Document of The World Bank FOR OFFICIAL USE ONLY Report No: 135777-DJ IMPLEMENTATION COMPLETION AND RESULTS REPORT TF 0A4568 ON A SMALL GRANT IN THE AMOUNT OF USD 0.5 MILLION TO THE Republic of Djibouti FOR Djibouti Household Survey 2017 (P162743) {April 2019} Poverty And Equity Global Practice Middle East And North Africa Region Regional Vice President: Ferid Belhaj Country Director: Marina Wes Senior Global Practice Director: Carolina Sanchez Practice Manager: Benu Bidani Task Team Leader(s): Gabriel Lara Ibarra ICR Main Contributor: Gabriel Lara Ibarra, Vibhuti Mendiratta ABBREVIATIONS AND ACRONYMS ADDS Djiboutian Agency for Social Development (Agence Djiboutienne de Développement Social) BETF Bank Executed Trust Fund CPS Country Partnership Strategy DISED Direction of Statistics and Demographic Studies (Direction des Statistiques et Études Démographiques) EDAM Djibouti Household Survey (Enquête Djiboutienne Auprès des Ménages) FDI Foreign Direct Investment FM Financial Management GDP Gross Domestic Product GoD Government of Djibouti IFR Interim Financial Reports NSDS National Strategy for the Development of Statistics USAID United States Agency for International Development TA Technical Assistance TABLE OF CONTENTS DATA SHEET ....................................................................... ERROR! BOOKMARK NOT DEFINED. I. PROJECT CONTEXT AND DEVELOPMENT OBJECTIVES ....................................................... 3 II. OUTCOME ................................................................... ERROR! BOOKMARK NOT DEFINED. III. KEY FACTORS THAT AFFECTED IMPLEMENTATION AND OUTCOME ................................ 11 IV. BANK PERFORMANCE, COMPLIANCE ISSUES, AND RISK TO DEVELOPMENT OUTCOME ERROR! BOOKMARK NOT DEFINED. V. LESSONS LEARNED AND RECOMMENDATIONS ............. ERROR! BOOKMARK NOT DEFINED. ANNEX 1. RESULTS FRAMEWORK AND KEY OUTPUTS ........................................................... 14 ANNEX 2. PROJECT COST BY COMPONENT ........................................................................... 18 ANNEX 3. RECIPIENT, CO-FINANCIER AND OTHER PARTNER/STAKEHOLDER COMMENTS ...... 19 ANNEX 4. SUPPORTING DOCUMENTS (IF ANY) .................... ERROR! BOOKMARK NOT DEFINED. The World Bank Djibouti Household Survey 2017 (P162743) DATA SHEET BASIC INFORMATION Product Information Project ID Project Name P162743 Djibouti Household Survey 2017 Country Financing Instrument Djibouti Investment Project Financing Original EA Category Revised EA Category Not Required (C) Organizations Borrower Implementing Agency Ministry of Finance Agence Djiboutienne de Developpement Social Project Development Objective (PDO) Original PDO The PDO is to support the Department of Statistics and Demographic Studies (DISED) to collect, analyze and disseminate data emanating from the Household Survey for Social Indicators (Enquete Djiboutienne Aupres de Menages - EDAM) 2017. FINANCING FINANCE_T BL Original Amount (US$) Revised Amount (US$) Actual Disbursed (US$) Donor Financing TF-A4568 500,000 500,000 500,000 Total 500,000 500,000 500,000 Total Project Cost 500,000 500,000 500,000 Page 1 of 19 The World Bank Djibouti Household Survey 2017 (P162743) KEY DATES Approval Effectiveness Original Closing Actual Closing 02-Mar-2017 09-Mar-2017 20-Aug-2018 20-Aug-2018 RESTRUCTURING AND/OR ADDITIONAL FINANCING Date(s) Amount Disbursed (US$M) Key Revisions KEY RATINGS Outcome Bank Performance M&E Quality Satisfactory Satisfactory Substantial RATINGS OF PROJECT PERFORMANCE IN ISRs Actual No. Date ISR Archived DO Rating IP Rating Disbursements (US$M) 01 07-Dec-2017 Satisfactory Satisfactory 0.50 ADM STAFF Role At Approval At ICR Regional Vice President: Hafez M. H. Ghanem Ferid Belhaj Country Director: Asad Alam Marina Wes Senior Global Practice Director: Carolina Sanchez Carolina Sanchez Practice Manager: Benu Bidani Benu Bidani Task Team Leader(s): Gabriel Lara Ibarra Gabriel Lara Ibarra ICR Contributing Author: Vibhuti Mendiratta Page 2 of 19 The World Bank Djibouti Household Survey 2017 (P162743) I. PROJECT CONTEXT AND DEVELOPMENT OBJECTIVES Context A. Country Context 1. Djibouti has experienced high economic growth rates in the last years. Taking advantage of its geographical location, Djibouti has become mainly a re-exporter of Ethiopian products and it has started to show its potential as a logistics platform. Between 2008 and 2014, the country has had an average annual economic growth rate of 4.9 percent, surpassing the average of developing countries in the Middle East and North Africa region (2.3 percent). The main drivers behind this trend have been public investment and the continued dynamism of the transport chain (Central Bank of Djibouti, 2014). 2. Djibouti is a rent-based economy, driven in large part from rents of its strategically located port and foreign military bases (Djibouti Systematic Country Diagnostic, 2018). Thus, Djibouti’s Gross Domestic Product (GDP) growth is largely related to evolution of the tertiary sector, with 67% of its GDP coming from this sector, especially from transport services. The primary sector is pretty small mainly because of hostile weather conditions which particularly have hampered the development of the agricultural sector. In addition, only 10% of the land is cultivable (Central Bank of Djibouti, 2014). 3. The era of high economic growth was boosted by the increase of foreign direct investment (FDI). FDI inflows started in 2003 and surpassed in some years (2007, 2008 and 2013) a level higher than 20% of GDP. The FDI is mainly concentrated in infrastructure for ports, roads, buildings and hotels. 4. After growing 8 percent annually between 2013 and 2016, the rate of GDP growth is estimated to slow to 5.7 percent in 2017. Projections indicate a revival in growth in 2018, driven by newly operating railway line and improved port facilities, but there are recognized risks to stability and continued growth. Economic growth has been highly capital intensive (the railroad to and a water pipeline from Ethiopia, and several port projects) with limited impact on job creation. Furthermore, this process has also been accompanied by the accumulation of a large public debt. Public debt grew from 57 percent of GDP in 2014 to 89 percent (estimated) in 2017 in Djibouti. The most recently acquired debt has a larger component of nonconcessional loans that could put further pressure in the country’s finances in the future. Finally, Djibouti’s revenues are also highly dependent on port activities and military bases. These phenomena are not likely to be conducive of an inclusive growth path, as large-scale infrastructure and logistics investments have limited potential to create employment, and the labor opportunities it demands are mainly for skilled workers only (Djibouti Systematic Country Diagnostic, 2018). 5. Unfortunately, the observed strong economic growth has not "trickled down" and the country has seen little improvements in poverty reduction. The extreme poverty incidence has remained high even after the high economic growth rates of recent years. The official estimates show that the poverty rate has only been reduced by one percentage point in the period 2002-2013 going from 24 percent to 23 percent. The incidence of poverty in rural areas has somewhat decreased, but it remains severe as Page 3 of 19 The World Bank Djibouti Household Survey 2017 (P162743) 44.5% of the population has been classified as extreme poor. This implies that around 45 percent of the population was not able to cover its food requirements. Other indicators such as the poverty rate 1 have shown slow progress as well. Thisrate, based on a cost of basic needs approach by adult equivalent, reflects the percentage of the population that is not able to afford a minimum threshold of food and non-food needs. Nationally, global poverty was 46% in 2002 and 40% in 2013. 6. While the results seem troubling, it should be noted that there are important differences in the approach to measure welfare and poverty across these years which raise caution in their interpretation. In summary, there are two key factors that may affect the comparability of the poverty estimates. First, the modules to collect consumption information were different across years (length and details of questions used, use of recall instead of a dairy, etc.) thus allowing for some difference to come from the change in this methodology. Second, the focus of the 2013 household survey was on updating the Consumer Price Index and collected information only from urban households. To obtain national estimates of welfare and poverty, estimates for the rural population were obtained using information from a subsample of rural households collected in a 2012 household survey. This combination of data sources could also affect the comparability of the poverty estimates discussed above. Thus, a new round of a household survey presents an important opportunity to set a well thought and reliable baseline for poverty monitoring, that at the same time incorporates best current practices in measurement of welfare. B. Sectoral and Institutional Context 7. The Government of Djibouti (GoD) has shown a commitment to a strong National Statistical Strategy. With support from donors, the first National Strategy for the Development of Statistics (NSDS) was developed for the period of 2006-2010. An updated strategy was developed for the period 2011-2015 embodied in law No123/AN/1/6eme L. The most recent NSDS only partially achieved the intended objectives - with the evaluation report highlighting that many of the activities could not secure timely funding. The NSDS third objective to "Improve the global quality and frequency of statistics of Household Living conditions" was met at 25%, whereas objective 4 "Promote access to data" was only met at 41%. The GoD is currently pursuing the definition and development of the NSDS 2016- 2020 and it is expected that these objectives will still be at the fore front of the strategy. Reliable data and statistics will be key in informing the government's new programs for development and action plans. 8. The Direction of Statistics and Demographic Studies (DISED) has relied on the Enquete Aupres des Menages (EDAM) since 1996 to produce welfare and poverty indicators. This effort has undergone several improvements along the years. The first effort came with the Enquête Auprès des Ménages – Indicateurs Sociaux (EDAM-IS), conducted in 1996, where only an exploratory questionnaire was fielded with the support of the World Bank and the United Nations Development Program (UNDP). DISED has continued to expand the frontier of data collected in the EDAM-IS, and 1The poverty rate uses a threshold that includes the costs of both food and non-food necessities. Households that cannot afford (i.e. have an estimated level of per adult equivalent consumption below) the cost of this basket are considered poor. Page 4 of 19 The World Bank Djibouti Household Survey 2017 (P162743) significantly improved the data collection in 2002 when an updated sampling frame was constituted in 2001/02 to better represent the living conditions of Djiboutian households. The EDAM-IS 2002 collected information on 12 consumption components from approximately 2400 households. 9. The 2009 Census provided an excellent opportunity to update and draw better samples to understand the expenditures of households. The questionnaire of 2012 was designed to improve the detail of the data collected. Modules on food, housing and utilities, education, health, clothing, communication, transport, purchases of durable goods and other non-food collected data on a total of 107 items. Lastly, the EDAM-Budget Consommation (EDAM-BC) collected very detailed information on consumption as part of DISED’s effort to update the consumer price index on a sample of only urban households. Using the information from the EDAM-BC 2013, a new poverty line and poverty rate were estimated. Thismethodology to measure poverty in Djibouti and the poverty rates for 2013 are included in a publication jointly produced by DISED and the African Development Bank (AfDB). 10. At the request of DISED, the Poverty GP has provided technical assistance (TA) since 2015 (P156794 and P164755) to help improve the poverty monitoring system in the country. Poverty measurement in Djibouti rests on the availability and richness of data collected in the EDAM. Through this engagement the team has managed to establish a good working relationship with DISED management and technical staff. The TA activities have allowed to conduct diagnostics on previous survey instruments, welfare measurement methodologies and data collection protocols. As part of the TA, the team focused on working closely with DISED on: i) improving the survey instrument, ii) finding feasible sampling strategies to include the nomadic population, iii) reviewing and updating the welfare and poverty measurement methodologies applied to the consumption data; and iv) building the capacity to curate and disseminate the microdata produced. Theory of Change 11. The project envisaged to undertake several activities: cartographic work to update the mapping of enumeration areas as well as a listing and sampling exercise to define the households that will be surveyed in the EDAM, update of the questionnaire of EDAM 2017 to keep it in line with current good practices, development of the questionnaire in Computer Assisted Personal Interview (CAPI) format, collection of data and stakeholder consultation meetings. 12. These set of activities culminated in the completion of data collection on 4474 households across Djibouti. This output from the project was achieved in a timely manner. 13. Through the completion of data collection as well as close collaboration with DISED on the activities related to preparation of associated datasets and measurement of welfare and poverty led to three outcomes. First, using the data from EDAM 2017 and through the provision of workshops and hands-on technical assistance, the capacity on welfare and poverty measurement of DISED staff was strengthened. The team presented the current good practices on welfare and poverty measurement, which were incorporated in the production of DISED’s official Djibouti 2017 poverty rates. Second, as the analytical work culminated in the preparation of a welfare and poverty report, its findings increased Page 5 of 19 The World Bank Djibouti Household Survey 2017 (P162743) the salience of the multidimensional needs of vulnerable groups in the Djiboutian population among key stakeholders informing the dialogue on how to tackle these issues. The report was disseminated through a presentation to line ministries and development partners in June 2018 and is published online in DISED’s website. Third, Djibouti’s Statistical system has benefitted from increased confidence due to the transparency of the implementation and dissemination of the EDAM 2017 survey. For instance, the metadata from EDAM 2017 along with all the implementation documentation has been made available publicly (a first in DISED history) through DISED’s online microdata catalog. 14. In the medium run, these outcomes will help in the formulation of well-thought out policies that have an impact on poverty reduction and boost shared prosperity. The survey will enhance WB dialogue with the GoD as it will fill recognized knowledge gaps and provide the analytical underpinnings for project design and prioritization to promote inclusiveness of the economy’s growth. 15. EDAM 2017 has laid the foundation to monitor welfare and SDGs in Djibouti. Line ministries are expected to be able to draw important information on the country's population and living conditions of households - particularly of the vulnerable groups. Social indicators are expected to inform the government's Vision 2035 as well. …to generate the ...leading to our ...which will achieve We will implement these ACTIVITIES... following ultimate OUTCOMES… OUTPUTS… VISION: Sampling framework and strategy for the EDAM Strengthened Capacity of Development of the EDAM 2017 DISED’s staff to questionnaire in a Computer Assisted collect, analyze Personal Interview format and disseminate data Well Training workshops for EDAM field Completion of the informed personnel Djibouti policy Household Survey making Data collection and support to preparation Publication of the of associated datasets Poverty and Welfare Analysis Technical assistance on welfare and Report poverty measurement Public Stakeholder Consultation Meetings documentation of EDAM survey and data Page 6 of 19 The World Bank Djibouti Household Survey 2017 (P162743) C. Higher Level Objectives to which the Project Contributes 16. The proposed activity relates to both the current Country Partnership Strategy (CPS) and the regional strategy. The first pillar of the CPS, Reducing Vulnerability, has identified the improvement of social safety net programs as well as the development of human capital through the improvement of health indicators and quality of education as key challenges. The improvement of the monitoring systems and welfare measurement in the country have the potential to inform and help design the targeting of the social safety net program and to provide a solid foundation for tracking human development indicators. The activity is also in clear alignment to the Middle East and North Africa (MNA) Strategy Pillar on renewing the social contract via the improvement of monitoring of urban poverty reduction and income opportunities of women and youth. Finally, the results obtained from the EDAM 2017 were available to provide timely inputs into the evidence required for the drafting of Djibouti’s Systematic Country Diagnostic, and ultimately inform the identification of key sectors for sustainable development in the Country Partnership Framework. 17. This project is aligned with the stated objective of the NSDS 2016-2020 to be able to answer to the requirements of producers and users of statistics in order to monitor policies and programs, including the national strategy of employment and the Sustainable Development Goals. The outputs of this project are also expected to contribute directly to the country’s capacity to monitor the in dicators needed to track progress on poverty reduction and human development, a key feature of Pillar IV of the country’s Vision 2035. Project Development Objectives (PDOs) 18. The PDO is to support the DISED to collect, analyze and disseminate data emanating from the Household Survey for Social Indicators (EDAM) 2017. Key Expected Outcomes and Outcome Indicators 19. Key expected results from this project include: (a) Data collection: Successful completion of the EDAM 2017 (b) Analysis: Publication of the Welfare and Poverty Analysis Report (c) Dissemination: Archiving of the 2017 EDAM for public access in an online microdata library (i.e. World Bank's Microdata Catalog) Page 7 of 19 The World Bank Djibouti Household Survey 2017 (P162743) Components 20. The project was executed under one component: Djibouti Household Survey 2017. The activities conducted under this operation had the objective of supporting the successful implementation of the design, data collection and analysis of the EDAM 2017. II. OUTCOMES Assessment of Achievement of Each Objective/Outcome Outcome 1: Data collection: Successful completion of EDAM 2017 21. The team2 along with DISED updated the questionnaire of EDAM 2017 to keep it in line with current good practices. The questionnaire included improvements on the food items module, the durable goods section, health module, and the module on housing services and dwelling-related expenditures. A first draft of the questionnaire was shared by DISED with various partners in December 2016, including a multi-sectorial technical committee, line ministries, and other international partners. The key objective was to elicit suggestions and recommendations that could help better attain the objectives of the EDAM and capture the welfare of Djiboutian households. A meeting with the Ministries and other partners was organized on January 11th, 2017 where the engaging discussion led to further refinements of certain questions in the EDAM. The final and revised draft of the EDAM questionnaire was then used in a pre-test in the field on a very small number of households during the last weeks of January. 22. The project entailed cartographic work to update the mapping of enumeration areas as well as a listing and sampling exercise to define the households that will be surveyed in the EDAM. The sampling strategy of the EDAM 2017 incorporated two important features. The first feature that was included in EDAM 2017 is the updating of the sampling framework to include slum areas (such as the community of Balbala). These areas were expected to have experienced large increases in population, due to immigration from within the country and elsewhere, thus raising caution on the usefulness of the 2009 census as a sampling framework. An updated list for these areas then informed the framework from which the EDAM sample was drawn. A second feature was the increased fieldwork efforts to include nomad households. According to the 2009 census, the nomadic population represented between 20- 40% of the Djiboutian population. However, due to logistical and resources constraints, among others, the nomads have not been surveyed in any of the previous EDAMs. Given the likelihood of being among the most vulnerable in the country, surveying them and to the extent possible including them in the calculation of national estimates of poverty and inequality has been a great improvement from all previous EDAMs. 2The team benefitted extensively from discussions and comments received from DEC’s survey team, SP and MFM colleagues working in Djibouti. The questionnaire design also underwent the review of DEC survey team as part of the pilot of the newly established household survey protocol. Page 8 of 19 The World Bank Djibouti Household Survey 2017 (P162743) 23. The questionnaire was programmed (using CSPro) in Computer Assisted Personal Interview (CAPI) format with the help of an expert and was administered using tablets. As a result, capacity of DISED on CSPro programming increased. 24. The project entailed the provision of a 3-week training workshop and hands-on training for field personnel (including supervisors and enumerators). The field personnel were also trained in the use of CAPI to administer the questionnaire to respondents. 25. Data collection was conducted to encompass 4474 households spread across the country. Data collection was done in two rounds: May 2017 and November- December 2017. The preparation of associated datasets with information from the survey collected was done immediately after data collection. Outcome 2: Analysis: Publication of the Welfare and Poverty Analysis Report 26. Using the data from EDAM 2017, the team provided capacity building on welfare and poverty measurement for the production of the official Djibouti 2017 poverty rates. The team presented the current good practices on welfare and poverty measurement. Two important revisions to the methodology included a technically-sound estimation of the value of housing services for certain groups of the population, and the estimation of flow of services for durable goods. The methodology was jointly defined with DISED. Finally, in February 2018, a reverse mission of DISED technical staff was organized in Washington DC. Frequent meetings allowed having rapid iterations and fruitful discussions on each component in the implementation of the new methodology. These feedback loops increased the efficiency of the whole process and shortened the process to produce welfare and poverty indicators. 27. The measurement of several other indicators such as access to services, amenities, dwelling characteristics and profile of the poor were also carried out by the team and DISED after data collection. DISED presented the preliminary findings of the survey in March 2018 to a group of representatives from line ministries and international development partners. The analytical work culminated in the preparation of a welfare and poverty report. The findings of the report were disseminated through a presentation to stakeholders, line ministries and development partners in June 2018. This report was also published on DISED’s website for wider use. Dissemination of the report was also supported through a WB blog post. Outcome 3: Dissemination: Archiving of the 2017 EDAM for public access in an online microdata library (i.e. World Bank's Microdata Catalog) 28. The metadata from EDAM 2017 has been made available publicly through DISED’s microdata catalog. This exercise entailed anonymization and preparation of public use files of surveys based on internationally accepted practices with the help of World Bank DEC team. Along with EDAM 2017, metadata of surveys conducted by DISED in the last two decades were also placed for public use in the microdata library. Page 9 of 19 The World Bank Djibouti Household Survey 2017 (P162743) Overall Outcome Rating Rating: Satisfactory 29. We consider the relevance of the PDO to be high as EDAM 2017 has laid the foundation to monitor welfare and SDGs in Djibouti. Line ministries are expected to be able to draw important information on the country's population and living conditions of households - particularly of the vulnerable groups. Social indicators are expected to inform the government's Vision 2035 as well. The survey will enhance WB dialogue with the GoD to adopt policies that have an impact on poverty reduction and shared prosperity. It will fill recognized knowledge gaps and provide the analytical underpinnings for project design and prioritization to promote inclusiveness of the economy’s growth. 30. Overall, DISED has become more involved in the policy making process as many line ministries refer to EDAM 2017 to draw information. 31. We consider the efficacy of the PDO to be substantial as the project was consistent with the objectives outlined. All the results indicators have been fully achieved. 32. The implementation of the EDAM 2017 constituted an important shift in the monitoring of welfare in the country: i) the improved questionnaire follows several best practices including capturing a well- crafted list of food items that represent the consumption patterns of Djiboutian households, expanding the asset ownership module to include purchase and current value of durable goods, observing receipt of specific social protection programs, incorporating education and health expenditures better into the flow of the questionnaire, and including a module on Water, Sanitation and Hygiene services; ii) the questionnaire design benefited from an open and transparent process where line ministries were able to provide comments so that the survey could better capture policy-relevant questions; and iii) the sampling strategy was revised to be able to observe nomadic households as part of the frame – a group previously left out of the survey data collection- and to produce welfare indicators at the arrondisement (i.e. district) level within Djibouti-ville. The work carried out for the EDAM 2017 is expected to be used as a baseline to monitor welfare and poverty when subsequent rounds of household survey data are collected. 33. Through the technical support on programming the questionnaire using CSPro, data collection, sampling, training of field personnel as well as welfare measurement, the country’s statistical capacity has been strengthened on these dimensions. 34. Support to develop and publish surveys’ documentation and associated metadata online was provided. Training and capacity building to DISED staff on anonymization and preparation of public use files of surveys conducted by DISED since 1996 on their website was conducted. 35. We consider that the project’s efficiency was substantial. Data collection was executed in a timely manner and analysis of the data led to the timely production of poverty and inequality estimates (barely six months after data collection finished). It also substantially improved the measure of households' welfare based on the most updated consumption basket of the Djiboutian population. Page 10 of 19 The World Bank Djibouti Household Survey 2017 (P162743) III. KEY FACTORS THAT AFFECTED IMPLEMENTATION AND OUTCOME During the course of implementation a few factors affected the original plan to carry on the survey. Nonetheless, flexibility and close coordination with the counterpart helped minimize or eliminate the impact on the expected outcomes. 36. There was an agreement reached, in coordination with the ministry of agriculture, to use information on water points' maps to help draw enumeration areas that can frame the data collection efforts for the nomadic households. However, this data on water points was not made available. Thus, a full review of the distribution of nomadic populations across all Census EAs was carried out. Given the wide variation in the share of nomads in rural EAs, a sampling strategy that randomly selected EAs would be able to capture a good share of the nomadic population. Thus, the data collection strategy was developed without the need to incorporate data on water points. 37. The TF was implemented via a cooperation between DISED and ADDS (an agency under the Secretariat of State for Social Affairs). DISED had limited experience in handling World Bank’s managed TFs. In contrast, ADDS had already played a role of implementing agency for several World bank’s projects. Given the explicit interest to produce timely data on households’ welfare, a decision was made to have ADDS as the implementing agency leading the administrative procedures of the TF including FM and procurement. DISED would be in charge of overseeing the data collection activities. The trust built with both ADDS and DISED was important in reaching an agreement between the parties. 38. Due to bureaucratic processes, the project was launched with a delay, as were the preparation activities for the survey. However, despite this, the fields teams were able to be on the ground by April 2017. To deal with the hot season and Ramadan, the sampling strategy was revised so that the data collection was divided into two phases. Phase 1 was completed successfully in May 2017 before the onset of Ramadan. The second phase of data collection was carried out in November- December 2017. The level of representativeness as originally planned was not affected. 39. Implementation of the survey was, for the first time in a household consumption survey, carried out via CAPI using tablets. Overall, the implementation using CAPI went well, with improved timeliness and quality of the data collected. However, the newness of the application resulted in problems in transmission of a subset of data for a small number of households in two regions of the country. The pre-defined checks to the data did not capture this potential problem. After completion of Phase 1 of data collection, and at the recommendation of the team, DISED agreed to: i) go back to the field to re- interview the households for which there was missing information; and ii) establish an updated protocol to ensure that the data was checked for completeness after data collection and transmission. The overall representativeness as originally planned was not affected. Page 11 of 19 The World Bank Djibouti Household Survey 2017 (P162743) IV. BANK PERFORMANCE, COMPLIANCE ISSUES, AND RISK TO DEVELOPMENT OUTCOME 40. All planned activities were implemented and completed on time, with no project extension which points to significant efficiency. Frequent missions to the country by staff ensured that any survey issues were dealt with rapidly. In addition, there were several technical meetings over the life of each project. The Bank team maintained close communications with the grant recipients and implementing agencies through video conferences and gave prompt support as needed. The Bank’s role was enhanced by provision of a more direct, hands-on training and capacity strengthening throughout the project. 41. There were no safeguards policy triggered on this project. The project focused on data collection and dissemination. 42. The World Bank provided fiduciary and procurement support through FM and procurement specialists. The World Bank responded well to requests for support which prevented further delays. Procurement of goods and services was conducted in line with the provision of the Grant Agreement and project procurement plans that; throughout the project implementation no procurement compliance issues were raised. The implementing agency showed strong commitment and ownership of the project. 43. Overall, the project operated a sound financial management system. The project complied with the Bank’s Financial Management operational policies and procedures. Two independent audits we re performed and validated the implementation of the project’s expenses 44. Disbursement was overall satisfactory. At ISR of December 2017, the entire amount of project funds had been disbursed. V. LESSONS LEARNED AND RECOMMENDATIONS 45. The timely implementation of the project was greatly supported by an accompanying TA project funded by other sources. Through all stages (design, preparation and implementation) of the survey, the project benefited significantly from presence of staff in Djibouti. The presence of staff in the field helped establish trust, close feedback loops and established an iterative process that allowed information to flow to all stakeholders quickly. The close coordination between DISED staff and staff on the field also ensured that the data was made available by DISED soon after fieldwork finished so that quality control and further analytics could be produced jointly and in a timely fashion. 46. Hiring experts in the fields of sampling and fieldwork preparation proved to be very beneficial in an environment where capacity is not high and the time demands of the implementation partner are many. The time-intensive and presence-heavy approach allowed the World Bank Team to support the implementation of the project in close cooperation with DISED. Page 12 of 19 The World Bank Djibouti Household Survey 2017 (P162743) 47. Flexibility and close coordination with the counterpart proved to be key in dealing with administrative delays. The project was launched with a delay, as were the preparation activities for the survey. However, despite this, the fields teams were able to be on the ground by April 2017. To deal with the hot season and Ramadan in May 2017, the sampling strategy was revised so that the data collection was divided into two phases without affecting the achievement of the outcomes. 48. There are unexpected pitfalls that can occur when a new technology is adapted. . In the case of data collection using CAPI, two lessons emerge from this project: i) enhancing feedback loops of data quality that allow correcting for data collection more quickly (in a week or less); and ii) assuring that the data collected is checked at several stages of the process (for instance, after collection, after transmission to central repository and after updating stored data). 49. Recognizing that DISED staff is typically stretched among several competing demands on their time, an option to help them block enough time to push through the technical-intensive process of developing a new welfare and poverty estimate had to be found. The team hosted a reverse mission of few DISED staff in Washington DC in February 2018. This collaborative approach helped move the analytical process substantially forward, not only due to the time devoted to the data from EDAM 2017, but the ability to host presentations by experts who typically are not able to join team missions but are working on related issues and can bring key lessons to the project. 50. DISED showcased their work on EDAM 2017 in various presentations and workshops, which has improved their saliency and agency within the country. In particular, the team is aware of at least two specific requests from line ministries for new data collection on the water and ICT sectors, and two more from international partners. These requests point to the trust that DISED has as a reliable office that can produce high-quality data for policy use. . Page 13 of 19 The World Bank Djibouti Household Survey 2017 (P162743) ANNEX 1. RESULTS FRAMEWORK AND KEY OUTPUTS A. RESULTS INDICATORS A.1 PDO Indicators Objective/Outcome: Completion and Dissemination of EDAM 2017 Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Successful completion of the Yes/No N Y Y Y 2017 Djibouti Household Survey 12-Feb-2017 01-Jan-2018 01-Jan-2018 01-Jan-2018 Comments (achievements against targets): Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Publication of the Welfare and Yes/No N Y Y Y Poverty Analysis Report 12-Feb-2017 30-Mar-2018 29-Jun-2018 29-Jun-2018 Comments (achievements against targets): Page 14 of 19 The World Bank Djibouti Household Survey 2017 (P162743) Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Archiving the 2017 Djibouti Yes/No N Y Y Y Household Survey in the World Bank Microdata Catalog 12-Feb-2017 30-Mar-2018 30-Mar-2018 30-Mar-2018 Comments (achievements against targets): A.2 Intermediate Results Indicators Component: Djibouti Household Survey 2017 Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Sampling framework and Yes/No N Y Y Y strategy for the EDAM 13-Feb-2017 28-Apr-2017 28-Apr-2017 28-Apr-2017 Comments (achievements against targets): Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Development of the EDAM Yes/No N Y Y Y 2017 questionnaire in a Computer Assisted Personal 13-Feb-2017 28-Apr-2017 28-Apr-2017 28-Apr-2017 Interview format Page 15 of 19 The World Bank Djibouti Household Survey 2017 (P162743) Comments (achievements against targets): Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Training workshops for EDAM Number 0.00 0.00 1.00 2.00 field personnel 13-Feb-2017 28-Apr-2017 28-Apr-2017 07-Nov-2017 Comments (achievements against targets): Unit of Formally Revised Actual Achieved at Indicator Name Baseline Original Target Measure Target Completion Stakeholder Consultation Number 0.00 1.00 1.00 3.00 Meetings with the line ministries and development 13-Feb-2017 28-Apr-2017 28-Apr-2017 25-Jun-2018 partners Comments (achievements against targets): Page 16 of 19 The World Bank Djibouti Household Survey 2017 (P162743) B. ORGANIZATION OF THE ASSESSMENT OF THE PDO Objective/Outcome 1: The objective is to support the DISED to collect, analyze and disseminate data emanating from the Household Survey for Social Indicators (EDAM) 2017. 1. Collection of Household Survey for Social Indicators (EDAM) 2017 Outcome Indicators 2. Analysis of data from Household Survey for Social Indicators (EDAM) 2017 3. Dissemination of EDAM 2017 1. Successful completion of the EDAM 2017 Intermediate Results Indicators 2. Publication of the Welfare and Poverty Analysis Report 3. Archiving of the 2017 EDAM for public access in an online microdata library 1. Update of questionnaire and field instruments in line with current good practices 2. Programming of the questionnaire using CsPro 3. Cartographic work and sampling for EDAM 2017 4. Provision of 3-week training workshop for field personnel in the use of CAPI to administer the questionnaire to respondents 5. Production of datasets for analysis, after quality control Key Outputs by Component 6. Capacity building on welfare and poverty measurement for the production of the (linked to the achievement of the official Djibouti 2017 poverty rates. Other social indicators on education, Objective/Outcome 1) employment etc. also produced. 7. Publication of a joint report with DISED on welfare and poverty on DISED’s website. Dissemination of the report was also supported through a WB blog post. 8. The findings of joint report disseminated through a presentation to stakeholders, line ministries and development partners in June 2018. 9. Archiving of metadata from EDAM 2017 on DISED’s microdata catalog Page 17 of 19 The World Bank Djibouti Household Survey 2017 (P162743) . ANNEX 2. PROJECT COST BY COMPONENT Amount at Approval Actual at Project Percentage of Approval Components (US$M) Closing (US$M) (US$M) Djibouti Household Survey 0 .50 0 Total 0.00 0.50 0.00 Page 18 of 19 The World Bank Djibouti Household Survey 2017 (P162743) ANNEX 3. RECIPIENT, CO-FINANCIER AND OTHER PARTNER/STAKEHOLDER COMMENTS Page 19 of 19