Document of The World Bank Report No: ICR00002201 IMPLEMENTATION COMPLETION AND RESULTS REPORT (IDA-40030 IDA-40031) ON A CREDIT IN THE AMOUNT OF SDR102.70 MILLION (US$155.0 MILLION EQUIVALENT) TO THE THE UNITED REPUBLIC OF TANZANIA FOR A LOCAL GOVERNMENT SUPPORT PROJECT April 30, 2013 Urban Development and Services Practice 1 (AFTU1) Country Department AFCE1 Africa Region CURRENCY EQUIVALENTS (Exchange Rate Effective April 1, 2013) Currency Unit = Tanzania Shillings (TSh) TSh 1.00 = US$0.0006 US$ 1.00 = TSh 1,620 FISCAL YEAR July 1 – June 30 ABBREVIATIONS AND ACRONYMS CBG Capacity Development Grant CDG Capital Development Grant/LGCDG CIUP Community Infrastructure Upgrading Program D by D Decentralization by Devolution DLA Dar es Salaam Local Authorities DMDP Dar es Salaam Metropolitan Development Project GoT Government of Tanzania ICR Implementation Completion Report IE Impact Evaluation LGA Local Government Authority LGCDG Local Government Capital Development Grant LGRP Local Government Reform Program LGSP Local Government Support Project M&E Monitoring and Evaluation MoF Ministry of Finance MTR Mid-term Review NGO Non-governmental Organization O&M Operations and Maintenance PDO Project Development Indicator PforR Program for results PM-RALG President’s Office Regional Administration and Local Government PMO-RALG Prime Minister’s Office of Regional Administration and Local PST Project Support Team REP Revenue Enhancement Program TRA Tanzania Revenue Authority TSCP Tanzania Strategic Cities Project Vice President: Makhtar Diop Country Director: Philppe Dongier Sector Director: Jamal Saghir Sector Manager: R. Mukami Kariuki Project Team Leader: Barjor Mehta ICR Team Leader: Andre A. Bald i TANZANIA Local Government Support Project CONTENTS Data Sheet A. Basic Information B. Key Dates C. Ratings Summary D. Sector and Theme Codes E. Bank Staff F. Results Framework Analysis G. Ratings of Project Performance in ISRs H. Restructuring I. Disbursement Graph 1. Project Context, Development Objectives and Design ............................................... 1 2. Key Factors Affecting Implementation and Outcomes .............................................. 6 3. Assessment of Outcomes .......................................................................................... 12 4. Assessment of Risk to Development Outcome......................................................... 20 5. Assessment of Bank and Borrower Performance ..................................................... 20 6. Lessons Learned ....................................................................................................... 22 7. Comments on Issues Raised by Borrower/Implementing Agencies/Partners .......... 23 Annex 1. Project Costs and Financing .......................................................................... 24 Annex 2. Outputs by Component ................................................................................. 25 Annex 3. Economic and Financial Analysis ................................................................. 26 Annex 4. Bank Lending and Implementation Support/Supervision Processes ............ 31 Annex 5. Beneficiary Survey Results ........................................................................... 33 Annex 6. Stakeholder Workshop Report and Results................................................... 34 Annex 7. Summary of Borrower's ICR and/or Comments on Draft ICR ..................... 35 Annex 8. Comments of Cofinanciers and Other Partners/Stakeholders ....................... 45 Annex 9. List of Supporting Documents ...................................................................... 46 Annex 10. Summary Community Infrastructure Upgrading Program Impact Evaluation ………………………………………………………………...47 ii A. Basic Information Local Government Country: Tanzania Project Name: Support Project Project ID: P070736 L/C/TF Number(s): IDA-40030,IDA-40031 ICR Date: 04/30/2013 ICR Type: Core ICR GOVERNMENT OF Lending Instrument: SIL Borrower: TANZANIA Original Total XDR 35.60M Disbursed Amount: XDR 97.38M Commitment: Revised Amount: XDR 102.70M Environmental Category: B Implementing Agencies: Prime Minister's Office Regional Administration and Local Government Cofinanciers and Other External Partners: B. Key Dates Revised / Actual Process Date Process Original Date Date(s) Concept Review: 07/30/2002 Effectiveness: 04/13/2005 04/13/2005 06/15/2006 Appraisal: 09/08/2004 Restructuring(s): 06/19/2009 Approval: 11/30/2004 Mid-term Review: 01/21/2008 Closing: 06/30/2008 06/30/2012 C. Ratings Summary C.1 Performance Rating by ICR Outcomes: Moderately Satisfactory Risk to Development Outcome: Moderate Bank Performance: Moderately Satisfactory Borrower Performance: Moderately Satisfactory C.2 Detailed Ratings of Bank and Borrower Performance (by ICR) Bank Ratings Borrower Ratings Quality at Entry: Moderately Satisfactory Government: Moderately Satisfactory Implementing Quality of Supervision: Moderately Satisfactory Moderately Satisfactory Agency/Agencies: Overall Bank Overall Borrower Moderately Satisfactory Moderately Satisfactory Performance: Performance: iii C.3 Quality at Entry and Implementation Performance Indicators Implementation QAG Assessments Indicators Rating Performance (if any) Potential Problem Project Quality at Entry No None at any time (Yes/No): (QEA): Problem Project at any Quality of No None time (Yes/No): Supervision (QSA): DO rating before Satisfactory Closing/Inactive status: D. Sector and Theme Codes Original Actual Sector Code (as % of total Bank financing) General water, sanitation and flood protection sector 10 10 Roads and highways 5 5 Sub-national government administration 85 85 Theme Code (as % of total Bank financing) Decentralization 23 23 Municipal finance 22 22 Municipal governance and institution building 22 22 Rural services and infrastructure 11 11 Urban services and housing for the poor 22 22 E. Bank Staff Positions At ICR At Approval Vice President: Makhtar Diop Gobind T. Nankani Country Director: Philippe Dongier Judy M. O'Connor Sector Manager: Rosemary Mukami Kariuki Jaime M. Biderman Project Team Leader: Barjor E. Mehta Matthew D. Glasser ICR Team Leader: Andre A. Bald ICR Primary Author: Andre A. Bald F. Results Framework Analysis Project Development Objectives (from Project Appraisal Document) The development objectives of the LGSP are: (i) to strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers systems; iv (ii) to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance. Revised Project Development Objectives (as approved by original approving authority) The objectives of the Project are to (a) strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers and demand-driven urban investments; and (b) increase access to infrastructure and services in the unplanned areas of Dar es Salaam and improve revenue performance for sustainable operations and maintenance. (a) PDO Indicator(s) Original Target Formally Actual Value Values (from Revised Achieved at Indicator Baseline Value approval Target Completion or documents) Values Target Years Percentage of Participating Local Government Authorities meeting minimum Indicator 1 : access conditions for accessing Local Government Capital Development Grants Value quantitative or 54% 80% 97% Qualitative) Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments (incl. % achievement) Percentage of Participating Local Government Authorities with 'clean' (i.e. Indicator 2 : unqualified) audit certificates Value quantitative or 20% 80% 54% Qualitative) Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments (incl. % achievement) Percentage of Transfers to Participating Local Government Authorities that are Indicator 3 : made within the first 30 days of each quarter Value quantitative or 0% 100% 44% Qualitative) Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments (incl. % achievement) Access to Services is improved: (a) the increase in the proportion of streets, (b) Indicator 4 : the reduction in households experiencing flooding; (c) percentage increase of enterprises. Value 25% of Phase 1 and 80%, 10%, 30% 97% Streets quantitative or 7% of Phase 2 streets v Qualitative) 22% of Phase 1 and 20% of Phase 2 households 1,838 in Phase 1 and 1,429 in Phase 2 enterprises Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments Flooding & Enterprises - Figures not collected at end of the project. Alternative (incl. % measures, discussed in section 3.1 of the ICR from the Impact Evaluation are achievement) highlighted. Dar es Salaam Local Authorities (DLA) revenue performance; (a) % increase in Indicator 5 : DLA won-source revenue by FY2011/12; (b) % unrestricted transfers and DLA revenues. 151% Own Source revenue = 41.0 Value Own source revenues = billion in July 2011 quantitative or Tshs 16.3 bn. 50% and 8% to May 2012, per Qualitative) 6.65% June 19, 2011. 11.9 % Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments (incl. % achievement) (b) Intermediate Outcome Indicator(s) Original Target Actual Value Formally Values (from Achieved at Indicator Baseline Value Revised approval Completion or Target Values documents) Target Years Indicator 1 : Percentage of Participating LGAs receiving 100 percent of the LGCDG Value 10 of 121 (8%) LGAs 20% (of 132 (quantitative qualified for a bonus in 20% LGAs) or Qualitative) 2006/07 Date achieved 11/30/2004 06/30/2008 06/29/2012 Comments (incl. % achievement) G. Ratings of Project Performance in ISRs Actual Date ISR No. DO IP Disbursements Archived (USD millions) 1 04/26/2005 Satisfactory Satisfactory 0.00 2 11/03/2005 Satisfactory Satisfactory 8.87 3 04/24/2006 Satisfactory Satisfactory 15.34 4 12/12/2006 Satisfactory Moderately Satisfactory 29.23 vi 5 05/02/2007 Satisfactory Moderately Satisfactory 49.66 6 10/26/2007 Satisfactory Moderately Satisfactory 51.73 7 04/15/2008 Moderately Satisfactory Moderately Satisfactory 56.64 8 12/15/2008 Moderately Satisfactory Moderately Satisfactory 72.46 9 06/23/2009 Satisfactory Satisfactory 84.52 10 11/25/2009 Satisfactory Satisfactory 86.75 11 03/29/2010 Satisfactory Satisfactory 97.73 12 12/10/2010 Satisfactory Satisfactory 131.58 13 06/28/2011 Satisfactory Satisfactory 142.92 14 12/31/2011 Satisfactory Satisfactory 144.58 15 07/11/2012 Satisfactory Satisfactory 151.18 H. Restructuring (if any) ISR Ratings at Amount Board Restructuring Disbursed at Restructuring Reason for Restructuring & Approved Restructuring Date(s) Key Changes Made PDO Change DO IP in USD millions 06/15/2006 N S S 23.55 Additional Financing Amendment to the DCA and the 06/19/2009 Y MS MS 84.52 AF Agreement If PDO and/or Key Outcome Targets were formally revised (approved by the original approving body) enter ratings below: Outcome Ratings Against Original PDO/Targets Moderately Satisfactory Against Formally Revised PDO/Targets Moderately Satisfactory Overall (weighted) rating Moderately Satisfactory I. Disbursement Profile vii 1. Project Context, Development Objectives and Design 1.1 Context at Appraisal Country Context When the Project started in 2003, Tanzania’s population was 35.9 million. It was among the world’s poorest countries, with a per capital Gross National Income of US$290. The country was shifting from a rural to urban economy and simultaneously making the transition from a one-party socialist state to a multi-party democracy. After its first national multi-party elections in 1994, the government started the process of decentralization by devolution, transferring greater responsibilities from central to local government. Sector Background Decentralization: The Government of Tanzania’s (GoT, or Government) decentralization process 1 aimed to improve intergovernmental fiscal transfers, strengthen revenue generation for Local Government Authorities (LGAs), and make service delivery more efficient and equitable. The GoT’s overarching policy, Decentralization by Devolution (D by D), was enacted in 1998 and shifted greater political, administrative, and financial powers to LGAs. Accordingly, their responsibilities for service delivery in education, health, agriculture, and infrastructure also increased. The Local Government Reform Program2 (LGRP) was the GoT’s primary instrument to implement D by D. LGRP was managed by the Prime Minister’s Office - Regional Administration and Local Government (PMO-RALG) and supported by donors3 though a common basket fund. Before LGRP and the introduction of the Local Government Capital Development Grant (supported through this project), direct development grants to LGAs were typically sectoral or area based. These resources were often earmarked, or controlled in a top-down manner. Further, the amounts for capital investments were small and not dependable for long-term capital works programming. Allocation of funds was not based on performance or a transparent/equitable formula. Overall, the intergovernmental finance system was fragmented, and there was no robust mechanism to finance infrastructure or to strengthen the capacity of LGAs. Improving the system was a necessary precursor for sustainable decentralized service delivery – and this provided the impetus for the Local Government Support Project. Urbanization: At the time, Tanzania was also urbanizing rapidly – with the share of the urban population increasing from 6 to 24 percent from 1967-2005. This demographic shift increased the stress on local governments to deliver basic services. Infrastructure coverage 1 Guided by the Local Government Reform Act 1982, second Poverty Reduction Support Program, and 1998`Policy Paper on Local Government Reform. 2 LGRP-I 1998-2008 and LGRP II 2009-2014 3 Netherlands, Finland, Ireland, Belgium, Canada, Germany, and the European Union. 1 and quality was poor and declining across sectors, and there was also widespread informality in the economy as well as in land/housing markets, with roughly 70 percent of the population residing in informal or unplanned settlements. Development trends in Dar es Salaam: Dar es Salaam was, and remains, the country’s largest and most important city for commerce and government. From 2002-2012, the population increased from 2.5 to 4.3 million – an average annual rate of 5.6%. The city also expanded outwards, with informal settlements covering 60 percent of the land. These demographic trends added to the significant challenges facing the Dar es Salaam Local Authorities (DLAs). Among others, a fragmented government structure, dependence on central government transfers, low capacity and lack of tools for revenue collection, and a backlog of basic infrastructure and services. Rationale for Bank Assistance and Contribution to Higher Level Objectives Against this background, the Local Government Support Project (LGSP, or Project) was developed to support both the decentralization agenda and key urbanization challenges in Dar es salaam. It therefore included two distinct components, one focused on strengthening the fiscal decentralization system, and one addressing urban priorities in Dar es Salaam, namely upgrading unplanned settlements and improving revenue enhancement. LGSP was consistent with the Country Assistance Strategy 2001-2003; which highlighted the priority areas of ‘Increasing Empowerment and Accountability’ and ‘Public Sector Reform and Institution Building.’ Decentralization was also supported by the Bank’s Economic Sector Work on Decentralization 2001, and the Government’s Poverty Reduction Strategy Paper as a foundation for service delivery to the poor and to improve local governance. LGRP was the government’s core public sector reform program to implement decentralization and had broad stakeholder and donor support. The Bank’s involvement would (i.) draw on its competencies in decentralization and urban management, (ii.) provide resources for local-level capital investments, and (iii.) help enhance donor support. For the urban management and upgrading components in Dar es Salaam – the Bank had significant experience in this field globally, and in Tanzania through the First and Second Urban Projects. The Bank was well positioned to address the community infrastructure needs and to help develop an upgrading project. 1.2 Original Project Development Objectives and Key Indicators The Project Development Objective (PDO) were: (i.) to strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfer systems; and (ii.) to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance. The original Key Performance Indicators (KPIs) are shown in Table 1. 2 1.3 Revised PDO and Key Indicators, and reasons/justification The PDO and Key Performance Indicators were revised in response to the additional financing and project restructuring (highlighted in Section 1.7). The project was extended for a total of 4 years. The first part of the PDO was revised to: “(i.) strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers and demand-driven urban investments.� The addition was to accommodate new project activities (namely financing preparation of the Tanzania Strategic Cities Program) approved through restructuring. The Key Performance Indicators were revised at restructuring to reflect: (i.) changes during implementation (expansion of the program coverage from 41 to 132 LGAs, and dropping sanitation interventions in CIUP); and (ii.) updated M&E practices (refer to section 2.3), such as the use of intermediate outcome indicators and targets. Table 1. Key Performance Indicators: Original and Revised Original Revised 1. Number of 41 targeted Local Government 80 percent of Participating Local Government Authorities receiving Local Government Capital Authorities meet minimum access conditions for Development Grants (LGCDGs) increased from accessing Local Government Capital Development 0 to 30 by 2007/08 (22 already qualified) Grants by FY2009/10. 2. Number of participating Local Government 80 percent of Participating Local Government Authorities with ‘clean’ audit performance Authorities with ‘clean’ (i.e. unqualified) audit records. certificates by FY 2009/10. 3. Proportion of resources transferred to Local Transfers to Participating Local Government Government Authorities which are made within Authorities are made within the first thirty days of each the first 30 days of each quarter increased to quarter by FY2009/10. 100% by 2007/08 4. Access to services in 16 beneficiary sub-wards Access to services provided under Part B.1 of the (approximately 10% of the Dar es Salaam Project is improved as measured by: population) improved, as measured by: (a.) the increase in the proportion of streets in the (a.) Reduction in travel time from home to surveyed area reachable by paved access roads. work, school/market/closest motor-able road (b.) the reduction in the proportion of surveyed (b.) Reduction in incidence of flooding of households experiencing flooding of premises. household premises (c.) the percentage increase in the number of (c.) Increase in frequency of waste collection enterprises in the surveyed area. (d.) Reduction in proportion of plots with no sanitation facilities 5. Dar es Salaam Local Authorities own source Dar es Salaam Local Authority revenue performance revenues increased by 50% by 2007-08 and improved as measured by: adequate funds are disbursed for O&M. (a.) 50 percent increase in Dar es Salaam Local Authority (DLA) own-source revenues by FY2011/12;and (b.) Eight percent of unrestricted transfers and DLA own-source revenues allocated for purposes of operations and maintenance by FY2011/12. 3 1.4 Main Beneficiaries The project was intended to benefit three groups:  Local consumers, through increased investments in infrastructure and improved services by LGAs. The investments In Component 2 for the Community Infrastructure Upgrading Program targeted an estimated 330,000 people in low-income households, estimated to cover 20 percent of unplanned areas in Dar es Salaam.  Local Government Administrations, through improved capacity, resources, and systems for service provision. The project originally supported 41 LGAs, but after additional financing and restructuring, it supported all 132 LGAs in the country.  PMO-RALG - which is responsible for administration of LGAs, through improved policy guidance, enhanced monitoring systems, and increased capacity to manage a decentralized capital grant system. 1.5 Original Components Component l: Support for Local Government Capital Development Grant System (USD35 million) This component supported the overall Local Government Capital Development Grant (LGCDG) system, which was also funded by GOT and bilateral development partners. The LGCDG system included (i.) a Capital Development Grant (CDG), and (ii.) a Capacity Building Grant (CBG). The LGCDG transfers were non-sectoral, and distributed on a formula basis to LGAs, who could invest in accordance with local needs as determined through a participatory planning and budgetary processes. The CBG would provide resources to help LGAs to build the capacity to access and manage the CDG. Component 2: Dar es Salaam Upgrading and Institutional Strengthening (USD18.8 million) This component had two sub-components: (a) the Community Infrastructure Upgrading Program, a community-driven infrastructure upgrading program that targeted unplanned areas in Dar es Salaam; and (b) the Local Revenue Enhancement/O&M Enhancement Program, which sought to improve revenue collection systems to support O&M expenditures. Component 3: Support to PO-RALG (USD7.9 million) This component supported the President’s Office – Regional Administration and Local Government (PO-RALG) in the implementation, monitoring, evaluation and audit of the Project, and the fiscal transfer program supported by Component 1, while building the capacity of PMORALG to execute these functions as part of routine activities. 1.6 Revised Components As a result of the Board approved restructuring (June 19, 2009), the description of Component 3 was revised to: ‘Strengthening the operation capacity of PMO-RALG 4 to support the carrying out of the coordination, implementation and monitoring and evaluation 4 An additional minor change to the project description due to the restructuring was to reflect (i.) the administrative change in name of the ‘President’s Office’ to the ‘Prime Minister’s Office’ – Regional Administration and Local Government, and (ii.) inclusion of Project activities to support preparation of additional LGA investments through TSCP. 4 of the Project including inter alia: (a) annual performance assessments of the Participating LGAs; (b) the administration and monitoring of CDGs and CBGs; (c) operations and maintenance budgeting for purposes of local revenue generation; (d) monitoring of service delivery and maintenance by Participating LGAs; (e) local government capacity building and management program; (f) project monitoring and evaluation, and progress reporting; (g) information dissemination; and preparation of additional investments in LGAs.’ 1.7 Other significant changes The Project received an additional credit early during implementation, and it was restructured, and the closing date extended twice for a total extension of 4 years. The specific changes were:  An additional Credit of US$98.0 million, Approved by the Board on June 15, 2006. This was used to scale-up support and activities to the Government’s Local Government Capital Development Grant (LGCDG); extend support to the LGCDG system from FY 2006/07 through FY 2009/10; and allow LGSP funding to be used nationally to cover all 132 LGAs. The closing date was extended by three years to June 30, 2011. The additional credit restored the Bank’s total financing to the amount planned during preparation (between US $90 to $130 million), but scaled-back at appraisal due to IDA-13 funding limitations.  Restructuring, Approved by the Board on June 19, 2009. This enabled the reallocation of cost savings and unallocated funds from Component 1 to finance cost increases of US$4.1 million in Phase I of the Community Infrastructure Upgrading Project (CIUP); scaling up (Phase II) of CIUP through an additional US$20.8 million; an allocation of US$7.5 million for preparation of the Tanzania Strategic Cities Project (TSCP); revision of the results framework; dropping the household sanitation activities; and a one year extension of the closing date to June 30, 2012. A waiver of OP 13.20 was approved by the Bank’s management (May 5, 2009) to allow an additional one-year extension beyond the three-year limit for Additional Financing. Table 2. IDA Financing (US$ Millions) Component Approval Additional Financing Restructuring Nov 30, 2004 June 15, 2006 June 19, 2009 Component 1. Support for Local 35.0 125.9 93.1 Government Capital Development Grant System. Component 2. Dar es Salaam upgrading 18.8 15.9 39.4 and Institutional Strengthening Component 3. Support to PMO-RALG 6.3 8.2 *20.1 Total 52.0 150.0 152.7 *includes unallocated amounts after the second restructuring ($4.6 million) 5 2. Key Factors Affecting Implementation and Outcomes 2.1 Project Preparation, Design and Quality at Entry Analytical There was a strong rationale and analytical foundation for the Bank’s interventions with respect to decentralization, urban management, and upgrading. All Project components carried out technical research and prepared quality feasibility studies as the basis for the concept design and appraisal.5 The preparation missions were composed of appropriate cross-sector Task-Teams, and findings were well documented and discussed. The Project was prepared in close consultation with GoT counterparts, donors, communities, and NGOs. The Project Appraisal Documents (PAD) highlighted the lessons learned from the Bank’s experience in decentralization and urban management. 6 Key lessons influencing the design of Component 1 included the need for: (i) a long-term ‘learning by doing’ approach to introduce changes to local governments systems; (ii) LGA implementation of local investments; and (iii) the use of performance and demand-based approaches. The design of Component 2 reflected lessons from similar upgrading operations, including the use of: (i) systematic selection criteria for communities, (ii) community involvement in planning and monitoring , (iii) community contributions as demonstration of demand, (iv) the use of per hectare cost norms to distribute benefits, (v) flexible standards to minimize resettlement impacts, and (vi) upfront commitments from LGAs for operations and maintenance. Assessment of the Project Design The Project design was relatively ambitious – though the justifications for, and the consideration of alternative approaches were articulated in the PAD. Three diverse activities (supporting fiscal transfer mechanisms, urban upgrading, and revenue enhancement activities) were implemented at the national, local, and community levels. There was intensive consultation and engagement with the donor community and among Government agencies at all levels. There were no direct linkages between Components 1 and 2 – and this is reflected by the Project’s ‘two PDOs.’ Both components could have been a stand-alone project, and the PAD indicates it would be a ‘simpler and more elegant approach’ to omit Compo nent 2 on upgrading. The upgrading component was retained because it offered opportunities to (i.) directly target the urban poor, and (ii.) start a pilot operation and engagement with local authorities in Dar es Salaam - one of the world’s fastest urbanizing cities. 5 Components 1 and 3 were prepared by Price Waterhouse Coopers (financed by the LGRP Common Basket Fund); Component 2a, including designs for Phase I and safeguards instruments by DHV/UCLAS (financed through a Japanese PHRD Grant); and Component 2b by consultant, Robert Faber. 6 Lessons from the Bank’s global experience were drawn from (i.) the African Urban Sector Strategy (Nov 2001); The Operation Evaluation Department Evaluation Summaries; and (iii.) the report, ‘Fiscal Decentralization and Sub -national finance in Africa (DANIDA and the World Bank, 2000). Regional operational experience was drawn from (i.) the Zimbabwe Local Government Capital Development Project; the Uganda Local Government Development Project, the Senegal Urban Development and the Decentralization project. 6 As noted above the Project was initially designed to be larger (up to US$ 150 million), but was reduced at appraisal due to IDA-13 funding constraints. The revised appraisal budget was US52.0 million, requiring the elimination of the final year of support to the LGCDG system (Component 1) and deferral of Phase II of Component 2. However, by Project completion, all of the original/intended activities were later carried-out - the project was restructured and additional financing provided. Each component required co-financing. For Component 1, LGAs had to provide 5% of the LGCDG from own source revenues as a minimum condition to access the grant. For Component 2, communities had to contribute 5% of capital costs and DLAs an additional 5% as a condition to start the works. The rationale for co-financing was to demonstrate demand and commitment. From the original project cost (USD60.8 million), IDA financed 86%, GoT 10%, and DLAs/communities 4%. At project close, the financing totaled USD158.8, shared 94% IDA, 3% GoT, and 3% DLAs/communities. Component 1 was designed to work through the GoT’s existing LGRP. Prior to the Project, many donors (pre-2004) had separate ad hoc programs to support LGAs. However, during preparation, the Bank worked closely with GoT and Development Partners to create a more systematic approach to support LGAs. As a result of this project, most donors committed their financing to work through the LGCDG and CBG established in Component 1. This was an important achievement of the Project. Assessment of Risks The PAD provided a realistic assessment of the broad risks and mitigation measures. Some of the risks considered ‘substantial’ and ‘high’ included:  The political and policy environment for local revenues will deteriorate  Failure of GoT to put in place adequate legal, policy, fiscal framework for LGAs  LGA capacity may not be sufficient to implement all of the control procedures  Counterpart funds may not be sufficient given declining levels of local revenues The Overall Risk rating was ‘Substantial.’ Areas that became problematic during implementation, but were assessed as ‘moderate’ included:  Operations and maintenance measures by councils are inadequate  LGAs receive transfers late due to delays in releasing funds from MoF  Political interference in project operation/in revenue enhancement efforts The assessment of risks did not flag the complexity of the overall project design – which had three distinct initiatives, would require diverse skill sets, and would be implemented through several levels of government. Some activities were pioneering in Tanzania. Others, such as improving own-source revenue under Component 2b were highlighted as risky at appraisal, but ultimately considered a high-risk high-reward endeavors, necessary for LGAs to be able to maintain infrastructure. The PAD identified in the lessons learned that the success of each key component required a long-term engagement. Given this, the original project implementation period of 3.5 7 years – which reduced when the budget was halved (effective 01/01/2005 to 06/0/2008) - was increased to 7.5 years to accommodate the additional financing added in 2006. 2.2 Implementation Components 1 and 3 supporting the Local Government Capital Development Grant The Project (through Components 1 and 3) transformed the existing fragmented intergovernmental transfer system and created the foundation for sustainable decentralized service delivery. It introduced and created a formula-based, transparent, and predictable fiscal flow mechanism to disburse funds to LGAs on the basis of institutional performance. Fiscal transfers to LGAs were disbursed through two windows: the LGCDG and CBG. The LGCDG provided funding for capital expenditures, which were selected by recipient LGAs using local planning and budgeting systems, while the CBG provided funding to support capacity building at the local level to enable efficient and effective use of the LGCDG.7 Prior to the introduction of the LGCDG, direct development grants to LGAs were small and not performance based. LGAs received additional development resources from sectoral and area-based development programs, contained within the respective line ministry’s budget, and the use of these resources was often earmarked or controlled in a top-down manner. With the support of this project, the LGCDG became formula-based. It introduced minimum qualification standards to build fiscal decentralization and ensure accountability in the use of public funds. Over time, the LGCDG evolved into the Local Government Development Grant (LGDG) Program, which mainstreamed performance based fiscal transfers. Beyond expectation, the Project galvanized the support of development partners who channeled their programs through the LGCDG system. As a result, it became possible for the program to be scaled-up from 41 LGAs (at appraisal) to cover all 132 LGAs in Tanzania, thus leveraging an additional US 181.5 million (estimated) in financing from development partners. The LGDG system and the larger local government reform process achieved a number of significant improvements in LGA performance. For instance, the requirement to include communities in LGA budgeting and planning improved participatory planning and budgeting, as well as accountability at the LGA level.8 The increased volume of funding 7 Today the Council Development Grant and the Capacity Building Grant are collectively known as the “LGDG core�. New sector windows (LGDG non-core) have been added as the government has rationalized and decentralized funding for development needs, in particular sectors where LGAs have core responsibilities. The LGDG system consists of the following s windows: (i.) LGDG Core, as described above; (ii.) LGDG Non-Core, added between 2007 and 2009, (iii.) Agriculture Sector Development Grant (ASDG) established in 2009 for agricultural extension and development; (iv.) Water Sector Development Grant (WSDG) established in 2007 for water and sanitation infrastructure and services; (v.) Health Sector Development Grant (HSDG) established in 2009 for primary healthcare services. 8 Tidemand, Msami, Impact of Local Government Reforms in Tanzania, 2010 8 channeled to LGAs also led to an increased amount of investment in LGA infrastructure and services9. The performance assessment element of the LGDG system also improved LGA systems by creating the incentives meet pre-defined targets.10 These improvements are reflected in public opinion. When asked in an independently conducted citizen survey (2003) if local government reforms are improving service delivery, 60.5% responded “yes�. This rate went up to 70.24% when the survey was repeated in 2009. There were some operational challenges during implementation, notably:  M&E: The M&E framework was not well executed. There were delays populating the baseline and interim results. The project underwent substantial changes at restructuring, although the revised targets seem to have been too ambitious, given the increased scale of the program (from 41 to 132 LGAs).  Assessment of impacts/outcomes: Specific reporting on the outputs and outcomes of the LGCDG were late and with inconsistent quality.  Government Contribution to the LGCDG system: The MTR found USD18 million in GoT’s own funds that were reportedly budgeted for the LGCDG system (and which were supposed to be transferred to qualifying LGAs) were transferred to non-qualifying councils to construct office buildings and other purposes. GoT ended the practice and recommitted to provide its own resources to LGAs though the LGCDG system.  Timeliness of GoT Transfers to LGAs through LGCDG: GoT was often late transferring LGCDGs and CBGs to local government, although performance improved over the course of the project. Without reliable transfers – it was difficult for LGAs to plan their capital works program. The Bank consistently raised the issue. However, reasons for the delays varied - e.g. fiscal constraints (occasioned by the financial crisis) in 2010/11 were a contributing factor during that period. Component 2: Dar es Salaam Upgrading and Institutional Strengthening Component 2a.CIUP successfully piloted upgrading in the three municipalities in Dar es Salaam. While CIUP faced implementation challenges – these were consistent with the risks anticipated at appraisal. For example, Phase I works were delayed, primarily due to collection of the requisite counterpart contribution from LGAs and communities. In addition, the MTR also references procurement delays, poor contract management, disputes on compensation rates, and the need to strengthen supervision arrangements. Some issues include:  Contract Management: Contract management of CIUP was problematic. Some ‘completed’ contracts required additional follow-up; there was a pattern of long contract extensions; and in some instances, contracts expired while works was on- going. These practices contributed to implementation delays and drove up costs.  Counterpart Funds: Counterpart contributions were consistently late, or were not fully paid. Communities were obligated to pay their full portion of counterpart funds, and 9 MDF Consulting, Local Government Development Grant System Mid Term Review, 2011 10 Liviga, Roell, Mhina, Effectiveness of “Decentralization by Devolution�: Financial Resources versus Absorption Capacity, 2010 9 did, but the time taken to collect community counterpart funds slowed implementation. However, the DLAs did not pay their full contribution. At project closing, it was noted that the DLA’s had used IDA funds to pay their counterpart contributions. These funds (USD 1.711 million) were deemed ineligible expenditures and had to be refunded.  CIUP Sustainability: The approach agreed by PMO-RALG and the Bank was that CIUP O&M should be addressed through LGA systems, instead of specific arrangements for CIUP communities. However, while communities paid their counterpart contribution before works started, these funds were not reserved for long- term O&M, and no alternative funding was budgeted. Component 2b. Revenue Enhancement and Operations and Maintenance The MTR flagged the difficulties and delays in implementing the Revenue Enhancement Program (REP), although one of the key performance indicators (local revenues have increased by 50%) had been met. A key factor affecting implementation was the reported lack of institutional or political will to tackle revenue enhancement. A further complication arose from the GoT’s decision, during the course of project implementation, for the Tanzania Revenue Authority (TRA) to ‘temporarily’ take over tax collection functions on behalf of DLAs. This resulted in putting the valuation work on-hold (delaying implementation by roughly a year) until there was clarity on institutional mechanisms. 2.3 Monitoring and Evaluation Design, Implementation and Utilization There were shortcomings in the design and execution of the M&E framework that resulted in unrealistic targets and measures that did not fully capture the overall positive achievements and transformative aspects of the Project. The original monitoring and evaluation (M&E) framework was revised during restructuring; to introduce intermediate outcome indicators; and reflect changes arising from the additional financing and restructuring. The project’s success (primarily when it was scaled-up nationally from 41 to 132 LGAs, and leveraged the resources from other donors) made M&E design more challenging with respect to (i.) developing realistic targets appropriate given the mix of capacities and the range of the expanded LGAs and (ii.) the administration of the grants. A baseline was not completed at appraisal, and successive supervision missions flagged delays in populating the framework. However, it was only after the MTR and restructuring, that the overall M&E performance improved for Components 1 and 3. The revised key performance indicators for Component 2 relied on the post-project Impact Evaluation. However, as the IE could not capture all of the key indicators using household surveys, the M&E for this Component fell short of expectations. Overall, M&E data was not well utilized to inform decision making. The Project used third party consultants to carry-out the analysis for the Mid-term Reviews. These reports were thorough, high quality, and provided a detailed stocktaking and recommendations for the client and Bank to make evidence based decisions. The final reports for Components 1 and 3 were well prepared with detailed analysis (the summary is 10 in Annex 7.) A final report on Component 2 was not prepared – instead the Impact Evaluation was used to assess the results of the CIUP. 2.4 Safeguard and Fiduciary Compliance Safeguards: The Project was appraised as category B, and triggered the following policies (i.) Environmental Assessment (OP4.01); and (ii.) Involuntary Resettlement (OP4.12). The Project’s key safeguards instruments supporting CIUP included an Environmental and Social Management Framework, Resettlement Policy framework, Environmental Management Action Plan, and Resettlement Action Plan. The instruments were updated to reflect additional project activities in Phase II of CIUP. There were no substantial safeguards issues identified during implementation. Except for one case of a ‘Moderately Satisfactory’ rating, safeguards were rated ‘Satisfactory’ throughout implementation. A specific environment and social safeguards mission was carried out after the MTR (since this was not included in the MTR), and numerous areas for improvement were identified. Subsequent Aide Memoires and ISRs do not report as rigorously on safeguards. Project documentation should have included more details on implementation and supervision issues to help justify the consistently high ISR ratings. Procurement: Procurement of goods, works, and consulting services was generally satisfactory. There were early delays with some of the CIUP contracts (mainly due to processing and approval for additional works, addendums, and extension of completion periods) which affected implementation. These issues were identified and followed-up appropriately. Procurement plans were regularly updated. Independent Procurement Review and Value for Money Audits were carried out (2007 and 2009) – findings, recommendations for improvement, and required actions were discussed between the Bank and client. However, no significant issues were identified in the reviews. Financial Management: There were several financial management issues identified throughout implementation. Firstly, the Supervision team identified (in December 2006 Aide Memoire) that the government was in breach of a legal covenant on its obligations to provide counterpart funding – an issue that was identified earlier but not fully resolved. Secondly, the MTR (Feb 2008) flagged USD1.71 million documented in the FMR that was not recorded in the Bank system. This triggered a significant effort by the Bank to work with PMO-RALG and Project Support Team Dar es Salaam (PST-DAR) to reconcile disbursement records. Thirdly, at the end of the project the Bank became aware that the DLA paid contracts using 100% IDA, with the understanding that GoT would cover the counterpart contribution. However, this did not happen, and ultimately, after the Project closed, the Bank requested the Ministry of Finance to refund USD1.71 million in ineligible expenditures i.e. the use of IDA funds for counterpart co-financing. Despite these shortcomings, fiduciary ratings however were generally ‘Satisfactory.’ It may have been prudent to downgrade the ratings to elevate specific problematic FM issues – which were overshadowed by the overall positive message. 2.5 Post-completion Operation/Next Phase Component 1, Support for Local Government Capital Development Grant: At project completion, the performance based fiscal transfer had become a permanent feature of 11 Tanzania’s fiscal architecture, with an increasingly greater share contributed from GoT’s own sources. For instance, while the Government’s share of the LGDG core disbursement in FY2005/06 was 7% (US$ 1.5 million out of US$ 22.4 million disbursed), this has increased progressively to 22% in FY2010/11 (US$ 15 million out of US$ 68.4 million disbursed) and to 40% in FY2011/12. In the context of GoT’s tight overall budgetary position, this trend demonstrates the success of LGSP in helping GoT create and sustain a transparent fiscal transfer mechanism, the LGDG. The Government has built on the success of the Project, requesting World Bank support for the design of a specific “urban� performance grant window within the LGDG system using the Program for Results (PforR) instrument. The Bank board approved a new operation, known as the Urban Local Government Strengthening Program (ULGSP), in October 2012. This is also the Bank’s first PforR operation in Africa. Leveraging the success achieved by LGSP, ULGSP will carry forward and reinforce the decentralization, institutional development, and local government capacity building agenda. Component 2a: CIUP: There are no specific requirements or mechanisms for post- completion transition arrangements for the upgrading works, outside the DLA’s normal responsibilities for O&M. The Project did not put into place specific O&M arrangements for CIUP. However, in line with plans to address the backlog of unplanned settlements in Dar es Salaam, a follow-up project, currently under preparation (Dar es Salaam Metropolitan Development Project) will continue the upgrading work, reinforcing the CIUP model with the city and drawing from the lessons of the CIUP Impact Evaluation. Despite the popularity and demand for CIUP – the expectations in the PAD that it would be scaled-up by the city, with the city’s own financing, did not materialize. Component 2b: Revenue Enhancement: The Project facilitated the identification and valuation of all taxable properties across the three DLAs. The tax base was greatly expanded and the tools are in place for the DLA to move forward by increasing the tax rates and improving billing and collection. Efforts to consolidate and expand the city’s revenue base will also be supported under the follow-up project, currently under preparation (Dar es Salaam Metropolitan Development Project). 3. Assessment of Outcomes 3.1 Relevance of Objectives, Design and Implementation Rating: High The overall relevance of the objectives remains High. The Project’s design, objectives, and implementation mechanisms are consistent with Tanzania’s development priorities. The decision to support the then-nascent GoT’s decentralization effort by focusing on performance-based grant mechanism proved correct, as demonstrated by the fact that the mechanism was scaled up nationally to cover all LGAs, and became the instrument of choice for transferring funds for the agriculture, water, health and most recently urban 12 sectors. The performance-based grant system (whose creation was supported mainly by LGSP) is now the main transfer mechanism for capital and sector specific grants. The relevance of LGSP’s focus on supporting decentralization is further demonstrated by the fact that decentralization is now a permanent feature of Tanzanian public policy and is explicitly referred to in GoT’s five-year (2010/11–2014/15) national growth and poverty reduction strategy, MKUKUTA II. MKUKUTA II is a medium-term plan that aims to achieve the goals of Tanzania’s Development Vision 2025 (TDV 2025) and the Millennium Development Goals (MDGs). TDV 2025 aspires to make Tanzania a middle income country characterized by high quality livelihoods; peace, stability, and unity; good governance; a well-educated and learning society; and a strong and competitive economy by 2025. One of the core pillars of MKUKUTA II, is the national decentralization reform, (dbyD) which is the Government’s overarching decentralization policy. The financing for these responsibilities comes from flows provided by the fiscal decentralization component, which was supported by the LGSP. The relevance of the urban upgrading component, which established a successful pilot that introduced the key principals of upgrading to three municipalities and improved access to infrastructure and services in target areas, is also high. CIUP provided the operational and knowledge foundation for addressing the challenges of a city that is rapidly urbanizing without adequate planning and infrastructure and service provision. The three municipalities and PMO-RALG have therefore requested a scale-up upgrading operation through the Dar es Salaam Metropolitan Development Project (now under preparation). The original project design and objectives are consistent with two of the four strategic goals of the CAS for FY2012-15: (i.) to build infrastructure and deliver services and (ii.) to promote accountability and governance. The Project is consistent with the Government’s Letter of Sector Development Policy on Urban Development (January 2010) which stresses the need for enhanced support to urban areas and for the long-term financing for urban infrastructure. All components established a foundation (institutional development, technical, and strategic) upon which the Bank’s current urban portfolio draws.11 3.2 Achievement of Project Development Objectives Rating: Moderately Satisfactory The PDO contains distinct objectives associated with the two major components. Out of the five PDO level indicators, the project fully achieved two and partially achieved three. The revised Project Development Objectives (PDO) were: (i.) to strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers and demand-driven urban investments, and 11 The Tanzania Strategic Cities Project, Urban Local Government Strengthening Program, and the Dar es Salaam Metropolitan Development Project (under preparation) 13 (ii.) to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance. The M&E framework (reference section 2.3) had some weakness, and overall did not fully capture the satisfactory results on the ground – thus additional complimentary measures are presented in Table 3 to reflect the results towards the PDO. Further, the indicators for Component 2 utilize the robust analysis from the Impact Evaluation for the corresponding measures of connectivity, flooding, and enterprise development. There are minor variations with the unit of measure (for 2 of the 3 indicators) in the framework that preclude assessing the Impact Evaluation derived indicator against the original targets. However, these indicators are judged to be accurate for evaluation of improved access to services – and the M&E framework indicates the final results were to be derived from the Impact Evaluation. Component 1, Support for Local Government Capital Development Grant: This component did achieve its broad objectives to strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of the intergovernmental transfer system. LGSP transformed the pre-existing ad-hoc, fragmented, and unpredictable intergovernmental finance system into a more transparent, results based, and equitable system. LGCGD became the primary intergovernmental transfer system to LGAs, and was the reason the Development Partners consolidated their support to LGAs instead of maintaining their earlier area-based programs. Since the start of the project in 2004, roughly US$350 million in grants 12 have been transferred to local government. These funds were used by LGAs for school construction, offices, health facilities, and agriculture. Value for money audits found 80 percent of investments were rated as ‘good’ or ‘best.’ Moreover, there is evidence (through citizen perception studies) of improved performance of LGAs over the life of the project. Key improvements as a result of the project are: (i.) the inclusion of communities in the budgeting and planning process, (ii.) accountability at the LGA level, and (iii.) increased investments in infrastructure and services due to the increased volume of funding for LGAs. The capacities and performance of LGAs has improved over the project period. Almost all LGAs now meet the minimum conditions for entrance in the LGCDG program. These conditions encompass financial management, planning and budgeting, LGA council’s functional processes, audit, and reporting and accountability. This significant increase in the capacity of local government is attributable to the LGSP’s capacity building efforts and built-in performance incentives for the annual assessments. Component 2a: CIUP CIUP used a participatory approach to plan and implement community-level infrastructure targeting 20 percent of the unplanned settlements in Dar es Salaam. The outputs for CIUP 12 IDA financed about 25% of the total grants under the LGDG system, GoT about 17%, and the remaining 58% from the Development Partners Common Basket Fund. 14 included 40km of two-way roads, 58km of one-way roads, 26km of footpaths, 548 culverts, 8.6 km of trunk drains, 3,153 pedestrian crossings, 83 solid waste containers, and 2,780 streetlights – impacting roughly 300,000 people in priority unplanned settlements. An Impact Evaluation13 (IE) was carried out on CIUP (refer to Annex 10 for the full report). Some of the findings include:  CIUP improved access to services for the average household. The time to reach a water supply decreased significantly to 1.5 minutes (from 3 minutes in the control), and toilets from 13 to 4 minutes. The proximity of streetlights on roads increased, from 2% in the control area to 46% in the treatment area, reducing time to reach a street light from 15.5 minutes to 3 minutes. Access to drainage directly outside dwellings increased from 14% to 37%. Households indicating that the access road to their dwelling was “somewhat improved� jumped from 7% to 48%.  Children in Project communities experienced reductions in diarrhea and acute respiratory illness (ARI) though no impact was found for adults. For children under 5, diarrhea almost halves, from 22% to 12%, and ARI incidence in the past 12 months drops from 27% to 18%.  School enrolment increased in Project communities, with the largest impact for girls older than 13. CIUP interventions are associated with an overall increase in student school enrolment from 85% to 89%. The cohort driving the observed impact was girls aged 13 – 16, with the control group having a 79% enrolment rate compared to 93% in the treatment group.  Operations and maintenance remains a challenge for sustainability. The IE found 48 percent of households reported streetlights were not working, and 29 percent flagged poorly functioning drainage. However, 94 percent reported garbage collection activities were well functioning, as were the community roads (98 percent). Component 2b: Revenue Enhancement The key performance indicators to improve revenue performance in Dar es Salaam were achieved. The percentage increase in DLA own-source revenue increased by 154 percent between 2005-2011 – well above the target level of 50 percent. The percentage of un- restricted transfers and DLA revenues increased from 6.60 to 11.9 percent between 2005- 11, above the target of 8 percent. Table 3. Results of PDO Indicators and Proxy Measures for CIUP Key Performance Indicator Results and Comments PDO 1: To strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfer systems Moderately Satisfactory PDO indicator 1. Target Exceeded – 97 percent 13 The Impact Evaluation compared the CIUP 31 sub-wards a control group of 12 sub-wards not offered the intervention. A baseline survey of 1,860 households in control and treatment sub-wards was conducted in 2006, upgrading began in 2007 and the follow up survey was completed in 2012. A stakeholder focus group was held on January 28, 2013 to assess the preliminary results of the IE – with partners from the Dar City Council, the National Bureau of Statistics, NGOs, and implementing agency technical staff. 15 80 percent of Participating Minimum access conditions include: financial management, planning and Local Government Authorities budgeting, LGA council’s functional processes, audit, and reporting and meet minimum access conditions accountability. At the onset of the project, 53% of the participating LGAs for accessing Local Government qualified for the minimum conditions (2004/05). In the final year, Capital Development Grants by (2010/11) this went up to 97%. FY2009/10. This was achieved despite the major increase in project’s scope from a target, prior to restructuring, of 41 LGAs, to 132 LGAs at the end. PDO indicator 2. Target partially achieved - 54 percent 80 percent of Participating Steady improvements were made towards meeting this indicator. By the Local Government Authorities end of the project the percentage of LGAs with clean audits had with ‘clean’ (i.e. unqualified) improved from 41 percent in 2003/04 to 54 percent in 2010/11(despite audit certificates by FY 2009/10. tripling the number of LGAs). Overall, the quality of audits for LGAs has improved. And as noted by the client’s final evaluation report, the target was not realistic given that audit reports span across four categories: Clean, qualified, adverse and disclaimed. Even in developed countries with significantly more LGA capacity, it is difficult, if not impossible to have 80% of local governments achieve the highest “score� – a clean audit. Arguably, having 0% adverse audit reports would have been the optimal indicator. As such, audits with adverse opinions dropped from 24% to 4% during the same reporting period. PDO indicator 3. Target partially achieved Transfers to Participating Local Due to the GoT’s overall tight budget position, this indicator was only Government Authorities are partially met. The GoT’s track record of timely transfers to LGAs made within the first thirty days fluctuated greatly throughout the project period ; while in 2008/09 83% of each quarter by FY2009/10. of transfers were made on time, the rate was 44% in 2009/10 The final evaluation report posits this was the least relevant benchmark, and as long as LGDG transfers were made regularly and predictably each quarter (rather than being bunched in Q4 when it is difficult to undertake activities) the timing of the disbursement during the first thirty days did not detract from the success of the transfer system. It is important to note, that despite delays in disbursement of funds, contrary to global practice, the performance assessments which determine the amount to be disbursed to LGAs have generally been completed on time. PDO 2: to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance Moderately Satisfactory PDO indicator 4. Targets partially achieved Access to services provided under of Overall, the findings from the Impact Evaluation show increased the Project is improved as measured access to infrastructure and services, assessed by connectivity to a by (3 indicators below): paved or gravel road, lower probability of flooding, and a decrease in enterprises. Positive results were assessed for 2 out of 3 indicators. The Impact Evaluation unambiguously assessed Component 2 as resulting in improved access to services for water supply, toilets, street lighting, roads, and drainage. (see discussion on Component 16 2a above) Indicator From IE Unit of measure based on the Impact Evaluation The increase in the Increase in Achieved: The GoT reports 97 percent against the original target of proportion of streets in connectivity 80 percent. the surveyed area to a paved The IE finds that the probability that a household was connected to reachable by paved or gravel a paved/gravel road increased from 4.7 to 27.9 percent from the access roads. road: baseline to Project close. The reduction in the Reduction in The probability that a household was flooded in the past 12 months proportion of surveyed Flooding reduced slightly, from 24.5 to 21.6 from the baseline to the IE households experience survey. Note: during the final year of the IE survey, there were flooding on preemies substantial flood events (including a 1 in 100 year flood) that may have skewed the reporting. The percentage Target set at This target could not be measured by the IE, as it measured sample increase in the number 30% households, not a total census of businesses. The counterpart did of enterprises in the not monitor the original measures. For comparison, the IE data surveyed area showed 60 percent of households in the CIUP communities had enterprises at the start of the project, and this dropped to 57 percent at the end. A comparable drop was found in the control area. Overall, the IE did not show significant improvements in employment or business entrepreneurship – most likely because both indicators started at high baselines (90 percent). PDO indicator 5. Target Exceeded Dar es Salaam Local Authority revenue performance improved as measured by: (a.) 50 percent increase in Dar es The percentage increase in DLA own-source revenue increased by Salaam Local Authority (DLA) 154% between 2005 and 2010/11 – well above the target of 50%. own-source revenues by FY2011/12;and (b.) Eight percent of unrestricted The percentage of unrestricted transfers and DLA revenues transfers and DLA own-source increased from 6.6% to 11.9% between 2005 and 2010/11 – above revenues allocated for purposes of the target of 8%. operations and maintenance by FY2011/12. 3.3 Efficiency Efficiency of Component 1 (Support for Local Government Capital Development Grant System). This component was designed to help achieve the first project development objective: “strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers systems�. Through this component, the bank helped GoT to establish a capital grants system that gives Local Government Authorities (LGAs) a much higher degree of freedom to allocate capital grants across sectors than the pre-existing system, whilst ascertaining that LGAs continued to meet good governance criteria. There is strong evidence that the system has improved resource allocation at the LGA level by replacing area-based grant systems with performance based ones. The share of grants allocated to sectors with generally high economic benefit/cost ratios (notably roads, drains and water supply) increased at the expense of investments in sectors with lower or negative economic benefit/cost ratios (such as public buildings and real estate). 17 Moreover, this component brought efficiency in fiscal transfers and in local government investment provision as it provided significant harmonization and greater transparency in resource allocation to LGAs. Furthermore, Component 1 helped align LGA service provision more closely with local needs by allowing performance-based funds to be expended on a discretionary basis, thus eliminating economic dead-weight loss caused by misalignment of funds and local needs. Efficiency of Component 2 (Dar es Salaam Upgrading and Institutional Strengthening). This component was designed to help achieve the second project development objective: “increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance.� It consisted of two sub-components: (i) the Community Infrastructure Upgrading Program (CIUP), and (ii) the Local Revenue Enhancement / O&M Enhancement Program. The economic benefits of the CIUP are substantially higher than the economic costs thereof, mainly because the program has resulted in a major improvement to public health in the project area. A recently concluded impact evaluation did not find a measurable impact of the CIUP on flood damage or on business establishment. However, because of higher-than-expected public health benefits, the economic rate of return (ERR) of the program was re-estimated at 15%, which was lower than the ERR of 20% that was estimated at the time of appraisal, but remained higher than the minimum required return of 12%. In conclusion, the cost of achieving both project objectives was deemed reasonable in comparison with both the benefits and with recognized norms. 3.4 Justification of Overall Outcome Rating Outcome Rating: Moderately Satisfactory The Project and priorities identified at appraisal remain relevant. The Project did not achieve all of its specific targets (noting the issues with M&E), though supporting measures demonstrates the broad achievements towards the PDOs and substantial impacts of the Project, including unexpected benefits under Component 2. The cost associated with the results were founds acceptable and efficient. Based on this, and considering the performance before and after restructuring, the overall outcome rating is evaluated as Moderately Satisfactory. Component 1. Strengthen Fiscal Decentralization The project was instrumental in making the GoT’s decentralization policies and strategies operational. LGSP established a single, harmonized, and transparent formula and performance-based system of intergovernmental transfers. The Project introduced the use of minimum conditions as basis for LGA participation, focused on timely disbursements of grants, and improved accountability in the use of LGA resources. This was accomplished working through country systems. Some aspects of the Project exceeded expectations. LGSP was instrumental in galvanizing the support of other development partners, who then channeled their programs through the LGCDG system. Because of this, the program (which originally intended to be a pilot supporting 41 LGAs) was rolled- out nationally to cover all 18 132 LGAs. LGSP transformed the LG finance system – roughly 20 percent of the total Tanzania Development budget goes to the local level, half of which is though LGDGs. Component 2 CIUP (Component 2a) successfully delivered community infrastructure and improved services per its objectives. The project was carried out in a consultation with the community – and adapted a low-risk approach that minimized safeguards impacts and was appropriate for the implementation capacity of the time. The project was effective in targeting the urban poor (impacting over 300,000 persons) living in unplanned settlements in more ways that originally intended (e.g. health and education outcomes identified through the IE). There remains strong demand from counterparts and communities for further engagements. The project successfully introduced the upgrading concept and improved the capacity of the three municipalities to implement such operations. GoT has requested to scale-up CIUP through the Dar es Salaam Metropolitan Development Project (under preparation). The Revenue Enhancement Program (Component 2b) – despite numerous challenges during implementation – facilitated the identification and valuation of all taxable properties in the DLAs. The original target was to carry-out the valuation of 100,000 properties starting from a baseline of 40,000 – however, by the end of the Project, 425,610 properties were inspected and valued. Results for “increasing the tax rate� could have been better, but this was a political issue outside the direct influence of the Project. For evaluation purposes, a project’s development objectives encompass both those stated in the project documents together with the key associated outcome targets. In the case of restructured projects in which key associated outcome targets are formally revised, the evaluation is based on the average grade (and associated numeric scale) of separate outcomes before and after restructuring, weighted by the level of disbursement at the time of restructuring. In this case, approximately 54 percent of the Credit had been disbursed at the time of restructuring. Table 4: Outcome Rating Assessment PDOs (A) (B) (A) + (B) Before Restructuring After Restructuring Final 1. Strengthen fiscal decentralization MS MS & improve management of the (4 x 54%) = 2.16 (4 x 46%) = 1.84 4.00 intergovernmental transfer system 2. Increase access to infrastructure MS MS and series in unplanned areas of Dar (4 x 54%) = 2.16 (4 x 46%) = 1.84 4.00 es Salaam Weight (% disbursed before/after 54% 46% 8.46 / 2 = 4.00 PDO change) Overall Rating - - Moderately Satisfactory 19 3.5 Overarching Themes, Other Outcomes and Impacts Poverty Impacts, Gender Aspects, and Social Development The results from the impact evaluation (section 2.2) identify some of the positive effects on the targeted beneficiaries of Component 2 for community upgrading. CIUP was effective in targeting the urban poor as intended at design – this due primarily to the early identification of the low-income communities and adherence to the selection criteria and methodology. As noted above, benefits beyond those initially intended, including positive health and education impacts, were an added benefit. For example, the CIUP interventions are associated with an overall increase in student school enrolment from 85% to 89%, with particular focus on girls over 13. 4. Assessment of Risk to Development Outcome Rating: Moderate The Project’s Moderate risk rating reflects the combined risk to the development outcome - at evaluation – of the three primary interventions carried out under the Project. The risks that the gains and systems supporting decentralization will not be maintained by the government are moderate. The government remains committed (demonstrated by the new ULGSP PforR) to decentralization, though there are risks in the ability to maintain past levels of effort outside a project based approach. There are also risks to the overall LGCDG system, which has relied more than originally anticipated on donor funds (which are now ending) while GoT has not taken on the funding commitments earlier envisioned. There are Significant risks to maintain the development outcome of the CIUP/upgrading work across the board as O&M is not adequately funded. Furthermore, unless the upgrading concept is continued as expected during appraisal, the gains made in improving access under Phases 1 and 2 will be lost. The Revenue Enhancement Work has moderate risks to the development outcomes, as the systems and tools are in place; the tax base has been expanded, capacity has been improved, and strong incentives are in place to at least maintain current efforts. Expanding the tax rate will be a challenging key next step, – but it does not put at risk the development outcomes. 5. Assessment of Bank and Borrower Performance 5.1 Bank Performance Quality at Entry Rating: Moderately Satisfactory At appraisal, the proposed budget was reduced from the USD90.3 million (estimates as basis for preparation) to USD52.2 million due to IDA-13 budget constraints. The design was adjusted to accommodate the unexpected change and also to prepare for future scale- up through additional financing. The project was designed to develop the long-term institutional infrastructure and capacity to support decentralization. Preparation of Component 1 was coordinated closely with the Development Partners, and designed to leverage the various sources of financing to support LGAs through a comprehensive and consistent mechanism. Participating LGAs (originally 20 47) were identified through established criteria, and operations manuals were prepared for appraisal. Component 1 utilized the existing GoT fiscal framework. Component 2 (CIUP and enhanced revenue collection) focused on key operational and urban management competencies supporting the DLAs. Component 2a, CIUP, followed a set methodology to target and select poor communities for upgrading, and the project concept was vetted by these communities. Implementation and capacity constraints were assessed, resulting in a lower-risk upgrading approach that did not include trunk infrastructure (and the associated resettlement). The requirement for counterpart contribution (DLA and communities) was justified as a sign of demand, but did not result in working O&M arrangements. Overall, the CIUP upstream analysis, implementation arrangements, and design were well conceived and based on lessons from prior Bank operations. The analytical work for preparation of the revenue enhancement activities (Component 2b) was developed through a team of international and national experts. The preparation team was transparent about the risks with this activity – ‘noting the substantial risk that it would not be effective.’ The analytical foundations of the Project were well prepared, strategically relevant, and closely coordinated with a broad array of stakeholders. The risks, mitigation measures, and benefits were assessed and discussed openly during preparation. The MS rating reflects the ambitious elements of the design. The M&E design and implementation was problematic, given that the targets set were overly ambitious or difficult to monitor, and do not seem to fully capture the full impacts/results of the project. The budget reduction introduced at appraisal (due to the IDA shortfall) was not ideal, however the task team and GoT made swift changes and contingency arrangements, that allowed additional financing to be restored at a later stage. Quality of Supervision Rating: Moderately Satisfactory The Project encountered operational challenges in-line with the anticipated risks. The Bank’s responses were generally adequate, timely, and proactive. The quality of supervision documentation was satisfactory. The Aide Memories were detailed, tracked the progress of follow-up recommendations, and identified key issues and risks. The ISR ratings tended to focus on the overall positive state of the Project, which overshadowed some of the technical/sector specific issues that required attention as raised in the Aide Memoires. In some instances (for FM, M&E, counterpart funding, and safeguards), ISR ratings may have been higher than warranted. The diverse mix of interventions made the Project challenging to supervise. However, the intensive level of the Bank’s supervision effort was a significant contributing factor to the positive outcomes. The presence of field-based TTL allowed the Bank to closely monitor and respond to implementation issues. Overall, the Bank’s supervision was responsive to the technical demands of both: (i) the civil works under CIUP; and (ii) the less formulaic interventions in capacity building and institutional development. Supervision was also well coordinated with Development Partners. 21 Overall Bank Performance Rating: Moderately Satisfactory 5.2 Borrower Performance Rating: Moderately Satisfactory Overall, the government (PMO-RALG), two Project Support Teams (PST PMO-RALG for Component 1 and 3; and PST-Dar for Component 2), and Dar es Salaam LGAs did a reasonable job supervising and implementing an innovative and complex project. The notable achievements of the borrowers’ performance included:  Strong leadership of PMO-RALG in (i.) coordinating development partners and leveraging resources to support the decentralization agenda, (ii.) coordinating LGAs involved in the program. PMO-RALG’s capacity grew, and now it is the focal point for urban interventions with all development partners.  A high level of responsiveness to the requirements of preparing a Bank project, including: (i) quality and timely analytical work for each component as basis for appraisal, (ii) feedback from communities/beneficiaries on the project design, (iii) preparation of operations manuals; and (iv) a good partnership in carrying-out the Impact Evaluation.  The project largely achieved its overall objectives due to the client’s adaptable approach (overcoming the reduction in the IDA budget at appraisal, and persevering through the institutional challenges of the revenue enhancement component, etc) and collaborative working relationship with the Bank and Development Partners. There were some shortcomings (detailed in section 2.2) in the government’s performance during implementation, notably:(i.) lack of contribution of counterpart funds for CIUP and ineligible use of IDA funds, (ii.) poor timeliness and quality of M&E systems and assessment of outcomes, (iii.) poor contract management, and (iv.) the low level of GoT contribution to the LGCDG systems and transfer of funds to non-qualifying councils. The project design was complex (piloting new initiatives), so overall the borrower did a solid job implementing an ambitious program. 6. Lessons Learned Key operational lessons from LGSP were incorporated into the design of the new (PforR) Urban Local Government Strengthening Program. These include: (i.) distinguishing between urban and rural local governments due to the different financial and capacity needs, (ii.) using independent assessors (for minimum access conditions and other indicators/reviews) to increase transparency, (iii.) providing a financial incentive for central government to disburse funds on time, and (iv.) adopting the same collaborative and comprehensive preparation practices. 22 A low-risk approach to upgrading achieves results. The Project adopted a low-risk approach for upgrading with minimal trunk infrastructure and resettlement. Given this was the DLA’s first upgrading project and the known capacity constraints – this was a prudent approach. The Impact Evaluation shows improved access to services, and even health and education benefits. The differences in some areas (local business development, self- investment in housing, incomes, reduction in flooding, etc) between the CIUP communities and control groups is not as striking as might be expected, particularly compared to other upgrading projects with trunk infrastructure. Nonetheless, the low-risk approach was appropriate given the complexity, counterpart capacity, and available resources. Now that the DLAs have experience with upgrading, future initiatives have room to be more ambitious and to try to improve coverage and connectivity to trunk infrastructure. Thinking beyond infrastructure and looking at human development factors is key to successful upgrading. The Impact Evaluation highlighted the expected improvements in infrastructure service. However, the gains in health and school enrolment for girls were not expected, and the reasons for the positive impacts are not fully understood. This highlights the far-reaching impacts of upgrading, and also points to the potential for further gains if the involvement of human development specialists is planned during both the design and implementation stages. Counterpart contributions may not be appropriate for a project of this nature, particularly for upgrading. Counterpart contributions were required at all levels for all components – the rationale at project design was to demonstrate demand and commitment. However, it is not clear that counterpart contributions increased ownership. Instead this became a constant issue discussed during implementation and added to the complexity to the project. Counterpart contributions were often paid late and delayed works/project activities; or were not paid at all raising FM challenges. As the counterpart contributions (5 % of capital costs) required from poor communities had no direct link with O&M requirements for the investments, this raises questions regarding appropriateness of arrangements for the collection and use of funds. An unintentional contribution to elite capture may have also been triggered (as land/house-owners who paid the counterpart contribution tended to dominate the planning process over their renters.). 7. Comments on Issues Raised by Borrower/Implementing Agencies/Partners (a) Borrower/implementing agencies The Borrower provided an ICR Executive Summary, which is included in Annex 7. (b) Cofinanciers N/A (c) Other partners and stakeholders The draft ICR was circulated for review and comments to all development partners/ local government donor coordination group – no comments were received. 23 Annex 1. Project Costs and Financing (a) Project Cost by Component (in USD Million equivalent) Actual/Latest Appraisal Estimate Percentage of Components Estimate (USD (USD millions) Appraisal millions) Component 1. Support for Local Government Capital Development 35.00 113.28 323.0 Grant System Component 2. Dar es Salaam Upgrading and Institutional 17.30 17.30 0 Strengthening Component 3. Support to PMO- 6.60 8.40 127.0 RALG Total Baseline Cost 58.90 156.70 Physical Contingencies 0.80 0.80 0.00 Price Contingencies 1.10 1.30 118.0 Total Project Costs 60.80 158.80 Front-end fee PPF 0.00 0.00 .0 Front-end fee IBRD 0.00 0.00 .0 Total Financing Required 60.80 158.80 259.0 (b) Financing Appraisal Actual/Latest Type of Estimate Estimate Percentage of Source of Funds Cofinancing (USD (USD Appraisal millions) millions) Borrower Counterpart 6.50 3.90 .00 International Development 52.00 150.00 .00 Association (IDA) Local Govts. (Prov., District, City) of Counterpart 2.30 4.90 .00 Borrowing Country Total 60.80 158.8 24 Annex 2. Outputs by Component PDO Objective 1. To strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfer systems; Indicator Baseline End target Actual 80 percent of participating LGAs meet 0% 80% 97% minimum access conditions for accessing (FY 2010/11) LGCDF by FY 2009/10 80 percent of participating LGAs with 20% 80% 54% ‘clean’ (i.e. unqualified) audit certificates (FY 2010/11) by FY2009/10 Transfers to participating LGAs are 0% 100% In 2008/09, 83.3% of transfers to LGAs effected within the first 30 days of each were made within the first 30 days of each quarter of each FY by FY 2009/10 quarter. However, in FY 2009/10 during the July –Dec 09 period, only 44% of the Note: IDA contribution to LDGD system transfers were made within the first 30 ended in Dec 2010. days of the quarter PDO Objective 2. to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operations and maintenance Indicator Baseline14 End Actual target Access to Services is improved: (a.) the 25% of 80% 97% increase in the proportion of streets, (b.) Phase I and (FY 2010/11) the reduction in households experiencing 7% of Phase flooding; (c.) percentage increase of II enterprises. 22% of 10% Figures not collected at end of the project. Phase I and Alternative measures, discussed in section 20% of 3.1 of the ICR from the Impact Evaluation Phase II are highlighted households 1,838 in 30% Figures not collected at end of the project. Phase I and Alternative measures, discussed in section 1,429 in 3.1 of the ICR from the Impact Evaluation Phase II are highlighted Dar es Salaam local Authorities (DLA) Own source 50% 151% Own Source revenue = 41.0 billion in revenue performance; (a.) % increase in revenue = July 2011 to May 2012, per June 19, 2011 DLA won-source revenue by FY2011/12; Tshs 16. (b.) % unrestricted transfers and DLA billion revenues 6.65% 8% 11.9 % 14 The Baseline indicators are reported, though were not based on the same data as the Impact Evaluation data – thus final results cannot be compared to the original baseline and are no relevant. 25 Annex 3. Economic and Financial Analysis The Project Appraisal Document contains an economic analysis of Component 1 (Support for Local Government Capital Development Grant System) and Component 2a (Community Infrastructure Upgrading Program), and a financial analysis of Component 2b (Local Revenue Enhancement / O&M Enhancement Program). 15 The cost of these components accounted for 86% of the total project cost at appraisal, and for 93% at project completion (Table A3.1). For each of the three (sub- )components, this annex summarizes the results of the economic or financial analysis in the PAD, and a re-evaluation of the results at the time of project completion. Table A3.1: Cost and Type of Economic or Financial Analysis Conducted by Component Cost (US$ million) Cost (% Total) Type of Analysis Component Appraisal Actual Appraisal Actual Conducted 1 Support for Local Government 35.0 113.3 58 71 Economic Capital Development Grant analysis, System qualitative 2a Community Infrastructure Economic Upgrading Program analysis, } 17.3 } 35.0 } 28 } 22 quantitative 2b Local Revenue Enhancement / Financial O&M Enhancement Program Analysis 3 Support to PMO-RALG. 6.6 8.4 11 5 None Contingencies 1.9 2.1 3 1 TOTAL 60.8 158.8 100 100 Source: World Bank Economic Analysis of LGCDG Results of analysis at appraisal. The Local Government Capital Development Grant (LGCDG) System would provide unconditional capital grants to Local Government Authorities (LGAs), and gradually replace a system of central government grants earmarked for specific sectors or specific areas. Part of the LGCDG grants would be based on the performance of LGAs in planning, budgeting and governance. At appraisal, it was assumed that the proposed system would generate three types of economic benefits: 1. Reduced transaction costs. The PAD acknowledges that the transaction costs of LGAs would remain “approximately equal� under the new system, but assumes that transaction costs of central government agencies and development partners involved in the administration of sector-specific or area-specific grants would be eliminated. 15 No economic or financial analysis was undertaken for Component 3, which largely consists of investments in technical assistance (the economic benefits of which are largely indirect and therefore difficult to assess). 26 2. Improved allocation of scarce resources by individual LGAs. The LGCDG system would provide LGAs with a much higher degree of freedom to allocate capital grants across sectors than the pre-existing system. Assuming that LGAs have better information about local needs than central government agencies (and would use this information for planning and budgeting), they will allocate capital grants to investment projects with a higher economic benefit/cost ratio than would otherwise be the case. 3. Improved allocation of scarce resources across LGAs. Part of the LGCDG grants will be channeled to LGAs that perform best in planning, budgeting and governance. Such LGAs are likely to better allocate capital grants than other LGAs, thereby further increasing the economic benefit/cost ratio of the LGCDG grants. Because investments financed by LGCDG capital grants were not pre-defined but demand-driven, it was not possible to conduct a quantitative economic analysis of the system. Re-evaluation of results at completion. Upon completion of the project, there was reason to believe that the LGCDG grant system had indeed resulted in improved allocation of scarce resources at the LGA level. However, the reduction in transaction costs and the impact of the performance-based element in the grant formula seemed less significant than envisaged at the time of appraisal. More specifically: 1. Reduced transaction costs. For two reasons, the PAD appears to be overly optimistic about the reduction in transaction costs that attributable to the LGCDG system. Firstly, the management and administration of the system requires more involvement from LGA officers than the relatively simple and straightforward system of sector-specific and area-based grants (this is likely to lead to better decisions about the allocation of scarce resources, but at an incremental transaction cost to the LGA). Secondly, the transaction cost at the central government level will not be eliminated completely, partly because the implementation of the LGCDG system will require management and coordination at the central level, but also because some of the sector-specific grants have remained in place (as was indeed envisaged at appraisal). 2. Improved allocation of scarce resources by individual LGAs. To estimate the net economic benefits of improved allocation of capital grants at the LGA level, it would be necessary to estimate the net economic benefits of investments projects financed before and after the introduction of the LGCDG system (any systematic increase in such benefits could be attributed to the new capital grant mechanism). Because of time and resource constraints, it was not possible to conduct this type of analysis. There is, however, reason to believe that the LGCDG system has improved resource allocation at the LGA level. This is because of a change in the sectoral composition of LGAs capital budgets since the sector-specific and area- based grant systems were largely replaced by the LGCDG system. In many LGAs, the share of grants allocated to sectors with generally high economic benefit/cost ratios (notably roads, drains and water supply) increased at the expense of investments in sectors with traditionally lower economic benefit/cost ratios (notably government offices). However, this effect was to some extent negated by central government interference in the sectoral allocation (because of an instruction of the Prime Minister, over 40% of total grants were allocated to the construction of secondary school buildings during FY2004/05 - FY2010/11). 3. Improved allocation of scarce resources across LGAs. The economic benefits from this source are likely to be limited, given fiscal incentives have weakened since the performance -based grant system was introduced. Since FY 2009/10 over 90% of LGAs qualified for 100% of the 27 performance-based portion, so that the system no longer makes a meaningful distinction between poor-performing and well-performing LGAs.16 In spite of the limited impact of the LGCDG system on transaction costs and a dilution of the performance-based grant element, the economic benefits of improved allocation of resources at the LGA level are highly significant, so that the component as a whole has, in all likelihood, generate substantial net economic benefits. Economic Analysis of CIUP Results of analysis at appraisal. The Community Infrastructure Upgrading Program (CIUP) financed improvements to basic infrastructure in 31 settlement areas in Dar es Salaam, which had a combined population of approximately 330,000 in 2004. At the time of appraisal, the economic rate of return (ERR) of the CIUP was estimated at approximately 20%, well above the Bank’s minimum required rate of 12%. The net present value (NPV) of the economic benefits of the program was about TSh 3.95billion (or US$3.6 million equivalent). The quantifiable economic benefits consisted of: 1. Reduced costs of flood damage. These were estimated as the sum of reductions in: (i) cleaning and repair costs borne by households (TSh 4,000 per household per year), and (ii) road maintenance costs (3.7% of investment cost per year). 2. Reduced health costs. These were estimated as the sum of reductions in: (i) costs of treatment of waterborne diseases (TSh 4,524 per household per year, which was equivalent to about US$4.5 in 2003), and (ii) income lost due to waterborne diseases (about US$4.7 per household per year). 3. Increased economic growth. At appraisal, it was estimated that CIUP-financed improvements to basic infrastructure would cause the establishment of two new business enterprises per hectare, and each enterprise would add value to the economy of TSh 5,000 per day (or TSh 1.5 million annually, assuming 300 business days per year). The net present value of economic benefits from increased economic growth accounted for 73% of the total NPV of the CIUP, followed by the reduced cost of flood damage and reduced health costs (which, taken together, accounted for the remaining 27% of the total NPV). Table A3.2: Type of Economic Benefits and Key Assumptions for CIUP Type of Economic Benefit / Key Appraisal Actual Comment Assumption 1 Reduced costs of flood damage Reduced cleaning and repair cost per HH 4,000 – No reduction relative to control per year (TSh) group* Reduced road maintenance costs (% 3.7 – Road maintenance cost is zero investment) in base case (no reduction possible) 2 Reduced health costs 16 Evaluation of LGSP Support for the Local Government Development Grant System. DEGE Consult in association with The Urban Institute. Dar-Es-Salaam. 2012. 28 Reduced costs of treatment of 4.5 42.3 Assumed values at appraisal waterborne diseases per HH per year substantially lower than (US$ equivalent) generally accepted estimates of Reduced income due to waterborne 4.7 42.3 reduced per capita health diseases per HH per year costs** (US$ equivalent) 3 Increased economic growth Additional new businesses established 2.0 0.0 No increase in business forma- (#/Ha) tion relative to control group* Added value per business per year (TSh 1.5 1.5 m) Source: World Bank * See impact evaluation for details ** Assumption based on: (i) 50% of unweighted annual per capita losses in Cambodia, Indonesia, the Philippines and Vietnam, corrected for general price inflation during 2003-2008, and (ii) an average HH size of 4.5. Re-evaluation of results at completion. In 2013, the Bank completed an impact evaluation of the CIUP.17 The evaluation was based on an extensive survey of 1,860 households, spread out over 31 settlement areas covered by CIUP (“treatment area�) and 12 settlement areas not covered by the program (“control area�). The same households were surveyed in 2006 (baseline) and 2012 (endline). The impact evaluation suggests that CIUP-financed interventions have resulted in major improvements to public health, but not in measurable changes to flood damage or business formation. With respect to the assessment of economic benefits, the key findings of the evaluation can be summarized as follows: 1. Reduced costs of flood damage. Despite CIUP-financed improvements in drainage, the proportion of households in the treatment area reporting that they experienced flooding in the past 12 months remained approximately 25% in both 2006 and 2012. Moreover, in 2012 households in the treatment area were 13 percentage points more likely to be flooded than households in the control area (even though households in the control area had not benefited from improvements to drainage infrastructure). The impact evaluation points out that these results should be considered with care, “since flooding is an inherently geographic shock�, and some areas may be more flood-prone than others due to their location and average rainfall. 2. Reduced health costs. The impact evaluation identified a major reduction in two major waterborne diseases (diarrhea and acute respiratory illness), the prevalence of which decreased at much higher rates in the treatment area than in the control area. This effect was limited to children; no impact was found for adults. In the treatment area, the incidence of diarrhea decreased from 22% in 2006 to 12% in 2012, and the incidence of ARI dropped from 27% to 18% in the same period. These decreases coincided with a reduction in the severity and duration of both diseases. Moreover, the PAD assumed unusually low per capita health costs (US$4.5 per household). In August 2008, the Bank published a comprehensive study on the economic benefits of improved sanitation in four Southeast Asian countries (Cambodia, 17 From the Ground Up: An Impact Evaluation of the Community Infrastructure Upgrading Programme in Dar Es Salaam. World Bank, 2013. 29 Indonesia, the Philippines and Vietnam).18 This study estimated the annual per capita health costs of poor sanitation at approximately US$9.4 per capita in constant 2003 prices. This is equivalent to about US$42 per household per year (assuming a household size of 4.5), which is nine times higher than the amount assumed in the PAD. 3. Increased economic growth. In 2012, about 57% of the labor force in the treatment area was self-employed, a substantially higher share than in the control area (44%). However, this difference was already present at baseline, indicating no change between the two groups over time. The impact evaluation therefore concluded that “CIUP interventions … did not have significant impacts on employment, business operations, income and expenditure, [or] private investment�. In contrast, the PAD anticipated that over 70% of the net present value of the project would be caused by increased economic growth. On the positive side, the impact evaluation also mentioned a series of economic benefits that were not anticipated at the time of appraisal, but are likely to have been caused by CIUP -financed investments and may improve economic growth in the long term. These most important of these are improved security, and improved female school participation rates. In summary, the economic analysis that was prepared during appraisal appears to have overestimated the economic benefits from reduced flood damage and increased economic growth, and underestimated the benefits of reduced health costs by a significant margin. Because the spreadsheet model that was prepared in 2003 to make these calculations is no longer available, it was not possible to re-assess the net economic benefits of CIUP directly. An alternative model was therefore used to re-assess the benefits of this component. Based on a series of highly conservative assumptions, the re-estimated ERR of 15% is lower than the ERR that was estimated at the time of appraisal (20%), but remains above the Bank’s minimum required economic rate of return of 12% (Table A3.3). 18 Economic Impacts of Sanitation in Indonesia – A five-country study conducted in Cambodia, Indonesia, Lao PDR, the Philippines, and Vietnam under the Economics of Sanitation Initiative, World Bank, Jakarta, August 2008. 30 Annex 4. Bank Lending and Implementation Support/Supervision Processes (a) Task Team members Responsibility/ Names Title Unit Specialty Lending Supervision/ICR Solomon Alemu Consultant AFTUW Zhiyu Jerry Chen Young Professional YPP Bella Lelouma Diallo Sr Financial Management Specia AFTFM Theresa Marissa J. Gamulo Procurement Analyst AFTUW Rumana Huque Sr Urban Spec. AFTUW Rosemary Mukami Kariuki Sector Leader AFTUW Mwanaisha Kassanga Program Assistant AFCE1 Jane A. N. Kibbassa Senior Environmental Specialis AFTEN Rowena J. Martinez Consultant AFTUW Faith-Lucy Matumbo Team Assistant AFCE1 Barjor E. Mehta Sr Urban Spec. AFTUW Donald Paul Mneney Senior Procurement Specialist AFTPC David Mayala Mulongo Urban Specialist AFTUW Elisa Muzzini Economist SASDU Nathaniel Paynter Water & Sanitation Specialist TWIWP Uri Raich Sr Urban Spec. AFTUW Mercy Mataro Sabai Sr Financial Management Specia AFTFM Rildo Santos Language Program Assistant AFTUW Zara Inga Sarzin Senior Urban Development Spec AFTUW Helen Z. Shahriari Sr Social Scientist AFTCS 31 (b) Staff Time and Cost Staff Time and Cost (Bank Budget Only) Stage of Project Cycle USD Thousands (including No. of staff weeks travel and consultant costs) Lending FY01 46.87 FY02 164.44 FY03 139.26 FY04 439.55 FY05 257.11 FY06 0.00 FY07 0.00 FY08 0.00 Total: 1047.23 Supervision/ICR FY01 0.00 FY02 0.00 FY03 0.00 FY04 0.00 FY05 149.26 FY06 295.10 FY07 215.43 FY08 281.90 Total: 941.69 32 Annex 5. Beneficiary Survey Results (if any) N/A 33 Annex 6. Stakeholder Workshop Report and Results The DIME Impact Evaluation team held a workshop on January 29, 2012 with key stakeholders (DLAs, PMO-RALG, NGOs, press, and other key ministries) to present the draft CIUP Study. An estimated 50 persons participated in the participatory workshop, and provided feedback on the draft, focusing on qualitative interpretation of the statistical data. 34 Annex 7. Summary of Borrower's ICR and/or Comments on Draft ICR Tanzania Local Government Support Project (LGSP), 2005-2011 [Component 1/3 Only] 1. Introduction In order to improve the allocation of local development funds, the Government of Tanzania (GoT) sought to introduce a Local Capital Development Grant System that would form a single, harmonized, objective and transparent mechanism to provide development grants to the LGA level. This vision was expressed in the Letter of Sector Policy on Fiscal Devolution of the Budget and Local Government Capacity Building (15th October 2004) that envisioned that “the LGCDG system will be established as the main modality through which funds for capital investment are transferred from central government and donors to LGAs. Ultimately the LGCDG system procedures and modalities will be the vehicle for the transfer of all development funds to LGAs, including sectoral programmes.� The introduction and strengthening of the LGCDG was funded under the Local Government Support Programme (LGSP) from 2005 through 2011. LGSP consisted of three components:  Component 1: Support for Local Government Council Development Grant System  Component 2: Dar es Salaam Upgrading and Institutional Strengthening,  Component 3: Support to PMO-RALG In addition to Government contributions and funding from the World Bank, LGCDG has been supported by contributions from the Netherlands, Finland, Ireland, Germany (KfW), Japan, Sweden and Belgium. 2. Background to the LGSP The formulation of LGSP was driven by two elements of the Government of Tanzania’s (GoT’s) overall development approach: the Poverty Reduction Strategy and the Local Government Reform Policy. Poverty Reduction Strategy. The Poverty Reduction Strategy Paper (PRSP) outlined the Government’s priority sectors in which improved service delivery was identified as vital to the overall goal of poverty reduction. These sectors included education, health, roads and water – the main sectors for local government capital investment. As a result, LGSP was designed to support investment in these priority sectors by providing capital to LGAs for expansion and rehabilitation of infrastructure, and targeted funds for Dar es Salaam LGAs to improve infrastructure in unplanned urban areas. Local Government Reform Policy. The Government has been pursuing a well-articulated policy of decentralization by devolution since 1996 that aims for the transfer of powers, functions and resources to elected multifunctional local governments. The implementation of the Local Government Reform Policy is coordinated by the Prime Minister’s Office – Regional Administration and Local Governance (PMO-RALG). Since 2000, local government reforms have been supported through a dedicated Local Government Reform Program (LGRP) that has aimed to support the reform process through a basket-funded program managed by a dedicated reform unit (the LGRP Team). 35 As a result of the ongoing decentralization reforms, local governments in Tanzania have important public service delivery responsibilities, including primary education, basic health services, water provision, agricultural extension and livestock promotion, roads, as well as the provision of other urban and local public services and infrastructure. In 2004, roughly one fifth of total recurrent public sector expenditure in Tanzania took place at the local government level. At the same time, however, a much smaller portion of the development budget was arriving at the local level. Prior to the introduction of the LGCDG, most ‘local’ development spending in Tanzania was provided through area-based programs (ABPs) or managed through sectoral development programs anchored with different central line ministries. Despite the fact that many of these programs were funded by the same development partners, local development resources were thus provided through a fragmented and uncoordinated system of intergovernmental finance. 3. Project Objectives LGSP’s project objectives were to: (i) to strengthen fiscal decentralisation, improve accountability in the use of local government resources, and improve management of intergovernmental transfer system and demand-driven urban investments; and (ii) to increase access to infrastructure and services in unplanned areas of Dar es Salaam and to improve revenue performance for sustainable operation and maintenance.19 A Development Credit Agreement (DCA) was signed between the Government of Tanzania and the World Bank for LGSP on January 19, 2005. The original project credit was for US$52 million with a closing date on June 30, 2008. The project received additional financing on August 30, 2006 which led to the total credit under LGSP of US$150 million. In 2009 the project was restructured to allow completion of works in Dar es Salaam and to fund preparation of the Tanzania Strategic Cities Project as well as extend the closing date to June 30, 2012. The project objectives were at that time revised with the inclusion of the underlines text above. 4. Project Components and Objectives The LGSP consisted of three components: Component 1: Support for Local Government Council Development Grant System – This component supported the introduction of a Local Government Development Grant (LGDG) system, which also received support from the Government of Tanzania (GoT) and other development partners. The core if the LGCDG system included two intergovernmental transfers: (a) the Council Development Grant (CDG), and (b) the Capacity Building Grant (CBG). The CDG provided discretionary cross-sectoral capital transfers, which were distributed on a formula basis to Local Government Authorities (LGAs) for investment in accordance with local needs as determined through local participatory planning and budgetary processes. The CBG provided resources to LGAs to help them to build their capacity to access and manage the CDG. Subsequent to the initial launch of the initial launch of the LGCDG, “sectoral grant windows� were added to the LGCDG system. In 2008, the name of the system was shortened to the Local Government Development Grant (LGDG) system. 19 The words underlined above were added to enable funding of the (preparation and implementation of the) proposed Tanzania Strategic Cities Project (which subsequently became effective in 2010). 36 Component 2: Dar es Salaam Upgrading and Institutional Strengthening, had two sub-components:  Sub-Component 2(a): Community Infrastructure Upgrading Programme (CIUP). This sub- component aims to improve access to infrastructure and services for roughly 20% of the population of Dar es Salaam who reside in unplanned areas.  Sub-Component 2(b): Local Revenue Enhancement and O&M Programme. This sub- component aims to significantly increase own-source revenues of Dar es Salaam City Council and the three Municipal Councils of Kinondoni, Ilala and Temeke to enable improved budgeting and implementation of Operation and Maintenance (O&M) programmes in the city/municipalities. Component 3: Support to PMO-RALG – This component supported the Prime Minister’s Office - Regional Administration and Local Government (PMO-RALG) in the implementation, monitoring, evaluation and audits of the Project, and the intergovernmental transfers supported by Component 1 while building capacity of PMO-RALG to execute the functions as part of routine activities. This component further supported preparation activities of the Tanzania Strategic Cities Project (TSCP). 5. Project Management A LGSP Project Support Unit (PSU) located within PMO-RALG initially coordinated the implementation of the LGCDG during the initial years of the project (under LGSP Component 1). However, in line with the objective of strengthening permanent institutional arrangements for fiscal decentralization, at the end of June 2007, the management of the LGCDG was mainstreamed within PMO-RALG under the Assistant Director for Local Government (LGA Finance). Two committees were established for technical oversight and policy guidance: 1. The LGDG Steering Committee provides policy oversight over the LGDG system. It is a senior-level inter-ministerial steering committee composed of the Permanent Secretaries from PMO-RALG, Finance, and relevant sector ministries. The role of the LGDG Steering Committee is to ensure that participating LGAs meet the required eligibility criteria; to approve the grant allocations in accordance to the allocation formulas; to make final decisions on appeals made by LGAs and to discuss and approve overall policy matters including revisions of formula, manuals, etc. 2. LGCDG Technical Committee, chaired by Deputy PS PMO-RALG, includes relevant heads of departments from PMO-RALG as well as the Ministry of Finance, relevant line ministries, participating DPs, ALAT, and other relevant stakeholders. The role of the LGDG Technical Committee is to review work plans, budgets and progress reports prepared by PMO-RALG; review the annual performance assessment of LGAs; and make recommendations to the LGDG Steering Committee, as appropriate. LGDG Quarterly technical Reviews have been conducted on a rotating basis in different regions, where different stakeholders in the field jointly review progress and issues related to the LGDG. In addition, a LG Capacity Building Consultative Group was established to review management and coordination of capacity building activities. During the first few years of LGSP, both the Steering Committee as well as the Technical Committee met on a quarterly basis and functioned effectively in shaping the LGDG system. However, as the implementation of the LGDG became more routine, the LGDG TC and SC have met only once per year in order to approve the national synthesis report and LGDG budget. In addition the LGDG Quarterly Technical Reviews have for the last two years only met twice a year rather than quarterly. 37 6. Project Cost and Financing Over the course of the LGSP project period, the core LGDG provided roughly US$ 350 million in capital development finance to the local government level, in addition to approximately $30 million in CBG funding. Roughly 25 percent of the total grants provided under the LGDG system were funded from IDA resources, whereas the government contributed approximately 17 percent. The remainder of the cumulative grant pool was provided by DPs. Table 1: Core LGDG - Disbursements by source, 2004/05 – 2011/12 (actual amounts, US$ millions)20 GOT IDA- LGDG DP-funded Core LGDG Financial Year allocations funded universe allocations (Total) (LGDG) allocations 2004/05 47 4.5 4.5 2005/06 113 5.3 18.2 16.0 39.5 2006/07 121 4.7 24.6 16.1 45.4 2007/08 121 2.1 41.6 14.7 58.3 2008/09 132 4.4 41.6 22.2 68.2 2009/10 132 33.5 43.0 15.3 91.7 2010/11 132 16.7 46.7 12.5 75.9 Total for period 66.5 220.1 96.8 383.5 Sources: Cumulative Annual Assessment Report 2004/05-2006/07 (July 2008); World Bank LGSP AM (March 2009); Annual LGDP Reports (2008/09 – 2010/11). 7. Project Results Improvements in local public service delivery and local governance Through its funding of the LGDG, LGSP has had a major impact on improved local infrastructure and local public service delivery in Tanzania. Since its inception in 2004/05, approximately US$350 million has been transferred in Council Development Grants under the Core LGDG in order to fund local infrastructure. Additional details on the local infrastructure projects that were funded under LGDG and other specific project results include:  Approximately 50% of local investments have been implemented by HLGs (District Councils and Municipalities and Town councils) and approximately 50% by LLGs (based on IPF allocated at ward level). The actual implementation of LLG projects were typically undertaken by Village Governments, School committees or other user committees. 20 For further details, see the "Evaluation of LGSP Implementation (Component 1 and 3) to Inform the Preparation of the Project Implementation Completion Report (ICR), June 2012." The table provides an indicative (rather than authoritative) break-down of disbursements by funding source, as historical project documents were not always consistent with regard to different disbursement levels by funding source. With regards to GoT contributions, the table only reflects direct LGDG grants provided by the GoT to LGAs through its regular budget processes, and does not reflect direct contributions of GoT to the LGSP Account (which was used to project operation and implementation costs in addition to grant funding). 38  The projects have in particular funded construction of secondary schools (43% of total expenditures) but also a large number of investments for LGA administrative purposes (20% for ward offices, village offices, etc.), health facilities (8%), primary education (8%), works (85), agriculture and livestock (4%) and smaller investments in water sector and other sectors. The expenditure patterns reflect to a large extent LGAs remaining funding priorities after allocation of significant sector grants for water, health, and agriculture under the LGDG system.  Value for money audits have been undertaken twice during the lifetime of the LGSP and indicate that approximately 80% of all investments are of “good� or “best� value. Only 1 - 2% of investments have been categorized as “poor value�.  Citizen perception studies form the independent REPOA indicate increased satisfaction with LG services and LGA operations during the early years of LGDG implementation (unfortunately similar data are unavailable for later years of the LGDG), Improvements in LGA capacities  The capacities of LGAs have improved significantly over the years as reflected in the increasing number of LGAs that fulfil minimum conditions and score high in annual performance assessments,  The introduction of the CBG has been established a modality for LGA planned and managed capacity building which will ensure that capacity building is directed in accordance to locally defined priorities. LGSP has supported the establishment of the overall framework for CBG delivery, including the development of ten standardized training modules have been developed: 1. LGA legislation, roles and responsibilities 2. Management and leadership skills in LGAs 3. LGA budgeting and budget management 4. LGA financial management and control 5. Procurement and contract management 6. LGA revenue mobilisation and O&M budgeting 7. LGA development planning and strategic planning 8. Project preparation, investment appraisal, environmental impact assessment and safeguard policies 9. Monitoring and evaluation 10. Data collection, information management and record keeping  Lower Level Local Governments have been significantly involved in planning, budgeting and actual implementation of the LGDGs, which has contributed to enhanced citizen participation in LGA affairs,  LGDG has introduced regular advertisement in newspapers of fiscal transfer and the assessment system provides incentives to LGAs that disseminate LGA finance data in villages and wards – in this manner the LGDG has spearheaded a process of greater local transparency of LGA budget and expenditure data, Improvements in fiscal decentralization and intergovernmental fiscal relations The most significant successes took place early on during the implementation of LGSP. In fact, the project’s main accomplishment was arguably achieved already in 2005 by establishing the LGCDG system as the main intergovernmental transfer modality to local governments, and by getting multiple DPs to agree to fold their area-based programs into the LGCDG system. In particular the Irish and Dutch programmes assisted with this transition. DP contributions were subsequently collected into a Common Basket Fund, from which point the management and funding flows of the LGCDG system fully took place through government budget systems. Performance assessment of 39 LGAs have been undertaken on an annual basis since 2004 and have – in particular in the early years - worked effectively to improve the quality of local government management. The increased commitment to fiscal decentralization by Governments stakeholders and DPs became evident by the fact that several “sectoral windows� were established from 2005 onwards that were structured along the same principles as the core LGDG into a wider LGDG system. This policy process was guided by the initial Government Letter of Sector Policy (October 2004), the amended letter of sector Policy (March 2006) and later the LGDG Memorandum of Understanding (2008). Performance indicators LGSP relies on three key performance monitoring indicators related to Component 1/3 of the project, notably: (1.1) the number of LGAs receiving Capital Development Grants; (1.2) the number of participating LGAs with “clean� audit performance records; and (1.3) the proportion of resources transferred to LGAs which are made within the first 30 days of each quarter increased. Although project performance fluctuated between years, Table 2 captures the best performance outcomes that have been achieved during the LGSP project period.21 Table 2: Summary of Project Outcome Indicators Project Outcome Indicators Baseline Outcomes under 2004/05 LGSP 1. Percentage of Participating LGAs meeting minimum 53.2% 97.7% conditions for accessing the LGDG. Target: 80 %. (25 of 47 LGAs) (129 of 132 LGAs) 2009/10 2. Percentage of Participating LGAs with “clean� (i.e. 55.3% 80.6% unqualified) audit certificates. Target: 80% (26 / 47 LGAs) (100 / 122 LGAs) 2006/07 3. Percentage of transfers to LGAs that are made within 0% 83.3% the first 30 days of each quarter. Target: 100%. (2008/09) With regard to the first performance measure (percentage of participating LGAs meeting minimum conditions), for the past five years, all but a handful of LGAs in Tanzania have met the Minimum Conditions for the LGDG system. In order words, the performance of the local government system has exceeded the targeted threshold (80%) consistently since 2007/08. The success of this outcome indicator is partly related to capacity building efforts but more significantly driven by LGAs incentive to perform in annual assessments. Local financial management performance has improved considerably over time, with the number of LGAs receiving adverse audit outcomes dropping substantially from 66 in 2000 to no more than 4 during any year since the launch of LGSP. Nonetheless, LGAs in Tanzania have had a harder time living up to the second outcome indicator: the percentage of participating LGAs with “clean� (i.e. unqualified) audit certificates. Whereas the target for this performance measure was set at 80% in 2004, the actual performance of LGAs on this measure has fluctuated around 50% since 2004/05. In 21 See the annual LGDG Progress Reports for more detailed discussions on performance on the project indicators from year to year. 40 fact, during 2006/07 this performance indicator was met, with 80.6 % of LGAs obtaining clean audit opinions. However, the fact that LGAs did not attain the targeted threshold during the other years is not necessarily indicative of poor performance of local governments in terms of financial management and accountability. In hindsight, it was an unrealistic indicator of local financial management, particularly as the CAG over the years has tightened the requirements for “clean� audit certificates. Finally, the percentage of transfers to LGAs that are made within the first 30 days of each quarter also fell short of the performance targets set at the beginning of LGSP. There are several reasons for these shortcomings related to DP partners concerns over reporting on past releases and Government’s challenges of revenue generation. Nevertheless a system of regular (albeit delayed) and formula based fiscal grant transfers to LGAs has been achieved; in 2008/09, 83 percent of transfers to LGAs were made within the first 30 days of each quarter. Overall attainment of the project objective The overall project development objective “to strengthen fiscal decentralization, improve accountability in the use of local government resources, and improve management of intergovernmental transfers� has been achieved to a significant degree, although this achievement has not been fully-captured by key performance indicators. LGSP/LGDG is the first and only modality giving formula based, discretionary development resources for the prioritization of LGAs. The need for such as grant system is now well accepted, and almost 20 percent of the development budget is now being provided to the local government level (of which approximately half through the LGDG system, including its sector windows). The overall Government budget structure has also been transformed since 2004 in a manner that allow for more transparency in the budget allocations that are being made to LGAs. 8. Major Implementation Issues Encountered The LGSP MTR (2008) identified some weaknesses in project implementation, particularly in the downstream monitoring and reporting of local sub-projects. In addition, it was identified that GoT’s contributions to the system were effectively funnelled alongside LGCDG system (on a non-formula basis), rather than being a true part of the LGCDG system. As a result, the second phase of the project was focused on achieving sustainability of the LGCDG system by addressing some of the program’s weaker features and by integrating the grant system further into regular government systems. The requirements contained in the LGDG MOU (2008) in order for other grant modalities to become part of the LGDG system (including requirements that allocations have to be performance- based and formula-based), seem to have discouraged other sector windows from fully joining the LGDG system. This has had the unintended effect of creating a somewhat of an artificial separation between “the LGDG system� and other (both government-funded and DP-funded) local development grants. In addition, it seems that by the final years of LGSP, changing political realities and macro-fiscal imbalances limited the ability of the GoT to make good on its desire to increase the overall level of funding to the local government level on a formula-basis through the LGDG system. Under significant fiscal constraints, the GoT in 2011/12 reverted to providing each LGA a small lump-sum grant, rather than following the regular LGDG allocation formula. 41 Although there continue to be basic administrative and cash-flow management challenges, local government development grants are now disbursed as grants on a quarterly schedule, rather than on an ad hoc basis or as a residual at the end of the financial year. 9. Project Sustainability During the first three years of LGSP, the LGDG system was implemented in a more or less projectised manner as a way to increase the access of local governments to development resources. At the central level, a Project Support Unit managed the project, while the project’s budget was lodged within PMO-RALG’s budget as another project. The LGDG has since progressed from “a project� to an integral part of GoT financing modalities in LGAs. The management of the LGDG was mainstreamed within PMO-RALG, and the LGDG grants were moved to the regional budget votes (within the same subvote as other local government grants). The core LGDG (the CDG and CBG) continues as the firmest and most consistent anchor of local government development grants. The GoT co-funding of the LGDG has progressed faster than originally envisaged in the 2003 design, although less consistent than envisioned in the 2008 LGDG MoU. The GoT realizes that future financing at the current level of LGDG funding will be a significant challenge. Co-financing from DPs is likely to decline due to dynamics in the global development community just as the government’s ability to provide contributions to the LGDG is limited as its faces macro-fiscal pressures. GoT is intent on ensuring that stakeholders engage each other in order progress with its fiscal decentralization agenda including the continued funding of LGDG at sustainable levels. This may include revision of the size of the LGDG grant pool compared to the 2008 MoU. The integrated planning and budget process at LGA levels is the primary tool to ensure O&M of LGDG funded infrastructure at local level. Independent Value for Money Audits undertaken for PMO-RALG (2008 and 2012) and the MTR in 2008 noted that there is scope for improvements O&M of completed facilities. The need is primarily in the form of further clarification and documentation of O&M procedures. In can be noted that the GoT throughout the last decade has ensured consistent and robust increases for funding of O&M through the recurrent grant system and LGA own revenue increases Both aspect of sustainability: sustainability of the wider intergovernmental fiscal transfer system and the specific sustainability of LGA funded investments deserve further attention in subsequent years of LGDG implementation. 10. Key Lessons Learnt The LGSP was essentially designed to demonstrate how a fiscal transfer system could operate and was initially in many aspects operated as a “project� with separate project unit etc. Nevertheless, Government commitment ensured a rapid and successful mainstreaming of core LGDG operations into a dedicated PMO-RALG LG Directorate. However, the wider ambition of reforming the fiscal decentralisation also requires strong commitment from MOFEA and improvement of selected operation procedures of its Regional and Local Government finance section. In hindsight the project might have benefitted from more significant involvement of MOFEA and specific activities that would support MOFEA in improving its operations related to management of fiscal transfers. Another important lesson is that LGAs are very willing to improve their performance in accordance to the annual performance assessments. These assessments are extremely valuable for PMO-RALG, 42 LGAs and other stakeholders as guidance to where capacities should be improved. There is a need to continue and strengthen these assessments. 11. Bank Performance Design Project preparation, design and quality at entry were satisfactory. The formulation of LGSP included an extensive and highly participatory design process, which drew lessons from similar World Bank operations and related international practices and built on the experiences of many area-based programs previously supported by bilateral DPs. Project preparation was highly responsive to GoT’s own vision for the intergovernmental transfer system, as there was considerable support among the political leadership and within the highest technical levels for a more decentralized public sector. During the initial project preparation stages, it was foreseen that the LGCDG system would be piloted on a small subset of LGAs for several years, but GoT decided that it sought a nation-wide development grant system and -with additional DP support- was able to upscale the LGCDG system to all LGAs at an early stage. Ahead of the finalization of the design of LGSP, the GoT had a strong voice in shaping the LGDG system by preparing a Letter of Sector Policy on Fiscal Devolution of the Budget and Local Government Capacity Building (October 2004) which envisioned that a “LGCDG system will be established as the main modality through which funds for capital investment are transferred from central government and donors to LGAs. Ultimately the LGCDG system procedures and modalities will be the vehicle for the transfer of all development funds to LGAs, including sectoral programmes.� Implementation The Bank provided instrumental partnership in the implementation of LGSP in terms of flexibility, as evidenced in the constant review of the PIP to accommodate new thinking and internal budget revisions, and agreeing to review and reschedule components in light of new developments and opportunities. This includes in particular the project restructuring in 2009 that allowed added emphasis on development on urban LGAs. Reviews and Supervision Bank supervision and implementation reviews were quite adequate and regular. It took the form of hands-on support without necessarily interfering with the day-to-day management of the programme. The Bank posted in the later period of LGDG a resident TTL, which helped in terms of coordination with regular monitoring activities such as participation on LGDG Technical committee meetings etc that also had to be coordinated with a significant number of participating bilateral DPs. The Bank provided technical guidance on new initiatives – in particular for development of the urban sector. 12. Borrower’s Performance Design The design of the LGSP Component 1 and 3 (the LGCDG) was undertaken with significant technical depth and over a long period by resident consultants that enabled a very participatory process. PMO-RALG was leading Government activities in the design phase but also ensured significant inputs from other stakeholders such as the LGAs, MOFEA, CAG and sector ministries. The Government developed a comprehensive letter of sector policy (signed by MOFEA) that outlined the broader fiscal decentralisation policy within which the LGSP and LGDG would operate. 43 Implementation Major implementation successes were noted at LGA levels as around 385 million USD of core LGDG was transferred for funding their local priorities. PMO-RALG successfully undertook efforts to build the capacities of LGAs and undertook core functions related to the management of the LGDG system – in particular management of annual performance assessments that were used as basis for grant allocations. However, challenges were faced in ensuring timely and reliable transfers of the grants to LGAs and subsequent robust reporting on fund transfers and utilisation. 13. Conclusion LGSP satisfactorily met its development objective: the project strengthened fiscal decentralization, improved accountability in the use of local government resources, and improved management of intergovernmental transfer system. In particular, the project has provided significant discretionary local development resources and introduced concepts and models for formula based grant allocations, fiscal incentives for LGAs to engage in good governance practices, participatory planning within known budget constraints, locally managed capacity building modalities, and decentralization of fiscal resources within HLG to LLGs. However, the LGDG system is currently also facing certain challenges, including difficulties in adhering to the allocation formula, difficulty in achieving improved reliability in the completeness and timeliness of transfers, and challenges in achieving complete and timely reports regarding local fund utilization. It is increasingly recognized that fiscal decentralization is a process rather than an end goal in itself. Tanzania’s progress on fiscal decentralization reforms is naturally linked to the wider decentralization reforms that arguably have developed along a different path in recent years than foreseen in the 1998 Local Government reform Policy. The need to consolidate the gains made under LGSP through adopting a more holistic approach to fiscal decentralization is pertinent. Continued efforts to improve fiscal decentralization and local financial management under LGSP II and the inclusion of local government finance issues in the recently developed PFMRP Phase IV (May 2012) are important next steps in this process. 44 Annex 8. Comments of Cofinanciers and Other Partners/Stakeholders The draft ICR was circulated for comments to Development Partners. No comments were received. There are no Co-financers. 45 Annex 9. List of Supporting Documents Independent Procurement Review and Value for Money Audit of Contracts Implemented by Local Government Authorities under the Local Government Development Grant System. FYs 2007-2011. June 2012. Prepared by Upimac, at the request of PMO-RALG. Evaluation of LGSP Support from the Local Government Development Grant System (Component 1 and 3). Reference LGSP/PE/26/10. Prepared by DEFE Consult in association with the Urban Institute, for PMO-RALG. Supervision Mission Aide Memoires (all) from the first on March 11, 2002 for Preparation; to the MTR on Jan 14, 2008, to the final on January 25, 2010 for Implementation Support. Implementation Status and Results Report (all 15) from Sequence 1 (04/26/2005) to Sequence 15. PRELIMINARY DRAFT: From the Ground Up; An Impact Evaluation of the Community Infrastructure Program in Dar es Salaam, Aidan Coville, World Bank and Yu-hsuan Su, World Bank/University of Washington. Nov, 2012. CIUP Stage Report: Community Environmental Management Plan, July 2004.UCLAS & DHV Consultants. Local Government Support Project Revenue Enhancement: Draft Final Report, Sept 17, 2004, by Robbert Faber. Mid-Term Review of the Local Government Support Project: Component 2; Prepared by NCG, GHK, and Howard Humphreys, Jan 2, 2008, for PMO-RALG. Mid-Term Review of the Local Government Support Project (LGSP), Component 1&3, by NCG, GHK, and Howard Humphreys, Dec 8, 2007. Mid-Term Review of the Local Government Strengthening Programme (2009-14) Draft Report, Nov 2012, by the DPG and PMO-RALG. Impact Evaluation of Infrastructure in Ethiopia and Tanzania: Final Report, by Maximo Torero, International Food Policy Research Institute. The Impact of Local Government Reforms in Tanzania 1998-2008. By Per Tidemand and Jamal Msami. Research on Poverty Alleviation. 2010. 46 Annex 10. Summary Community Infrastructure Upgrading Program Impact Evaluation Summary The World Bank makes substantial investments in slum upgrading projects around the world, with the broad objectives of alleviating poverty, improving health and well-being and strengthening the social fabric and economic opportunities within these communities in a holistic and integrated manner. While the intrinsic value of improved living conditions is clear, policymakers can benefit from understanding more precisely which dimensions and to what degree upgrading programs can influence livelihoods when setting policies to promote social and economic development. Rigorous research is beginning to emerge on the effects of slum upgrading on important health outcomes like diarrhea and acute respiratory illness (ARI) and crowding in private investments; however very little is known about the broader impacts of the upgrading process that serve to motivate these interventions in the first place. Isolating the causal impact of upgrading programs presents a challenge. Simply identifying a strong correlation between an intervention and an outcome of interest does not necessarily imply that the program caused the observed change in outcomes. The reason for this is clear in the case of informal settlement upgrading projects. By doing a simple “before and after� analysis, measuring the beneficiary outcomes at baseline and then during a follow up after the intervention has been implemented, any change observed in the outcome could be the result of any number of external factors that may have influenced these individuals during the period between baseline and follow up. To measure the true impact of the CIUP we need to identify a counterfactual or “control� group of households having not received the program that proxies for what the upgraded, or “treated� households would have been like in the absence of the program. The more plausible the control group is for comparison, the more plausible any impacts measured using this control group as a basis are. Since the choice of upgrading sites has both practical and political motivations, finding a control group that mimics these selection criteria is critical for generating plausible impact results. As a gold standard, randomized controlled trials (RCTs) ensure comparability between control and treatment groups by randomly selecting who receives treatment (the intervention) and who does not. However, these are often infeasible for upgrading projects for practical implementation reasons in which case alternative “quasi-experimental� methods need to be used. A recent Cochrane review (2013) 22 identified 5 rigorous slum upgrading evaluations globally with no clear consensus on the results, highlighting the importance of research such as this to help fill an important knowledge gap as well as feeding into improved project design. This impact study uses a quasi-experimental approach to evaluating the Community Infrastructure Upgrading Programme (CIUP) financed by the World Bank with the aim of improving the lives of slum dwellers in Dar Es Salaam through targeted investments in community infrastructure such as roads, drainage systems and streetlights. The study makes use of a point system used by the Dar Es Salaam City Council to identify and allocate CIUP interventions to mitaa (sub-wards) in the city and matches the 31 mitaa selected for upgrading with a comparable control group of 12 mitaa receiving the same scores but not being offered the 22 Turley, R, Saith, R, Bhan, N, Rehfuess, E, Carter, B (2013) “Slum upgrading strategies involving physical environment and infrastructure interventions and their effects on health and socio-economic outcomes�, Cochrane Review 47 intervention. A baseline survey of 1860 households in control and treatment mitaa was conducted in 2006, upgrading began in 2007 and the follow up survey was completed in 2012. Stakeholder focus groups were held to contextualize preliminary quantitative results. A group of 26 participants convened for a working session in Dar Es Salaam on 28 January 2013 to engage on the preliminary results. The session included partners from the Dar City Council, National Bureau of Statistics, implementing engineers, local universities and NGOs and the World Bank. The focus groups gathered information on (i) potential explanations for seemingly anomalous results; (ii) suggestions for the inclusion of impact measures that had not been covered in the original set of results; and (iii) ways in which the work could feed into policy debate and project design in the future. Results presented in this report reflect these discussions. Figure 1: Distribution of surveyed households (Control = Yellow; Treatment = Red) We find that the CIUP interventions increased household sizes and decreased out-migration, halved diarrhea rates for children under 5, and increased female school enrolment rates, but did not have statistically significant impacts on employment, business operations, income and expenditure, housing prices, private investment or social cohesion. The rest of this note expands on these results, but first highlights the limitations of the study to ensure that results are interpreted appropriately. 48 Caveats Although careful identification of suitable control sites for the purpose of comparison and rigorous econometric methods underlie the results presented, it is important to interpret the results with consideration for their limitations, including issues related to (i) sample representivity; (ii) the inherent heterogeneity of the interventions provided to communities and (iii) the validity of the control group to be used as a comparison from which to estimate impact. The CIUP was designed as an inclusive, holistic and participatory approach to the upgrading process in Dar Es Salaam, with the objective of providing important public infrastructure to unplanned settlements lacking these facilities. A total of 31 mitaa were supported through the program providing (i) paving of trunk roads, (ii) footpaths; (iii) public toilets; (iv) drainage systems; (v) streetlights; (vi) pedestrian crossings; (vii) upgraded water kiosks and (viii) solid waste containers. Community meetings were held on a regular basis from the project inception until completion, with the particular focus of eliciting local female participation in the decision process of upgrading priorities. The results presented reflect the outcomes identified in this program, but should not necessarily be seen as outcomes to be expected for in situ upgrading programs more broadly. It is important to note that the program primarily focuses on public infrastructure, and does not cover interventions commonly found in other in situ upgrading programs. In particular, land titling, water and sanitation provision, and social development programs were mostly absent from the CIUP. The program also makes provision for variations in the package of interventions provided based on the community needs. As such, although all treatment mitaa participated in the CIUP, how each mitaa (or household) benefited could vary considerably across the intervention areas. The results reflect only the surveyed mitaa, and are not representative of Dar Es Salaam. Enumerator Areas (EAs) were randomly selected from within mitaa, but mitaa were selected purposefully based on where the CIUP interventions were taking place and which control mitaa best resembled these groups. As such, inferences should not be made about other mitaa in Dar Es Salaam that were not part of the study. High migration rates and an inability to track the same households over time means that results should be interpreted as changes to communities and not changes to households. Various regression models have been included in the analysis in an attempt to disentangle improvements in location (through selective in-migration) vs. improvements for original dwellers. The large migration over the 6 years (over one third of households) and the selective migration (higher migration rates in control areas) make these analyses challenging. However, results are still generated in many cases for (i) overall change in community characteristics; (ii) in-migrant characteristics; and (iii) non-migrants, or original residents to provide a comprehensive overview. This summary report reports on (i) only, whereas a more comprehensive paper to be published at a later stage will include (ii) and (iii). On the validity of the control group, a careful selection process identified comparison mitaa that were as similar as possible to treatment mitaa; however some baseline differences across groups should be kept in mind when interpreting results. The GoT initially considered all 310 unplanned settlements across the 3 districts in Dar Es Salaam. Scores between 23 and 33 were provided to each area, and mitaa were then deemed eligible for upgrading if they reached a score of 30 or more. This reduced the number of eligible mitaa to 63. A restriction was placed on the selection of eligible mitaa to ensure that 300Ha were to be upgraded in each of the three districts, with a total of 900Ha, and any mitaa falling below 10Ha in size would be combined with a neighboring mitaa to be considered for upgrading to benefit from economies of scale. The selection 49 process generated 31 mitaa that were chosen for participation in the CIUP. These clear selection criteria ensured that the process was transparent and void of potential political interference. The process also helped to identify candidate mitaa to be considered as control group comparisons for the purposes of the evaluation. Since a rigorous ranking system was conducted, all mitaa meeting the same ranking criteria were likely to be similar across a range of measures, and households within these mitaa were likely to be following the same growth trajectory. Baseline results, however, showed some important differences between the two groups before any intervention took place that should be kept in mind when interpreting the results. For the treatment/control baseline comparison, we find that household demographics, housing ownership, income and wealth, school enrollment, access to services and dwelling characteristics were similar across the groups in 2006; however, employment, health and length of residence differed significantly. The typical household had 3.8 household members, with a 39 year old household head having 8 years of education and a 75% chance of being a male. On average, 0.9 children per household were enrolled in school. Employment, while very high, was slightly lower in the treatment area, with 87% of household heads employed compared to 91% in the treatment area (most of whom were “self-employed�). Household monthly income hovered between 140 000 and 150 000 TZS (approximately $100) which is not significantly different across groups. Treatment households had stayed in their dwellings for slightly more than two years longer than control households and were more likely to have paying tenants on their property, although the proportion of households owning their dwelling was indistinguishably different across groups, at around 38%. Ownership of individual assets differed across groups but not systematically, providing no strong indication that one group had a broader asset base than the other. Health outcomes measured at baseline suffered from poor data quality, which makes it difficult to use this information; however, the available data indicate that diarrhea rates were lower for treatment households, while the incidence of acute respiratory illness was the same across groups at baseline. Services and infrastructure across groups in 2006 was relatively similar, with 45% – 48% of households having electricity connections and walking 5 – 6 minutes to access their primary water source. Housing structures did not resemble the corrugated iron shacks usually synonymous with unplanned settlements, and 94% of households had a cement floor and 97% of dwellings had walls made of brick or cement. Dwellings in the control area were, however, slightly bigger, with an average of 2.7 rooms, compared to 2.4 rooms in the treatment area. Results Extensive physical infrastructure was provided in treatment mitaa, together with community outreach meetings reaching nearly 10 000 women. A total of 65 community events reaching out to 15 827 people in CIUP wards were held to ensure full community participation in the upgrading process. The project aimed to have at least 50% female participants at these events, and records show that this outreach was effective in achieving a 58% female participation ratio. The project completed 40km of two-way roads, 58km of one-way roads, 26km of footpaths, 132km of road side drains, 8.6km of trunk drains and 26km of footpaths. A further 548 culverts, 3153 pedestrian crossings, 83 solid waste containers and 2780 street lights were provided during the upgrading process. The provision of public toilets was minimal, with a total of 24 being provided across the entire upgraded area. 50 Table 1: Completed CIUP activities based on project monitoring data Total Phase 2 Phase 1 TWO WAY ROADS Km 40.3 14.9 25.5 ONE WAY ROADS Km 57.8 33.9 23.9 FOOTPATHS Km 26.3 10.7 15.6 ROAD SIDE DRAINS Km 132.5 63.6 69.0 CULVERTS Lines 548 319 229 PEDESTRIAN CROSSING Number 3153 1020 2133 DRIFTS Number 564 166 398 SOLID WASTE CONTAINERS Number 83 35 48 STREET LIGHTS Number 2780 1630 1150 PUBLIC TOILETS Number 24 17 7 TRUNK DRAINS Km 8.6 3 5.6 Treatment mitaa experienced significantly more upgrading activity than control mitaa during the CIUP intervention period, but only in CIUP-related interventions. We find no difference between control and treatment areas in health, schooling, sewerage, electricity, or land regularization programs. Since none of these were part of the CIUP package of intervention, this serves only to confirm that other interventions are unlikely to confound our results (i.e. other programs were equally distributed across control and treatment areas). We do, however, see strong, positive impacts in self-reported knowledge of community interventions taking place directly related to CIUP, such as toilets, garbage collection, internal roads, water drainage and street lights. Responses on “water supply� provide no differential between control and treatment areas; however, we explore this further by breaking water supply into: (i) piped water; (ii) yard tap; and (iii) water kiosks. As expected with the CIUP interventions, a significant increase in upgrading support to water kiosks is observed. Interestingly, however, control households report having received upgrading to their piped water supply that was not provided in treatment areas. This upgrading translated into better access to infrastructure for the average household. Distance to a water supply decreases significantly to 1.5 minutes (from 3 minutes currently in control), and toilets from 13 to just under 4 minutes. Households report streetlights being available on their roads up from 2% in the control area to 46% in the treatment area, reducing time to reach a street light from 15.5 minutes to 3 minutes. Access to water drainage directly outside dwellings increased from 14% to 37% and time to point decreased from 7.5 minutes to 1 minute. Finally, households indicating that the access road to their dwelling was “somewhat improved� jumped from 7% to 48%. Garbage collection systems were already relatively high in the control area at 71%; however household access to a collection system increased further to 91% through CIUP. Other community facilities such as schools and health clinics saw a slight reduction in time taken to reach location, but this is insignificant. Ostensibly, this may result from the improved pathways and roads leading to a reduction in time reaching a facility even when new facilities have not been included in the CIUP intervention, and, as we have seen previously, any potential confounding health and education interventions had not been implemented differentially across treatment and control areas. The graph presents these results, with a red star indicating that the result is statistically significant. 51 Although large infrastructure investment resulted in increased access for households, the sustainability of this infrastructure poses a challenge. Of the households that reported having streetlights on their street outside their home, 48% of these people reported that the lights were not working. Similarly, 29% of households reported their drainage systems as not in working order. However, garbage collection activities (94%) and community roads (98%) were mostly seen as functioning in the way they were expected to. This raises important questions about the sustainability of community infrastructure and the need for ongoing support from local government moving forward. Sustainability of these interventions must be borne in mind when interpreting reported impacts. The impact of improved infrastructure on flooding is ambiguous and complicated by the recent extreme flooding in Dar Es Salaam in December 2011. The endline survey was conducted from April to June 2012. The worst flooding in Dar Es Salaam in 50 years was experienced in Dec 2011, with informal settlements particularly affected. Since flooding is an inherently geographic shock, the comparison groups (in other locations in the city) should be considered with care. Data from the surveys indicate that treatment mitaa were more susceptible to flooding, with 22% of Figure 2: Time taken to reach faciliy 18 16 14 * * Minutes 12 10 8 * 6 4 CIUP 2 Control 0 * households having reported have their home flooded in the past 12 months compared to 18% in the control areas. Given that large improvements in drainage systems were recorded, the result seemed inconsistent and became a core part of the focus group discussion held in January 2013 to identify potential reasons for this. It was found that the 2011 floods had placed such a strain on bulk infrastructure that some of the bulk drainage systems (not part of CIUP) burst, releasing substantial flooding into a few select mitaa. Without providing information to Government officials about the outliers identified in the data, these officials were requested to submit a list of mitaa that were, by their records, affected directly by this bulk infrastructure flooding. A list of 5 mitaa was provided and analysis was conducted with and without these mitaa. It was found that, indeed, 31% of households in the identified “outlier� mitaa reported having a flooding incident compared to 18% in the rest of the study mitaa. When removing these outliers, control and treatment wards report identical levels of flooding (18%). While, overall, the amount of flooding in treatment areas is higher than in controls there are two considerations to make: (1) Despite the extensive flooding, the proportion of households in the treatment area reporting that they experienced flooding in the past 12 months remains at 22% in 2006 and 2012; and (2) Although no reductions were found in the likelihood of being flooded, the severity of flooding in CIUP mitaa was mitigated. We find that 52 69% of households experiencing flooding in control areas needed to make repairs, while this reduced to 58% of households in treatment areas. Households reporting flooding as a major problem also decreases slightly from 57% in the control area to 52% in the treatment area. Household sizes increase as a result of the upgrading process. Household size increases significantly more in treatment areas over time (from 3.8 to 5.13 people per household, compared to an increase from 3.84 to 3.97 people in the control areas). When restricting the analysis to non- migrants in the cross-sectional analysis, the impact is even larger, indicating that this is not just a result of larger in-migrant households, but a choice of original residents in the treatment area to expand their households. This is most likely a combination of life cycle differences (treatment households tend to have lived in the area for longer and are slightly older). If purely to do with treatment households being older (and thus more likely to have children), we would expect the result to fall away when controlling for age of household head. When controlling for age and length of tenure, results reduce slightly but persist, suggesting that other channels such as potential pull factors bringing more household members to upgraded areas may be driving the observations. When breaking down the demographic shift in household size due to the increase, we see that, while the proportion of households with a spouse present does not change between treatment and control areas, the number of “relatives� increases by 55% and the number of children (of the household head) increases by 33%, which substantiates the possibility that other family members (rather than tenants being mis-interpreted as being part of the household) are joining the household and this is not all driven by increases in the nuclear family that would be the case if this was a pure “life cycle� result. Dwelling owners increase the likelihood of renting out rooms to tenants, while renters pay equivalent amounts for accommodation in treatment and control areas. Resident dwelling owners rent out at least one of their rooms 68% of the time in the treatment group compared to 48% in the control group highlighting the potential for households to leverage on the community upgrading to increase rental income opportunities. However, rental prices are similar across groups (27500 TZS per month in control areas compared to 29200 TZS in treatment areas) and total rental income per month for treatment households is not significantly different from control households. * Increased private investment in the form of housing renovations and upgrades is restricted to toilet improvements. Although there is a slight increase in the total amount of investment for upgrading the household’s dwelling over the past 6 years (up from 290 000 to 360 000 TZS – a difference of approximately $50), this is not statistically significant, the overall investment amount 53 is low, and the measurement is noisy, making it difficult to obtain more precise estimates. There is very little difference across groups on the probability of conducting any particular upgrading activity (eg. plastering, extensions, upgrading building material, etc.); however there is a significant increase in the likelihood that a household has upgraded their toilet in the last 6 years (from 10% in the control group to 18% in the treatment group), which is driven almost entirely by original residents rather than in-migrants, which could be explained by a number of possible channels. Firstly, there is some evidence of a limited sanitation marketing campaign that formed part of the CIUP intervention, and community meetings held to discuss CIUP interventions often focused on the importance of sanitation and hygiene, although based on evidence of similar interventions, it seems unlikely that the full impact could be expected to come through these interventions. It is possible, rather, that some of this impact is driven by the household size increase experienced in intervention areas, increasing the need for toilet facilities. However, this assertion cannot be verified using the available data. Employment rates and business entrepreneurship does not deviate from high values at baseline. Employment is high in both groups at close to 90%, with treatment employment rates slightly (but not significantly) lower. The majority of people are self-employed with their own business. Although there are substantially more people running businesses in the treatment areas (57%) relative to control areas (44%), this difference was present at baseline, indicating no change between the two groups over time. Average yearly household income of 3 660 000 TZS (about $2500) and expenditure (both actual and per capita) also remain indistinguishable across groups. Both control and treatment areas have thriving business activity and similar diversity in businesses; however the data available cannot assess whether there is a differential in- migration of business activity in treatment areas. We are restricted to measuring economic activity in terms of employment, entrepreneurship, rental income opportunities and private investment. This neglects the potential increase in business creation and migration into treatment areas resulting from improved public infrastructure. Without a business sample/census the best information available comes from community questionnaires answered by key informants in each of the mitaa which provide evidence of multiple business types in both control and treatment areas, but does not record differences in quantity. 70% 60% 50% 40% CIUP 30% Control 20% 10% 0% Before After Figure 4: Percentage of households running at least one business 54 Children experience large reductions in diarrhea and acute respiratory illness (ARI) but no impact is found for adults (19 and older). Focusing on the two main sicknesses associated with slums, we find an overall decrease in diarrhea and ARI incidence in the treatment area. This result, however, masks the distributional impact of the CIUP on health. Breaking the results down by age, we observe that, consistent with the literature, there is a much stronger impact on children under 5. Diarrhea almost halves, from 22% to 12%, and ARI incidence in the past 12 months drops from 27% to 18%. These illnesses are also more severe and last longer in control areas. When restricting the analysis to include only non-migrant households (implying that the child would have spent their entire life in the settlement if they stayed with the family from birth), the impacts increase even further with diarrhea rates reduced by 16 percentage points and ARI reduced by 11 percentage points. This result is in line with similar literature with regard to the distributional impact on children; however the size of the effect is substantial, especially considering the limited effect the upgrading had on, for instance, flooding. We identify that there was a difference in diarrhea rates 45% 5: 12 month morbidity rates by age group Figure 40% 35% * 30% * 25% 20% * * CIUP 15% Control 10% 5% 0% Diarrhea ARI Other Diarrhea ARI Other illness illness children <5 adults >18 between children at baseline (with CIUP wards being somewhat better off); however, while the baseline survey did collect health information, the measures are not directly comparable to the endline survey and a direct difference-in-difference estimation cannot be performed. As such, it is expected that these results be seen as an upper bound for health impacts. School enrolment increases, with the largest impact found for school-aged girls older than 13. We find that the CIUP interventions are associated with an overall increase in student (age 6 – 18) school enrolment from 85% to 89%. Exploring further, the results show that all of the impact is driven by increased enrolment in the treatment group for girls, while there is no impact on boys’ enrolment. We further break the analysis down by age and find no impact on children aged 6 – 12 (primary school education is compulsory in Tanzania). The cohort driving the observed impact are girls aged 13 – 16. 13 – 16 year old girls in the control group have a 79% enrolment rate compared to 93% for the same-aged girls in our treatment group. No such difference is seen for boys, where both control and treatment boys have a 93% enrolment rate. An impact is still observed for older children (aged 17 – 19), but this is slightly lower and only borderline significant. In this cohort, control group girls have a 62% enrolment rate compared to a 74% enrolment rate for girls in the 55 treatment group. Figure 6 highlights the stark contrast between control group females and the rest of the cohort. 100.0% 90.0% 80.0% Boys in T (89%) 70.0% Boys in C (89.3%) Girls in T (86.4%) 60.0% Girls in C (80.5%) 50.0% 40.0% 6 7-8 9-10 11-12 13-14 15-16 17-18 Figure 6: School enrolment by age and gender The large improvement in school enrolment was discussed in detail in the January 2013 stakeholder focus group. Although striking, a number of plausible reasons for this improvement were discussed: 1. Improved security, as a result of streetlights, footpaths, etc, make it safer for children to go to school. This would affect female adolescents more than their male counterparts especially when considering crimes related to sexual harassment, especially prevalent in slum areas. 2. Improved health has been shown to improve female school participation rates where health and schooling are complements, while increasing labor force participation for males, where improved health increases strength, and thus attractiveness in the (manual) labor market23. 3. Increased female decision-making power has been shown in the literature to increase household resources directed to children. Improved female empowerment, resulting from the continuous community engagement events focusing particularly on involving females in the decision making process may translate into increased decision power in the household by females and more emphasis on schooling (especially for girls). 4. An increase in the household size resulting in changing household member roles reduces the burden and opportunity costs for females (who are often the ones required to tend to household chores), making schooling relatively more attractive. We then ran tests to confirm or reject these hypotheses. For improved security, we find no correlation between the household’s perception of safety in the settlement and the likelihood that a child is enrolled in school. Similarly, we find no differential correlation between health and 23Pitt, M. M., Rosenzweig, M. R., & Hassan, N. (2010). Human capital investment and the gender division of labor in a brawn-based economy. Economic Growth Ctr Yale. 56 schooling across gender. It is not possible to directly test a relationship between the community meetings and female empowerment, however, a commonly identified result in the literature shows that increased female decision making in the household leads to a reduction in expenditure on cigarettes and tobacco and an increase in expenditure on schooling. We find consistency in the data – treatment households are 3 percentage points less likely to spend money on cigarettes and tobacco and 11 percentage points more likely to spend money on monthly schooling costs. While far from identifying a rigorous causal relationship, the results support a message that improved female empowerment induced through the community engagement may play a part in reducing high school drop out for girls. The “household size – enrolment� relationship, however, provides the strongest evidence of a potential causal pathway. We see a strong relationship between the number of children in the household and whether girls are enrolled in school. Accounting for the treatment group status, and overall household size, each extra child in the household is associated with a 2.5 percentage point increase in the likelihood that a girl child attends school; however, the number of children in the household has no relationship with boys’ enrolment. Taking this further, we find that over 50% of children that are not enrolled in school cite work as the reason for leaving school. This is even more common for females (55%) than males24. Restricting our analysis only to females that have dropped out of school, we find a borderline significant increase in the control group by 18 percentage points in selecting “work� as the reason for leaving school. This substantiates an argument that an increase in household size (induced by the upgrading) influences female school enrolment rather than male enrolment, ostensibly because of the change in opportunity costs, releasing time for females (who are the most likely to be required to help with housework, etc.) to attend school. The upgrading process keeps social interactions intact, but does not measurably improve social cohesion. The study focuses on revealed rather than stated preferences by asking questions that relate to direct actions within the community. At the first level, we would like to see if people are aware of community activities such as volunteer programs, neighborhood security groups and organized sports clubs, for instance. We then assess whether any change in awareness of community activities is accompanied with actual participation. We also explore neighbor dynamics by finding out how much people rely on (and help) their neighbors on a range of day-to-day activities such as child care, transport and job search. The results indicate no systematic significant differences across groups in any of these measures. Since the alternative to in situ upgrading usually includes household relocation that is associated with the breaking down of social and economic networks, the fact that this upgrading does not disturb the social fabric is, in itself, an important finding. However, in situ upgrading is often motivated for with the assumption that these programs will improve social cohesion in the community, which does not seem to be the case in this context. Recommendations and Conclusion This report presents a unique opportunity to fill an important knowledge gap on slum upgrading interventions. With only a handful of rigorous evaluations done on slum upgrading activities (none of these on World Bank projects that we are aware of), and the importance of finding ways to improve slum conditions in a rapidly urbanizing world, this evaluation offers insights into a broad range of impacts that similar community infrastructure upgrading projects may expect. 24 Note that “work� in this context has combined both paid and house work. 57 Despite a focus on rigor and quality, a number of caveats to the study need to be carefully included in any interpretation of the findings. With some important differences between treatment and control group at baseline, and an inability to track households migrating out of the study wards, the control group as an effective comparison for analysis can be called into question. The bias of any results is unknown, but various robustness checks, and a focus on the theory of change, combined with an ex-post qualitative review to contextualize any seemingly anomalous results helps to provide justification (and skepticism) when it is due. Strong results in access to infrastructure, household sizes, health, and increased rentals show that, even with community infrastructure interventions that do not include the full range of standard upgrading activities can still have an important impact on communities. Although it included a comprehensive community engagement component, the project was void of any specific social/health/education/titling programs that are commonly included in similar programs. We see no change in social cohesion in these communities, and perhaps this is to be expected given the context; but important impacts, such as reduced diarrhea, increased household sizes and an increase in the number of households renting out parts of their house for income increases substantially. While some intuitive results have been verified through this study, a number of unexpected results that should feed into further policy discussion have been identified. In particular, the potential pull factors drawing new household members to the upgraded communities has a number of knock-on effects that are critical for the policy debate. On the one hand, upgrading planning needs to account for the potential in-migration that results from the intervention and the implication this has on the ability for the community infrastructure to provide effectively to these populations. On the other hand, policymakers need to understand more carefully what a change in household size might mean for the ways in which households, and household members make decisions. For instance, the link between household size and school attendance, while tenuous, provides new insight into ways in which slum upgrading programs may serve the development objectives of a wider range of stakeholders. Ex ante impact expectations need to take into consideration baseline levels, and what room for improvement there is when setting targets. We see no increase in employment rates, (most) upgrading indicators and households running businesses. This may not be surprising given the high levels from which the outcomes of interest started at baseline. Employment rates for household heads were 90% at baseline, while almost all households were made with concrete walls and floors. This leaves very little room for improvement, with marginal increases becoming harder to achieve at these high baseline levels. As such, finding no impact in these indicators can be expected. There is still room for improvement: poor maintenance, low private investment and unclear economic benefits provide space for reflection on what can be done to achieve these objectives in future projects. We see that drainage systems and streetlights in particular suffer from sustainability challenges. Identifying supporting community incentive schemes, for instance, to overcome collective action problems associated with common goods could be considered to maximize potential development impacts. The project did not include any tenure security component. While it cannot be identified here, there is strong evidence that improved tenure security substantially increases private investment. The results from the study show limited private investment taking place, calling into question whether the added effort of including such a component would be justified. Other, incremental tenure options such as the provision of street addresses for instance, could be included and tested in future programs, identifying ways to stimulate private investment through improved perception of tenure security while limiting the complexity of the intervention. 58 Future research can focus on areas that this study has not been able to address adequately: namely, spillover effects and migration. It is expected that investments in community infrastructure in one area are likely to provide potential benefits to neighboring communities, through improved hygiene, security and economic opportunities that are not necessarily confined to the treatment area. If neighboring control communities are improved through such spillovers, the comparison between treatment and comparison groups would underestimate the program impact. Differential in- and out-migration also changes the households in a treatment community such that, by tracking only people that remain in the community, we do not get a clear picture of the overall program impact. Lower-income households could be forced to move out due to increase living costs, making way for higher-income groups – resulting in a false interpretation about benefits accruing to households (in this case the community may improve, but at the expense, not for the benefit of, the household). Alternatively the upgrading scheme may catalyze growth, lifting households out of poverty traps and opening opportunities to move out of the slum. These two contrasting conclusions would lead to two very different policy recommendations. This study does not include analysis of out-migrants and results should thus be interpreted as improvements to the community rather than to households. The evaluation results can now be used to stimulate debate and feed into project and policy design. Dissemination activities will continue, and culminate in a policy note and working paper that can be used as a step towards filling an important knowledge gap in slum upgrading activities. Annex 1: Summary of main impacts Indicator Control Treatment Impact Diarrhea (<5) 12% 22% 10% Household size 3.84 5.04 1.2 School enrolment (all) 85% 89% 4% School enrolment (girls age 13 -16) 62% 74% 12% Toilet upgrading 10% 18% 8% Renting rooms 48% 68% 20% Annex 2: Baseline Comparability p- Variable Treatment Control value 5% sig Demographics Male 0.49 0.53 0.00 YES Age in years 24.22 23.53 0.08 Enrolled in school 0.90 0.92 0.24 Male (HHH) 0.75 0.77 0.26 Age (HHH) 39.49 38.29 0.06 Household size 3.80 3.83 0.82 Employment employed=1 0.38 0.41 0.03 YES employed=1 HHH 0.87 0.91 0.01 YES 142772 Total household income (TZS) 149739.06 .27 0.52 Total household income (TZS, drop 128432 >650,000 = top 1%) 131388.42 .13 0.51 Education Years of education completed 6.86 6.80 0.72 59 p- Variable Treatment Control value 5% sig Years of education completed (HHH) 8.00 8.05 0.81 Age 6 - 12 school enrolment 0.90 0.92 0.24 Age 6 - 12 female school enrolment 0.90 0.94 0.14 Health Experienced diarrhea in last month 0.02 0.05 0.01 YES Experienced acute respiratory illness in last month 0.05 0.07 0.21 Tenure Length of time occupying the same dwelling (years) 8.70 6.47 0.00 YES Number of paying tenants 2.46 1.38 0.00 YES Rental per room per month for 2855.5 tenants (TZS) 5860.68 6 0.09 Increase in rent from last year for tenants (TZS) 216.82 144.45 0.25 Expected sale price of dwelling 28565960. 191408 (TZS) 00 20.00 0.04 YES 10564. Total rent (if HH does not own) 10633.08 85 0.97 Increase in HH rent since last year (TZS) 551.50 632.20 0.36 A HH member owns the dwelling 0.38 0.40 0.34 Services and structure HH has an electricity connection 0.45 0.48 0.32 9031.4 Monthly electricity charge (TZS) 8686.31 7 0.68 Number of hours per day of electricity provision 15.60 14.33 0.01 YES Number of blackouts per week 3.02 3.03 0.87 Monthly expenditure on water 6182.9 (TZS) 7585.65 2 0.08 Length of time to walk to primary drinking source 5.56 5.11 0.15 HH boils water before drinking 0.67 0.68 0.57 HH filters water before drinking 0.51 0.53 0.57 Monthly expenditure on waste collection 712.30 504.23 0.00 YES HH has a telephone 0.57 0.54 0.20 9814.8 Monthly telephone charge 10086.94 1 0.77 Dwelling has a brick/cement floor 0.94 0.94 0.90 Dwelling has brick/cement walls 0.97 0.98 0.02 YES C I.Sheet/wood rood=1 0.91 0.89 0.06 Number of rooms in HH (excluding bathroom, garage) 2.43 2.74 0.01 YES 60 p- Variable Treatment Control value 5% sig HH premises have been flooded in past year 0.25 0.34 0.00 YES Time use Time spent on activity yesterday (minutes) Work outside home 608.18 608.80 0.95 School classes or homework 253.29 300.05 0.59 Grinding grains, cooking and food shopping 147.89 126.02 0.00 YES Doing dishes, laundry, ironing, or cleaning your house 146.13 114.20 0.00 YES Collecting water 61.48 67.33 0.28 Collecting firewood 48.17 75.71 0.00 YES Doing other household chores (child caring, sick caring) 156.16 174.14 0.20 Feeding yourself (breakfast, lunch, dinner), sleeping 669.73 634.17 0.00 YES Activities not mentioned before? 212.15 216.81 0.78 Assets A HH member owns at least one of the following assets Radio 0.77 0.81 0.02 YES TV 0.33 0.35 0.28 Iron (electric or coal) 0.46 0.57 0.00 YES Stove 0.90 0.94 0.01 YES Fan 0.34 0.39 0.05 Sofa 0.22 0.23 0.88 Table 0.86 0.84 0.49 Chairs 0.73 0.78 0.02 YES Beds 0.98 0.95 0.01 YES Bicycle 0.10 0.07 0.17 Motorcycle/scooter 0.00 0.01 0.26 Car or truck 0.02 0.03 0.18 Cart 0.04 0.04 0.64 Pack animals 0.00 0.00 0.39 Sewing machine 0.10 0.06 0.01 YES Loom 0.01 0.01 0.12 Tools 0.12 0.05 0.00 YES Urban land (Hectares) 0.08 0.12 0.01 YES Rural land (Hectares) 0.16 0.08 0.00 YES Crime HH burglary in last 6 months 0.07 0.07 0.66 At least one HH member was a victim of theft in last 6 months 0.17 0.19 0.13 61