but overall there were no major difficulties. ARRU, which was a new institution at project launch, handled the complex procedures efficiently. The PCR suggested correctly that a consolidation of the multitude of small contracts into homogeneous sets would have simplified project management for both the Borrower and Bank supervision missions. 3.4 The Loan disbursement percentages for each category reflected the estimated foreign exchange component of each eligible item. They were originally 45 percent for civil works and self-construction loans, 30 percent for sites and services, and 80 percent for technical assistance. As shown in the statistical introduction to this report, actual disbursements lagged projections by 31 percent every year. Eventually the actual disbursement profile was close to the standard for Tunisia. Although the Loan Agreement onto categories in demand, Bank staff complained in a 1988 Aide-Memoire that only 25 percent of the Loan had been disbursed despite two amendments. Partly to boost disbursement, Bank staff had recommended the unusual approval of a second amendment request when only US$4.1 million had been disbursed. At one point in 1990, the sewage component was 60 percent committed, but only 15 percent disbursed. Disbursements picked up significantly in 1990 with 25 percent of the 22 Loan disbursed compared to 40 percent for the six previous years. There were payment delays to the enterprises partly due to the municipalities' cumbersome approval process. CPSCL also contributed to slowing disbursements by taking no less than three months to pass on the Loan funds. As late as 1989, funds forwarded through CPSCL did not reach ARRU before six months,6 which compounded ARRU's problem of working capital shortages to launch its ongoing program. 3.5 Land Acquisition. The land required by the project was a concern during preparation because it was known from experience that land often produced bottlenecks in implementing urban housing projects according to schedule. The major difference with the past was the power given to ARRU to expropriate land to carry out its mission (SAR, para. 3.18). Because compensation for expropriations can be challenged by the owners for many years in Tunisian courts, ARRU tried to purchase all the needed land through direct negotiation with the owners. Yet, this "expropriation A l'amiable"' approach was not successful in all cases and court actions, by unwilling owners, created delays, especially at Ettadhamen where the area involved was very large. Beneficiary Targeting 3.6 The project was designed to satisfy the housing needs of the urban poor and the selection of beneficiaries was given particular attention. For serviced sites, which would house the largest number of beneficiaries targeted by the project, a Selection Committee chaired by the municipality and including AFH and CNEL, prepared initial lists. Priority was given to residents whose houses were demolished by the project (upgrading component). A quota of 20 percent was reserved to CNEL savers and the members of a small social fund (FOPROLOS). Non-priority beneficiaries had to have monthly incomes of TD50-1 10 in 1981 Dinars (indexed on the guaranteed minimum wage), which was very close to the Bank-defined absolute urban poverty threshold (TD60, SAR, para. 1.07). Beneficiaries were required to start construction within one year of plot delivery and its resale before 5 years was constrained. 3.7 Through its several amendments, the project continued to focus on lower than middle income (TD50-150 per month or 1-2.5 times the minimum guaranteed salary).' Of the originally estimated 10,700 beneficiaries to occupy the sites and services, 62 percent were expected to be urban poor; for the 28,800 beneficiaries of upgrading, the target was 31 percent (SAR, para. 4.09). 3.8 Mortgage Granting. CNEL was reluctant to lend outside its contractual savings scheme which makes accumulating savings a prerequisite for obtaining a mortgage loan. Eventually CNEL granted 2,444 loans to eligible households. Relative to the number of beneficiaries this is, however, a small figure. In 1990, ARRU asked the Bank to revise the ceiling for self-construction loans which had not been revised since 1987 while inflation measured by consumer prices had increased by 23 percent. However, the move was late in the implementation cycle and was followed by less than 500 new loans to be granted by CNEL. 6. Supervision Report, December 7, 1989. 7. The expression is quoted from the PCR, para. 5.07. 8. SMIG (Salaire Minimum Garanti) 23 3.9 Affordability. Given the target population, affordability was a key consideration. The estimates were made assuming a relatively high down payment (20 percent) and a debt- servicing burden no greater than 15 percent for upgrading and 25 percent for sites and services. The tests showed both upgrading and sites and services components to be affordable to between 85 and 92 percent of the 39,500 targeted beneficiaries (SAR, para. 4.08). Given the thoroughness of this exercise, it is difficult to comprehend the extremely low mortgage loan recovery rates achieved by CNEL (PAR, para. 4.5). 3.10 The PCR does not provide actual figures, but ARRU is currently completing a detailed self-evaluation of Urban III which should shed light, inter alia, on the profile of beneficiaries after project completion. However, the summary of the impact in Hafsia (see Annex) showed that beneficiaries were reluctant to answer questions about their income, a reality which makes ex post analysis more difficult. 3.11 Legalization of Ownership. A socially significant by-product of the project was to provide security of land tenure to the squatters whose dwellings were rehabilitated. This was achieved by issuing new titles to plot holders after they paid for the infrastructure improvements (e.g., connection fees). Rehabilitation of the Tunis Medina (Hafsia Component) 3.12 The Hafsia component was a very successful pilot rehabilitation scheme. It had an uncertain start. The financial sources for the local costs of the component had not been identified during loan negotiations. Over a year after effectiveness, it was agreed that the Hafsia Special Account would be funded on the Tunisian side by the housing improvement fund, Fond National pour I'Amelioration de l'Habitat (FNAH), the Municipality of Tunis, the sale of land to developers, and eventual cost recovery from the housing beneficiaries (through down payments and mortgage loan servicing). The Hafsia rehabilitation component was such a success only five years after the project had been launched that it was extended to the entire Medina of Tunis. A 1990 supervision report described the improvements to the area as having "completely transformed the social life in what was before a decayed place". Middle-class professionals were returning without displacing poor people. Less than 250 tenants living in precarious houses were resettled, most in Ettadhamen, the largest project site. A recent study measured the medium-term impact of this successful component (see Annex). 3.13 There were some difficulties in closing the Hafsia Special Account as it involved some clearing of debts by the various participants. For instance, ARRU made a profit out of the Hafsia subproject, but it delayed a TDI80,000 payment to the Tunis Municipality. In 1989, a Bank supervision report admitted that the account was "not performing normally, blocking the rehabilitation process".' The following year, there was an argument between the Housing Bank and ARRU on the funding of the Hafsia account. In 1992 the Bank requested an amendment to the Hafsia contract to replace the Fond National pour I'Am6lioration de l'Habitat (FNAH) with a Planning Ministry contribution (TDI50,000) while keeping co- financing from ARRU at TD400,000 and from the project at TD450,000. All these incidents are understandable given the pilot nature of the financial and administrative set-up. 9. Supervision Report, March 13. 1989. 24 Economic Rate of Return 3.14 The economic rate of return (ERR) is likely to be satisfactory. At appraisal the ERR estimates ranged from 18 percent for the sites and services and the Hafsia components and 24 percent for the less costly upgrading component. The PCR did not provide the recomputed ERRs. OED estimates that the actual rates are in this vicinity or better, based on the rental values for housing, which are visibly higher than projected at the outset. 25 4. Institutional Development Technical Assistance and Training 4.1 Technical Assistance. TA was originally a marginal component mainly because the Government was reluctant to use foreign borrowed funds for that purpose. During project preparation, TA was included in the project cost to ensure its last resort financing by the Bank. Eventually, the funds earmarked for TA were extensively used which was a new trend, at least in urban lending. The TA component was reallocated several times to meet the changing needs of various sector ministries over the ten-year implementation period. Some assistance was provided to the Ministry of Housing for a study of urban planning in mid-size cities (1990). Pre-investment funds for preparing studies for Urban IV project and a future transport project were also included. In 1992, TD400,000 were allocated for a study to be carried out by the Ministry of Interior to prepare the Municipal Development Project (Urban V) which aimed, among other goals, to give CPSCL greater administrative autonomy. In 1993, Ministry of Transport requested the transfer of funds earmarked for studies to the acquisition of computer equipim-ent. 4.2 Training. ARRU also made a concerted use of the loan funds earmarked for training. Overall, training benefitted a large number of public servants. The PCR does not comment on the success or failure of this component. 4.3 Accounting and External Audit. Accounting, or lack of it, at two agencies revealed deep institutional problems. The accounts for each component were kept by each municipality under the Tunisian public accounting plan, then audited by the Contr6leur G6n6ral des Finances. The accounts for AFH, ARRU, and CPSCL were kept under the commercial accounting plan. Although CPSCL was merely an account managed by an experienced development bank (BDET), its accounts for 1985-87 were still not prepared as of end 1988, let alone the external audit of these accounts. The Bank requested these audits to be prepared before the end of 1989 and agreed that the Contr6leur G6n6ral des Finances in the Ministry of Finance would carry out the task. The audit for the period 1975-91 was very good with only TD2,300 of discrepancy over the entire period, but it was nonetheless qualified because the debt-service tables of the municipality subloans were still not ready. AFH's 1985-87 accounts were heavily qualified and this led to a management shake-up (para. 2.8). The accounting system had to be changed retroactively to 1982. CNEL's accounts were not audited since its intermediation was a minor activity. Cost Recovery 4.4 Housing projects such as Urban III are normally sustainable either directly (through mortgage loan repayments) or indirectly (through taxes). From either view point, the cost recovery record of this project is not exemplary and the PCR acknowledged the failure without elaborating on it as the Audit does here (PCR, para. 8.01). It was expected that net cash flows would stay positive after all project funds had been disbursed (1990) and would reach US$5.2 million by 1992: "This surplus would allow the replicability of the project on future sites." (SAR, para. 3.11). The stated objective (Issues Paper, March 1983) was to recover about 73 percent of total costs directly from beneficiaries and the balance also partially from them but indirectly through existing utility tariffs and local taxes to be introduced. While there was no problem with the utilities charging fees (notably for connections), as both they and their clients were used to the practice, the concept of taxing the beneficiaries of infrastructure improvements remained essentially an untested concept. 26 4.5. The negotiations had targeted June 30, 1984 as the date for introducing the frontage tax, a property tax based on the street side length of each lot which was supposed to help repay half of the CPSCL loans (SAR, para. 4.04). This is one of the few (yet, unfortunately key) project features which did not elicit "ownership" by the Borrower. In 1987, the Ministry of Housing informed the Bank that the municipalities of Le Kef, Jendouba and Ettadhamen had decided to apply the frontage tax. Yet, the decrees allowing municipalities to levy this tax were not promulgated. A 1989 supervision report acknowledged that cost recovery was "still on hold", but the Bank did not push the issue for the next two years partly because it was aware of the political resistance to this tax."o 4.6 The Bank took a firm stance for the first time in 1991 when it asked both the Ministry of Interior to take measures to collect the frontage tax from the beneficiaries and the Housing Bank to start enforcing recovery procedures on defaulting subloans. In 1992 ARRU launched an effort to mobilize the concerned municipalities into accelerating the introduction of the frontage tax . To this date, little progress (Part II of the PCR mentions "one decree" introducing the tax on one site) has been made on this issue partly because of its political nature and of the known resistance of many Tunisians to any form of property tax. In the meantime, the ONAS amendment has allowed partial recovery by charging the beneficiaries a connection fee for sewerage." 4.7 As of December 1992, no municipality had yet started to reimburse CPSCL mainly because the latter had not prepared the loan repayment schedules. This hindered CPSCL's necessary move toward autonomy. On the mortgage lending side, the Housing Bank (formerly CNEL) eventually improved its loan collection, but it was still no more than 25 percent in 1990. In June 1995, it was still 30 percent overall, but 50 percent for loans in Hafsia. By any standard, this outcome is dismal. Bank projects in housing finance in various countries have shown that arrears over 5 percent are rare and that collection is usually successful when the threat of repossessing the house is raised. The performance of CNEL/Housing Bank is all the less defensible in that collection on its own mortgage portfolio meet acceptable standards. 4.8 The SAR displayed a Recovery Table (p. 17) broken down as follows: loan repayments (10.7 percent), direct sales (26 percent), utilities charges and tariffs (11.2 percent), frontage tax (29.6 percent), and not recovered (22.5 percent). The PCR did not update this table. OED estimates that the "not recovered" category has increased to over 50 percent because the loan recovery has been poor and the frontage tax has been introduced in only one of the six sites according to Part II of the PCR drafted by the Borrower (PCR, para. 5.03 (a)). This shortfall had to be compensated by an added fiscal burden. The PCR claims that Urban III behaved the same as Urban II on this score, and that the lessons of the past were not taken into account in its design (PCR, para. 12.04); in fact, Urban 11 was still being implemented and not yet at the cost recovery stage when Urban III was appraised and negotiated. The problem was more the reluctance to enforce Bank remedies than a faulty design. Indeed, the loan recovery rate has been poor despite planning for up to five percent of the credit risk to be absorbed by CNEL (SAR, para. 3.07). It is not clear whether the Government, which was bearing the credit risk remainder, enforced this clause. The CNEL officials met during the Audit were not concerned by their weak performance and its contrast with the much better recovery rate obtained on their own mortgage loan portfolio. 10. The January 1990 Supervision Report is, for example, typical of the absence of the issue. I1. Letter to the Ministry of Planning, June 26, 1989. 27 4.9 The project involved other limited subsidies. The Bank loan was onlent at rates below the Bank rate. For example, ARRU received TDI million in 1991 as an operational subsidy which contributed to make it profitable. It has also been exempted of the value added tax (VAT) since its creation while most recently it was competing with AFH, Socit6 Nationale Immobili6re de Tunisie (SNIT) and private developers, all burdened with VAT. The defaults on the subloans by the Housing Bank have been also absorbed by the Government and the five percent penalty has not been enforced. 4.10 Sustainability. The PCR listed four reasons in favor of project sustainability: (i) many components had been duplicated in the follow-up projects, with evidence of continuity; (ii) ARRU's experience ensured the likely continuity of the rehabilitation programs; (iii) AFH and BH have started working together; and (iv) private developers could replicate some of the components. This is not disputable. Yet, project sustainability means that the project is capable of generating the current flow of benefits and to recover the recurrent expenditures and the cost of servicing the debt so that it does not affect public finances, and if desirable, funds can be recycled to finance other similar projects; in that sense, for Urban III, it is unlikely as long as cost recovery through loan repayments and frontage tax is low or non- existent. The PCR acknowledged that reality.  29 5. Conclusions and Lessons Learned Borrower Performance 5.1 In spite of its poor record on cost recovery, the performance of the Borrower is rated as satisfactory. The Government took a series of decisions which provided a conducive environment to implement the project. The Ministry of Housing was created to give more prominence to the new housing policy orientation. A specialized agency (ARRU) was created within one year of project identification, a step which showed that the Government was serious about implementing the new policy of housing rehabilitation then being demonstrated under Urban II. The appointment of a very competent staffl2 to head ARRU was also a key decision in the project's success. This gave an invaluable impetus to the launching of a large scale upgrading component. 5.2 The performance of other public agencies was not impressive. To date, the collection of improvement and self-help construction loans granted under the project by CNEL is much worse than for its own loans. AFH was not interested in changing its modus operandi although it created a special division to handle the project. This risk was known in advance, but little could be done short of carrying out a major reorganization, which eventually took place during the preparation of the Urban V project (1987). Although less than 10 years separate the creation of AFH (1973) and ARRU (1981), the contrast in efficiency is dramatic. It reflects and indicates to some extent the rapid change in Government policy in the late 1970s and early 1980s which gave sector agencies broader means to carry out their mandate. 5.3 ARRU must also be commended for the unusual initiative among implementing agencies to carry out a combined audit and impact study of its performance under the Urban III project. Four local consulting firms (one for two sites) are soon to complete their analyses. Bank Performance 5.4 Bank performance is rated as satisfactory. The Bank was heavily involved in a developmental strategy in housing in Tunisia since the late 1970s involving processing a series of projects to deal with all the key aspects of urban policy. The project followed along the lines of Urban II and built on the institutional framework developed thus far. It maintained the two main orientations: to continue the upgrading work in Tunisia and to introduce both upgrading and sites and services in secondary cities. The project involved two ministries (MOT, MOE), two financial institutions (CPSCL, Housing Bank), three agencies (ARRU, AFH, ONAS) and indirectly the Ministry of Interior, the Ministry of Transport, Regional Planning, and the District of Tunis. It also had all kinds of components, each one representing a different approach. This complexity made coordination a challenging task and the ensuing delays are not surprising. 5.5 The project took an innovative turn when the Bank accepted the Government's request to make the upgrading component into a line of credit. Although by then the Bank had processed several housing finance projects worldwide, it was one of the first times it had shifted from pre-identified subprojects to a credit line approach for upgrading activities. 5.6 A clear failure of the Bank was in enforcing subloan recovery through the frontage tax. Since this mode of recovery accounted for almost 30 percent of the project cost, an early firmer stance, was warranted. The Bank was also unsuccessful at strengthening CPSCL-a 12. He is now the current Minister of Public Works and Housing. 30 lending window to the municipalities-it remained a lending window throughout the project. Given the political sensitivity of financing municipalities, there was reluctance among the Tunisian authorities to allow greater autonomy to CPSCL. Eventually the consensus shifted and another Bank project helped transform CPSCL into a municipal bank. Ratings 5.7 The Audit assesses the outcome of the project as satisfactory. It delivered more serviced plots, more rehabilitation housing and more sewerage connections than planned, and the Medina renovation was an unexpected success. The sustainability of the projet is rated as uncertain because of the poor recovery record, and institutional development as moderate because of the slow pace of reform at CPSCL. Bank performance and Borrower performance are judged satisfactory. The PCR-based ratings were the same, except for sustainability which was rated as likely on the basis of incomplete financial data. 5.8 The Hafsia component is deemed as highly satisfactory because it exceeded the expectations that were placed in this pilot operation. A preliminary study (see Annex) and a site visit by the Audit have confirmed its benefits both in improving the housing stock and in reviving the social status of an historic area. Especially encouraging, ARRU succeeded in doing rehabilitation works below cost despite the difficult conditions encountered in the old city.14 With hindsight, it is tempting to portray it as the obvious thing to do. But is worth recalling that some participants were skeptical-and accordingly less than cooperative-at the start and even during implementation of the project. Lessons Learned 5.9 An important lesson is that complex projects can work if the apex implementation agency is capable and is given sufficient autonomy to take the actions needed to launch a new activity. The Borrower, in Part II of the PCR, made a similar observation (para. 5.04 (a)). Tackling many urban issues in parallel is tempting. On balance, however, the transaction costs and risks generated by multi-component projects, such as Urban III, suggest that a simpler approach has merits; this is even more so because projects are generally part of a program approach and a component with a lower chance of success can be postponed to a follow-up project. The PCR recommended that less complex projects be designed in the future (para. 12.02). OED concurs with this recommendation. 5.10 When a large number of civil work contracts are expected, administrative ways to alleviate the management burden both on the implementing agencies and on the Bank should be found, e.g. by bundling small contracts into coherent sets and delegating the process on the model of the SOEs on the disbursement side. 5.11 The PCR recommended that to avoid bottlenecks at project launch, the necessary land be purchased beforehand. This is also a sensible recommendation, although, often the Government's commitment to the project is firmed up at about the same time that the Bank loan is submitted to the Board. Rather than making the process more rigid, it would be more realistic to require a better risk analysis to identify the likely causes of delay in implementation. 13. Although operations indicated that reforming the CPSCL was not a specific objective of the Project, it was implicitly covered under the third objective listed in the SAR (para. 2.01) of "strengthening the institutional framework of agencies involved in the urban and housing sectors". 14. The narrow streets deterred contractors from bidding for the works. 31 Annex Impact of the Hafsia Component 1. A recent study" shows the medium-term economic and social benefits and drawbacks of the pilot rehabilitation component of the historical Medina district in Tunis. As the execution of this component was considered a success by both the professionals and the general public, it was useful to take stock of the experience several years later. Progress, or lack of it, was measured against the objectives. The social objectives included the identification of the housing with the relocation of displaced people in the same area as much as possible, minimum equipment of each dwelling with running water, kitchen and lavatory, creation of community facilities (clinics, public baths, kindergartens, etc), upgrading of street and infrastructure networks. The objectives economic included: increased welfare reflected in higher rents for the rehabilitated dwellings, increased employment through commercial and handicraft activities, and civil works to be carried out as much as possible with local labor, construction of new housing for middle-income households able to support the costs. 2. The purchasers of new housing in Hafsia were highly educated high level professionals mostly between 36 and 50 years of age, in large part because of the Housing Bank's selecting process. The beneficiaries had themselves selected Hafsia because 62 percent could walk to their place of work and because they were familiar with the area, with 56 percent having rented in Hafsia or surroundings for 1-7 years prior to buying. The revenues of this group, the majority civil servants, has not changed significantly since the project. The monthly debt repayment averaged TD 137 or 29 percent of the average income. 3. The new tenants of rehabilitated housing in Hafsia were mostly under 50, had secondary education, and were either mid-level professionals or heads of small handicraft firms. Thirty percent moved from Tunis suburbs and another 30 percent from other cities (Kairouan and Sfax). The location was selected for the proximity to work. The monthly rent was between TD 120 and 250, but without income data the relative burden was not assessed. 4. The group of Hafsia residents who borrowed to rehabilitate their dwellings comprised older household heads (54 percent over 50 of which half of them retirees), who are less educated than the first two groups. Their main reasons for carrying out the rehabilitation were the availability of low-interest loans and the encouragement from ARRU and ASM. The prospect of adding value to their assets came next. An unknown percentage of them rehabilitated without loans; most of them were above 50 and with either primary education or none; 55 percent did not work; and 40 percent of those working received daily wages thus explaining their ineligibility for bank loans. The greatest proportion (80 percent) of all the groups (with or without) interviewed walked to their work place. The improvements (assessed at TD3,000-12,000) were financed with savings, contributions from the extended family, and rental of the extensions. 5. Twelve households were moved out and relocated in the area. Household heads were educated (60 percent secondary schooling, 20 percent university), mostly below 50 years of 15. "Etude de l'mpact Social et Economique du Projet Hafsia" by Association Sauvegarde de la Medina de Tunis and Harvard University, Graduate School of Design, March 1994. 32 Annex age and all working in the service sector. They paid between TD70-120 per month toward debt service which is lower than the lowest rents paid by the above tenants mentioned in para.3 above. 6. Some dwellers were temporarily relocated to municipality-owned dwellings for later relocation on the Urban III site of Ettadhamen, but they refused to move out when offered serviced plots. They had little education, and held low-level jobs in the administration or the service sector in the area. 7. The households who accepted to buy plots in the Douar Hicher settlement in Ettadhamen, a suburb of Tunis, were slow to move on the site since only 2 of a sample of 12 were residing in their constructions. Although former tenants, they have not been motivated to move because the loans for self-construction were too small (TD2,000) and they have to take public transportation to their unchanged work places near Hafsia. The sampling revealed that most of the housing offered by private developers was transformed into commercial outlets (mostly cloth making boutiques).    rut D> (DO 00- (J0 >0